45980 lines
906 KiB
JSON
45980 lines
906 KiB
JSON
{
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 2.9994125709810064,
|
|
"global_step": 7659,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 8.695652173913044e-08,
|
|
"loss": 1.9692,
|
|
"step": 1
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 1.7391304347826088e-07,
|
|
"loss": 1.9342,
|
|
"step": 2
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 2.608695652173913e-07,
|
|
"loss": 2.0081,
|
|
"step": 3
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 3.4782608695652175e-07,
|
|
"loss": 1.921,
|
|
"step": 4
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 4.347826086956522e-07,
|
|
"loss": 1.8265,
|
|
"step": 5
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 5.217391304347826e-07,
|
|
"loss": 1.8762,
|
|
"step": 6
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 6.086956521739131e-07,
|
|
"loss": 1.8263,
|
|
"step": 7
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 6.956521739130435e-07,
|
|
"loss": 1.8854,
|
|
"step": 8
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 7.82608695652174e-07,
|
|
"loss": 1.7811,
|
|
"step": 9
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 8.695652173913044e-07,
|
|
"loss": 1.8129,
|
|
"step": 10
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 9.565217391304349e-07,
|
|
"loss": 1.6965,
|
|
"step": 11
|
|
},
|
|
{
|
|
"epoch": 0.0,
|
|
"learning_rate": 1.0434782608695653e-06,
|
|
"loss": 1.6645,
|
|
"step": 12
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.1304347826086956e-06,
|
|
"loss": 1.6348,
|
|
"step": 13
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.2173913043478262e-06,
|
|
"loss": 1.6074,
|
|
"step": 14
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.3043478260869566e-06,
|
|
"loss": 1.6287,
|
|
"step": 15
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.391304347826087e-06,
|
|
"loss": 1.5391,
|
|
"step": 16
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.4782608695652176e-06,
|
|
"loss": 1.6493,
|
|
"step": 17
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.565217391304348e-06,
|
|
"loss": 1.6162,
|
|
"step": 18
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.6521739130434784e-06,
|
|
"loss": 1.6207,
|
|
"step": 19
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.7391304347826088e-06,
|
|
"loss": 1.6454,
|
|
"step": 20
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.8260869565217394e-06,
|
|
"loss": 1.4043,
|
|
"step": 21
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 1.9130434782608697e-06,
|
|
"loss": 1.5508,
|
|
"step": 22
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.0000000000000003e-06,
|
|
"loss": 1.5376,
|
|
"step": 23
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.0869565217391305e-06,
|
|
"loss": 1.501,
|
|
"step": 24
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.173913043478261e-06,
|
|
"loss": 1.5048,
|
|
"step": 25
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.2608695652173913e-06,
|
|
"loss": 1.5785,
|
|
"step": 26
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.347826086956522e-06,
|
|
"loss": 1.4287,
|
|
"step": 27
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.4347826086956525e-06,
|
|
"loss": 1.5123,
|
|
"step": 28
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.5217391304347826e-06,
|
|
"loss": 1.5019,
|
|
"step": 29
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.6086956521739132e-06,
|
|
"loss": 1.5321,
|
|
"step": 30
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.695652173913044e-06,
|
|
"loss": 1.4512,
|
|
"step": 31
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.782608695652174e-06,
|
|
"loss": 1.4769,
|
|
"step": 32
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.8695652173913046e-06,
|
|
"loss": 1.5001,
|
|
"step": 33
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 2.956521739130435e-06,
|
|
"loss": 1.6008,
|
|
"step": 34
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 3.043478260869566e-06,
|
|
"loss": 1.5097,
|
|
"step": 35
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 3.130434782608696e-06,
|
|
"loss": 1.4475,
|
|
"step": 36
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 3.217391304347826e-06,
|
|
"loss": 1.4221,
|
|
"step": 37
|
|
},
|
|
{
|
|
"epoch": 0.01,
|
|
"learning_rate": 3.3043478260869567e-06,
|
|
"loss": 1.4042,
|
|
"step": 38
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 3.391304347826087e-06,
|
|
"loss": 1.4726,
|
|
"step": 39
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 3.4782608695652175e-06,
|
|
"loss": 1.4858,
|
|
"step": 40
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 3.565217391304348e-06,
|
|
"loss": 1.4601,
|
|
"step": 41
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 3.6521739130434787e-06,
|
|
"loss": 1.5113,
|
|
"step": 42
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 3.739130434782609e-06,
|
|
"loss": 1.4247,
|
|
"step": 43
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 3.8260869565217395e-06,
|
|
"loss": 1.4912,
|
|
"step": 44
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 3.91304347826087e-06,
|
|
"loss": 1.4185,
|
|
"step": 45
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.000000000000001e-06,
|
|
"loss": 1.4073,
|
|
"step": 46
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.086956521739131e-06,
|
|
"loss": 1.5049,
|
|
"step": 47
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.173913043478261e-06,
|
|
"loss": 1.4605,
|
|
"step": 48
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.260869565217392e-06,
|
|
"loss": 1.506,
|
|
"step": 49
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.347826086956522e-06,
|
|
"loss": 1.4675,
|
|
"step": 50
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.434782608695653e-06,
|
|
"loss": 1.4712,
|
|
"step": 51
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.5217391304347826e-06,
|
|
"loss": 1.4624,
|
|
"step": 52
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.608695652173913e-06,
|
|
"loss": 1.4258,
|
|
"step": 53
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.695652173913044e-06,
|
|
"loss": 1.4553,
|
|
"step": 54
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.782608695652174e-06,
|
|
"loss": 1.476,
|
|
"step": 55
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.869565217391305e-06,
|
|
"loss": 1.398,
|
|
"step": 56
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 4.9565217391304355e-06,
|
|
"loss": 1.5027,
|
|
"step": 57
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 5.043478260869565e-06,
|
|
"loss": 1.3629,
|
|
"step": 58
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 5.130434782608697e-06,
|
|
"loss": 1.476,
|
|
"step": 59
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 5.2173913043478265e-06,
|
|
"loss": 1.4943,
|
|
"step": 60
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 5.304347826086957e-06,
|
|
"loss": 1.4311,
|
|
"step": 61
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 5.391304347826088e-06,
|
|
"loss": 1.4914,
|
|
"step": 62
|
|
},
|
|
{
|
|
"epoch": 0.02,
|
|
"learning_rate": 5.478260869565217e-06,
|
|
"loss": 1.4628,
|
|
"step": 63
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 5.565217391304348e-06,
|
|
"loss": 1.4714,
|
|
"step": 64
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 5.652173913043479e-06,
|
|
"loss": 1.4158,
|
|
"step": 65
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 5.739130434782609e-06,
|
|
"loss": 1.4574,
|
|
"step": 66
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 5.826086956521739e-06,
|
|
"loss": 1.3495,
|
|
"step": 67
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 5.91304347826087e-06,
|
|
"loss": 1.417,
|
|
"step": 68
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6e-06,
|
|
"loss": 1.4455,
|
|
"step": 69
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.086956521739132e-06,
|
|
"loss": 1.4599,
|
|
"step": 70
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.173913043478261e-06,
|
|
"loss": 1.3407,
|
|
"step": 71
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.260869565217392e-06,
|
|
"loss": 1.5316,
|
|
"step": 72
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.3478260869565225e-06,
|
|
"loss": 1.4693,
|
|
"step": 73
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.434782608695652e-06,
|
|
"loss": 1.4312,
|
|
"step": 74
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.521739130434783e-06,
|
|
"loss": 1.5118,
|
|
"step": 75
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.6086956521739135e-06,
|
|
"loss": 1.5135,
|
|
"step": 76
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.695652173913044e-06,
|
|
"loss": 1.5334,
|
|
"step": 77
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.782608695652174e-06,
|
|
"loss": 1.5475,
|
|
"step": 78
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.869565217391305e-06,
|
|
"loss": 1.4111,
|
|
"step": 79
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 6.956521739130435e-06,
|
|
"loss": 1.4256,
|
|
"step": 80
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.0434782608695665e-06,
|
|
"loss": 1.3907,
|
|
"step": 81
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.130434782608696e-06,
|
|
"loss": 1.4366,
|
|
"step": 82
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.217391304347827e-06,
|
|
"loss": 1.3315,
|
|
"step": 83
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.304347826086957e-06,
|
|
"loss": 1.3649,
|
|
"step": 84
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.391304347826087e-06,
|
|
"loss": 1.364,
|
|
"step": 85
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.478260869565218e-06,
|
|
"loss": 1.478,
|
|
"step": 86
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.565217391304348e-06,
|
|
"loss": 1.4379,
|
|
"step": 87
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.652173913043479e-06,
|
|
"loss": 1.3606,
|
|
"step": 88
|
|
},
|
|
{
|
|
"epoch": 0.03,
|
|
"learning_rate": 7.739130434782609e-06,
|
|
"loss": 1.3104,
|
|
"step": 89
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 7.82608695652174e-06,
|
|
"loss": 1.3317,
|
|
"step": 90
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 7.91304347826087e-06,
|
|
"loss": 1.3445,
|
|
"step": 91
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.000000000000001e-06,
|
|
"loss": 1.4153,
|
|
"step": 92
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.086956521739131e-06,
|
|
"loss": 1.3725,
|
|
"step": 93
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.173913043478263e-06,
|
|
"loss": 1.4074,
|
|
"step": 94
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.260869565217392e-06,
|
|
"loss": 1.4869,
|
|
"step": 95
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.347826086956522e-06,
|
|
"loss": 1.3383,
|
|
"step": 96
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.434782608695653e-06,
|
|
"loss": 1.2853,
|
|
"step": 97
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.521739130434783e-06,
|
|
"loss": 1.3187,
|
|
"step": 98
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.608695652173915e-06,
|
|
"loss": 1.383,
|
|
"step": 99
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.695652173913044e-06,
|
|
"loss": 1.4277,
|
|
"step": 100
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.782608695652174e-06,
|
|
"loss": 1.2761,
|
|
"step": 101
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.869565217391306e-06,
|
|
"loss": 1.3389,
|
|
"step": 102
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 8.956521739130435e-06,
|
|
"loss": 1.3889,
|
|
"step": 103
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.043478260869565e-06,
|
|
"loss": 1.4318,
|
|
"step": 104
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.130434782608697e-06,
|
|
"loss": 1.4817,
|
|
"step": 105
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.217391304347826e-06,
|
|
"loss": 1.4164,
|
|
"step": 106
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.304347826086956e-06,
|
|
"loss": 1.331,
|
|
"step": 107
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.391304347826087e-06,
|
|
"loss": 1.4398,
|
|
"step": 108
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.478260869565217e-06,
|
|
"loss": 1.4444,
|
|
"step": 109
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.565217391304349e-06,
|
|
"loss": 1.4672,
|
|
"step": 110
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.652173913043478e-06,
|
|
"loss": 1.3687,
|
|
"step": 111
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.73913043478261e-06,
|
|
"loss": 1.3822,
|
|
"step": 112
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.82608695652174e-06,
|
|
"loss": 1.3665,
|
|
"step": 113
|
|
},
|
|
{
|
|
"epoch": 0.04,
|
|
"learning_rate": 9.913043478260871e-06,
|
|
"loss": 1.3683,
|
|
"step": 114
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1e-05,
|
|
"loss": 1.3459,
|
|
"step": 115
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.008695652173913e-05,
|
|
"loss": 1.3379,
|
|
"step": 116
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.017391304347826e-05,
|
|
"loss": 1.3883,
|
|
"step": 117
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0260869565217393e-05,
|
|
"loss": 1.4035,
|
|
"step": 118
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0347826086956523e-05,
|
|
"loss": 1.34,
|
|
"step": 119
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0434782608695653e-05,
|
|
"loss": 1.3536,
|
|
"step": 120
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0521739130434783e-05,
|
|
"loss": 1.3223,
|
|
"step": 121
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0608695652173914e-05,
|
|
"loss": 1.3283,
|
|
"step": 122
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0695652173913046e-05,
|
|
"loss": 1.3709,
|
|
"step": 123
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0782608695652175e-05,
|
|
"loss": 1.4307,
|
|
"step": 124
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0869565217391305e-05,
|
|
"loss": 1.4222,
|
|
"step": 125
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.0956521739130435e-05,
|
|
"loss": 1.3426,
|
|
"step": 126
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1043478260869566e-05,
|
|
"loss": 1.3803,
|
|
"step": 127
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1130434782608696e-05,
|
|
"loss": 1.3967,
|
|
"step": 128
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1217391304347827e-05,
|
|
"loss": 1.3996,
|
|
"step": 129
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1304347826086957e-05,
|
|
"loss": 1.3021,
|
|
"step": 130
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1391304347826089e-05,
|
|
"loss": 1.344,
|
|
"step": 131
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1478260869565218e-05,
|
|
"loss": 1.4421,
|
|
"step": 132
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1565217391304348e-05,
|
|
"loss": 1.3349,
|
|
"step": 133
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1652173913043478e-05,
|
|
"loss": 1.3849,
|
|
"step": 134
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.1739130434782611e-05,
|
|
"loss": 1.3892,
|
|
"step": 135
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.182608695652174e-05,
|
|
"loss": 1.3467,
|
|
"step": 136
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.191304347826087e-05,
|
|
"loss": 1.385,
|
|
"step": 137
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.2e-05,
|
|
"loss": 1.3764,
|
|
"step": 138
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.208695652173913e-05,
|
|
"loss": 1.3213,
|
|
"step": 139
|
|
},
|
|
{
|
|
"epoch": 0.05,
|
|
"learning_rate": 1.2173913043478263e-05,
|
|
"loss": 1.3347,
|
|
"step": 140
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2260869565217393e-05,
|
|
"loss": 1.378,
|
|
"step": 141
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2347826086956523e-05,
|
|
"loss": 1.3809,
|
|
"step": 142
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2434782608695652e-05,
|
|
"loss": 1.2931,
|
|
"step": 143
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2521739130434784e-05,
|
|
"loss": 1.4498,
|
|
"step": 144
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2608695652173915e-05,
|
|
"loss": 1.3872,
|
|
"step": 145
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2695652173913045e-05,
|
|
"loss": 1.3818,
|
|
"step": 146
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2782608695652175e-05,
|
|
"loss": 1.42,
|
|
"step": 147
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2869565217391305e-05,
|
|
"loss": 1.3181,
|
|
"step": 148
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.2956521739130436e-05,
|
|
"loss": 1.3977,
|
|
"step": 149
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3043478260869566e-05,
|
|
"loss": 1.3901,
|
|
"step": 150
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3130434782608697e-05,
|
|
"loss": 1.3495,
|
|
"step": 151
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3217391304347827e-05,
|
|
"loss": 1.3808,
|
|
"step": 152
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3304347826086958e-05,
|
|
"loss": 1.3117,
|
|
"step": 153
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3391304347826088e-05,
|
|
"loss": 1.3448,
|
|
"step": 154
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3478260869565218e-05,
|
|
"loss": 1.273,
|
|
"step": 155
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3565217391304348e-05,
|
|
"loss": 1.3701,
|
|
"step": 156
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.3652173913043479e-05,
|
|
"loss": 1.3223,
|
|
"step": 157
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.373913043478261e-05,
|
|
"loss": 1.3174,
|
|
"step": 158
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.382608695652174e-05,
|
|
"loss": 1.3493,
|
|
"step": 159
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.391304347826087e-05,
|
|
"loss": 1.3793,
|
|
"step": 160
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.4e-05,
|
|
"loss": 1.4044,
|
|
"step": 161
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.4086956521739133e-05,
|
|
"loss": 1.3699,
|
|
"step": 162
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.4173913043478263e-05,
|
|
"loss": 1.3449,
|
|
"step": 163
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.4260869565217392e-05,
|
|
"loss": 1.3663,
|
|
"step": 164
|
|
},
|
|
{
|
|
"epoch": 0.06,
|
|
"learning_rate": 1.4347826086956522e-05,
|
|
"loss": 1.3462,
|
|
"step": 165
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.4434782608695654e-05,
|
|
"loss": 1.4223,
|
|
"step": 166
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.4521739130434785e-05,
|
|
"loss": 1.3377,
|
|
"step": 167
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.4608695652173915e-05,
|
|
"loss": 1.3824,
|
|
"step": 168
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.4695652173913045e-05,
|
|
"loss": 1.3529,
|
|
"step": 169
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.4782608695652174e-05,
|
|
"loss": 1.3475,
|
|
"step": 170
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.4869565217391306e-05,
|
|
"loss": 1.4049,
|
|
"step": 171
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.4956521739130436e-05,
|
|
"loss": 1.3665,
|
|
"step": 172
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.5043478260869567e-05,
|
|
"loss": 1.3378,
|
|
"step": 173
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.5130434782608697e-05,
|
|
"loss": 1.3178,
|
|
"step": 174
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.5217391304347828e-05,
|
|
"loss": 1.2825,
|
|
"step": 175
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.5304347826086958e-05,
|
|
"loss": 1.3643,
|
|
"step": 176
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.5391304347826088e-05,
|
|
"loss": 1.2708,
|
|
"step": 177
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.5478260869565217e-05,
|
|
"loss": 1.2544,
|
|
"step": 178
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.5565217391304347e-05,
|
|
"loss": 1.3892,
|
|
"step": 179
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.565217391304348e-05,
|
|
"loss": 1.4187,
|
|
"step": 180
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.573913043478261e-05,
|
|
"loss": 1.2725,
|
|
"step": 181
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.582608695652174e-05,
|
|
"loss": 1.2869,
|
|
"step": 182
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.591304347826087e-05,
|
|
"loss": 1.3597,
|
|
"step": 183
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6000000000000003e-05,
|
|
"loss": 1.3441,
|
|
"step": 184
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6086956521739132e-05,
|
|
"loss": 1.3058,
|
|
"step": 185
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6173913043478262e-05,
|
|
"loss": 1.269,
|
|
"step": 186
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6260869565217392e-05,
|
|
"loss": 1.374,
|
|
"step": 187
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6347826086956525e-05,
|
|
"loss": 1.3341,
|
|
"step": 188
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6434782608695655e-05,
|
|
"loss": 1.3492,
|
|
"step": 189
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6521739130434785e-05,
|
|
"loss": 1.3548,
|
|
"step": 190
|
|
},
|
|
{
|
|
"epoch": 0.07,
|
|
"learning_rate": 1.6608695652173914e-05,
|
|
"loss": 1.3149,
|
|
"step": 191
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.6695652173913044e-05,
|
|
"loss": 1.3061,
|
|
"step": 192
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.6782608695652177e-05,
|
|
"loss": 1.3742,
|
|
"step": 193
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.6869565217391307e-05,
|
|
"loss": 1.2934,
|
|
"step": 194
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.6956521739130437e-05,
|
|
"loss": 1.4055,
|
|
"step": 195
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.7043478260869566e-05,
|
|
"loss": 1.3673,
|
|
"step": 196
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.71304347826087e-05,
|
|
"loss": 1.3152,
|
|
"step": 197
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.721739130434783e-05,
|
|
"loss": 1.3087,
|
|
"step": 198
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.730434782608696e-05,
|
|
"loss": 1.4102,
|
|
"step": 199
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.739130434782609e-05,
|
|
"loss": 1.3258,
|
|
"step": 200
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.747826086956522e-05,
|
|
"loss": 1.3095,
|
|
"step": 201
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.756521739130435e-05,
|
|
"loss": 1.4177,
|
|
"step": 202
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.765217391304348e-05,
|
|
"loss": 1.236,
|
|
"step": 203
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.773913043478261e-05,
|
|
"loss": 1.3322,
|
|
"step": 204
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.782608695652174e-05,
|
|
"loss": 1.4095,
|
|
"step": 205
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.791304347826087e-05,
|
|
"loss": 1.3399,
|
|
"step": 206
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8e-05,
|
|
"loss": 1.3913,
|
|
"step": 207
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.808695652173913e-05,
|
|
"loss": 1.2687,
|
|
"step": 208
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8173913043478263e-05,
|
|
"loss": 1.3884,
|
|
"step": 209
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8260869565217393e-05,
|
|
"loss": 1.3952,
|
|
"step": 210
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8347826086956523e-05,
|
|
"loss": 1.3261,
|
|
"step": 211
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8434782608695653e-05,
|
|
"loss": 1.358,
|
|
"step": 212
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8521739130434782e-05,
|
|
"loss": 1.3863,
|
|
"step": 213
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8608695652173912e-05,
|
|
"loss": 1.4177,
|
|
"step": 214
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8695652173913045e-05,
|
|
"loss": 1.3593,
|
|
"step": 215
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8782608695652175e-05,
|
|
"loss": 1.2539,
|
|
"step": 216
|
|
},
|
|
{
|
|
"epoch": 0.08,
|
|
"learning_rate": 1.8869565217391305e-05,
|
|
"loss": 1.3564,
|
|
"step": 217
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.8956521739130434e-05,
|
|
"loss": 1.3563,
|
|
"step": 218
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9043478260869568e-05,
|
|
"loss": 1.3583,
|
|
"step": 219
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9130434782608697e-05,
|
|
"loss": 1.3808,
|
|
"step": 220
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9217391304347827e-05,
|
|
"loss": 1.3213,
|
|
"step": 221
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9304347826086957e-05,
|
|
"loss": 1.343,
|
|
"step": 222
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9391304347826087e-05,
|
|
"loss": 1.3407,
|
|
"step": 223
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.947826086956522e-05,
|
|
"loss": 1.3097,
|
|
"step": 224
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.956521739130435e-05,
|
|
"loss": 1.3223,
|
|
"step": 225
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.965217391304348e-05,
|
|
"loss": 1.357,
|
|
"step": 226
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.973913043478261e-05,
|
|
"loss": 1.2906,
|
|
"step": 227
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9826086956521742e-05,
|
|
"loss": 1.3784,
|
|
"step": 228
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9913043478260872e-05,
|
|
"loss": 1.3136,
|
|
"step": 229
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 2e-05,
|
|
"loss": 1.2967,
|
|
"step": 230
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.999999910585279e-05,
|
|
"loss": 1.3025,
|
|
"step": 231
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999996423411313e-05,
|
|
"loss": 1.3544,
|
|
"step": 232
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999991952676056e-05,
|
|
"loss": 1.2567,
|
|
"step": 233
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999985693647814e-05,
|
|
"loss": 1.3489,
|
|
"step": 234
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999977646327706e-05,
|
|
"loss": 1.3038,
|
|
"step": 235
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.999996781071717e-05,
|
|
"loss": 1.3517,
|
|
"step": 236
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999956186817973e-05,
|
|
"loss": 1.3423,
|
|
"step": 237
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999942774632182e-05,
|
|
"loss": 1.3207,
|
|
"step": 238
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999927574162207e-05,
|
|
"loss": 1.2518,
|
|
"step": 239
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999910585410757e-05,
|
|
"loss": 1.4058,
|
|
"step": 240
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999891808380872e-05,
|
|
"loss": 1.321,
|
|
"step": 241
|
|
},
|
|
{
|
|
"epoch": 0.09,
|
|
"learning_rate": 1.9999871243075913e-05,
|
|
"loss": 1.249,
|
|
"step": 242
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999848889499556e-05,
|
|
"loss": 1.2701,
|
|
"step": 243
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.99998247476558e-05,
|
|
"loss": 1.3887,
|
|
"step": 244
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999798817548963e-05,
|
|
"loss": 1.2965,
|
|
"step": 245
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999771099183677e-05,
|
|
"loss": 1.3736,
|
|
"step": 246
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999741592564903e-05,
|
|
"loss": 1.3189,
|
|
"step": 247
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999710297697917e-05,
|
|
"loss": 1.3552,
|
|
"step": 248
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999677214588314e-05,
|
|
"loss": 1.2019,
|
|
"step": 249
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.999964234324201e-05,
|
|
"loss": 1.2501,
|
|
"step": 250
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999605683665243e-05,
|
|
"loss": 1.1609,
|
|
"step": 251
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.999956723586457e-05,
|
|
"loss": 1.3158,
|
|
"step": 252
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999526999846864e-05,
|
|
"loss": 1.2813,
|
|
"step": 253
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.999948497561932e-05,
|
|
"loss": 1.4197,
|
|
"step": 254
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999441163189458e-05,
|
|
"loss": 1.2934,
|
|
"step": 255
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999395562565104e-05,
|
|
"loss": 1.3249,
|
|
"step": 256
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.999934817375442e-05,
|
|
"loss": 1.3107,
|
|
"step": 257
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.999929899676588e-05,
|
|
"loss": 1.32,
|
|
"step": 258
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999248031608276e-05,
|
|
"loss": 1.3486,
|
|
"step": 259
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999195278290723e-05,
|
|
"loss": 1.2936,
|
|
"step": 260
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999140736822654e-05,
|
|
"loss": 1.3116,
|
|
"step": 261
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999084407213823e-05,
|
|
"loss": 1.3754,
|
|
"step": 262
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9999026289474303e-05,
|
|
"loss": 1.3127,
|
|
"step": 263
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.999896638361449e-05,
|
|
"loss": 1.3244,
|
|
"step": 264
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9998904689645095e-05,
|
|
"loss": 1.3706,
|
|
"step": 265
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9998841207577148e-05,
|
|
"loss": 1.2823,
|
|
"step": 266
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9998775937422002e-05,
|
|
"loss": 1.2799,
|
|
"step": 267
|
|
},
|
|
{
|
|
"epoch": 0.1,
|
|
"learning_rate": 1.9998708879191336e-05,
|
|
"loss": 1.3394,
|
|
"step": 268
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9998640032897135e-05,
|
|
"loss": 1.33,
|
|
"step": 269
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9998569398551713e-05,
|
|
"loss": 1.2995,
|
|
"step": 270
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.99984969761677e-05,
|
|
"loss": 1.3282,
|
|
"step": 271
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999842276575805e-05,
|
|
"loss": 1.2941,
|
|
"step": 272
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9998346767336034e-05,
|
|
"loss": 1.3228,
|
|
"step": 273
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999826898091524e-05,
|
|
"loss": 1.3309,
|
|
"step": 274
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999818940650958e-05,
|
|
"loss": 1.4293,
|
|
"step": 275
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9998108044133286e-05,
|
|
"loss": 1.2681,
|
|
"step": 276
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9998024893800902e-05,
|
|
"loss": 1.3695,
|
|
"step": 277
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9997939955527304e-05,
|
|
"loss": 1.3243,
|
|
"step": 278
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999785322932768e-05,
|
|
"loss": 1.3277,
|
|
"step": 279
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999776471521754e-05,
|
|
"loss": 1.2628,
|
|
"step": 280
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999767441321271e-05,
|
|
"loss": 1.3438,
|
|
"step": 281
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999758232332934e-05,
|
|
"loss": 1.2995,
|
|
"step": 282
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9997488445583898e-05,
|
|
"loss": 1.332,
|
|
"step": 283
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9997392779993174e-05,
|
|
"loss": 1.3607,
|
|
"step": 284
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999729532657427e-05,
|
|
"loss": 1.2977,
|
|
"step": 285
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9997196085344622e-05,
|
|
"loss": 1.2289,
|
|
"step": 286
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9997095056321974e-05,
|
|
"loss": 1.288,
|
|
"step": 287
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999699223952439e-05,
|
|
"loss": 1.3088,
|
|
"step": 288
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9996887634970262e-05,
|
|
"loss": 1.3526,
|
|
"step": 289
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.999678124267829e-05,
|
|
"loss": 1.3078,
|
|
"step": 290
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9996673062667503e-05,
|
|
"loss": 1.3392,
|
|
"step": 291
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9996563094957245e-05,
|
|
"loss": 1.3578,
|
|
"step": 292
|
|
},
|
|
{
|
|
"epoch": 0.11,
|
|
"learning_rate": 1.9996451339567187e-05,
|
|
"loss": 1.28,
|
|
"step": 293
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9996337796517313e-05,
|
|
"loss": 1.3211,
|
|
"step": 294
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9996222465827924e-05,
|
|
"loss": 1.2399,
|
|
"step": 295
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9996105347519644e-05,
|
|
"loss": 1.3059,
|
|
"step": 296
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999598644161342e-05,
|
|
"loss": 1.2779,
|
|
"step": 297
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9995865748130518e-05,
|
|
"loss": 1.2721,
|
|
"step": 298
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9995743267092514e-05,
|
|
"loss": 1.3042,
|
|
"step": 299
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999561899852132e-05,
|
|
"loss": 1.3611,
|
|
"step": 300
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999549294243915e-05,
|
|
"loss": 1.3223,
|
|
"step": 301
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9995365098868555e-05,
|
|
"loss": 1.3204,
|
|
"step": 302
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999523546783239e-05,
|
|
"loss": 1.2652,
|
|
"step": 303
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9995104049353847e-05,
|
|
"loss": 1.3096,
|
|
"step": 304
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9994970843456417e-05,
|
|
"loss": 1.2849,
|
|
"step": 305
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9994835850163926e-05,
|
|
"loss": 1.2669,
|
|
"step": 306
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999469906950051e-05,
|
|
"loss": 1.1981,
|
|
"step": 307
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9994560501490638e-05,
|
|
"loss": 1.3265,
|
|
"step": 308
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9994420146159083e-05,
|
|
"loss": 1.2814,
|
|
"step": 309
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9994278003530944e-05,
|
|
"loss": 1.3665,
|
|
"step": 310
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9994134073631648e-05,
|
|
"loss": 1.288,
|
|
"step": 311
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9993988356486927e-05,
|
|
"loss": 1.2457,
|
|
"step": 312
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999384085212284e-05,
|
|
"loss": 1.3248,
|
|
"step": 313
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999369156056577e-05,
|
|
"loss": 1.2529,
|
|
"step": 314
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999354048184241e-05,
|
|
"loss": 1.2985,
|
|
"step": 315
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.999338761597978e-05,
|
|
"loss": 1.3134,
|
|
"step": 316
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9993232963005213e-05,
|
|
"loss": 1.3167,
|
|
"step": 317
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9993076522946372e-05,
|
|
"loss": 1.2507,
|
|
"step": 318
|
|
},
|
|
{
|
|
"epoch": 0.12,
|
|
"learning_rate": 1.9992918295831226e-05,
|
|
"loss": 1.2985,
|
|
"step": 319
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9992758281688075e-05,
|
|
"loss": 1.3049,
|
|
"step": 320
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9992596480545533e-05,
|
|
"loss": 1.2279,
|
|
"step": 321
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9992432892432534e-05,
|
|
"loss": 1.3575,
|
|
"step": 322
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9992267517378337e-05,
|
|
"loss": 1.2743,
|
|
"step": 323
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.999210035541251e-05,
|
|
"loss": 1.2904,
|
|
"step": 324
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9991931406564944e-05,
|
|
"loss": 1.3192,
|
|
"step": 325
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9991760670865864e-05,
|
|
"loss": 1.2828,
|
|
"step": 326
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.999158814834579e-05,
|
|
"loss": 1.2637,
|
|
"step": 327
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9991413839035583e-05,
|
|
"loss": 1.3582,
|
|
"step": 328
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9991237742966412e-05,
|
|
"loss": 1.379,
|
|
"step": 329
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9991059860169764e-05,
|
|
"loss": 1.2473,
|
|
"step": 330
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9990880190677454e-05,
|
|
"loss": 1.3241,
|
|
"step": 331
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9990698734521614e-05,
|
|
"loss": 1.2904,
|
|
"step": 332
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9990515491734687e-05,
|
|
"loss": 1.3474,
|
|
"step": 333
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9990330462349447e-05,
|
|
"loss": 1.2204,
|
|
"step": 334
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9990143646398982e-05,
|
|
"loss": 1.3197,
|
|
"step": 335
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9989955043916704e-05,
|
|
"loss": 1.3118,
|
|
"step": 336
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.998976465493633e-05,
|
|
"loss": 1.3122,
|
|
"step": 337
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9989572479491917e-05,
|
|
"loss": 1.3493,
|
|
"step": 338
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9989378517617828e-05,
|
|
"loss": 1.3371,
|
|
"step": 339
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.998918276934875e-05,
|
|
"loss": 1.3068,
|
|
"step": 340
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9988985234719687e-05,
|
|
"loss": 1.3389,
|
|
"step": 341
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9988785913765967e-05,
|
|
"loss": 1.1777,
|
|
"step": 342
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.9988584806523227e-05,
|
|
"loss": 1.3442,
|
|
"step": 343
|
|
},
|
|
{
|
|
"epoch": 0.13,
|
|
"learning_rate": 1.998838191302744e-05,
|
|
"loss": 1.32,
|
|
"step": 344
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998817723331489e-05,
|
|
"loss": 1.2696,
|
|
"step": 345
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998797076742217e-05,
|
|
"loss": 1.2616,
|
|
"step": 346
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998776251538621e-05,
|
|
"loss": 1.3483,
|
|
"step": 347
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998755247724425e-05,
|
|
"loss": 1.2835,
|
|
"step": 348
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9987340653033847e-05,
|
|
"loss": 1.2421,
|
|
"step": 349
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9987127042792888e-05,
|
|
"loss": 1.3002,
|
|
"step": 350
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998691164655957e-05,
|
|
"loss": 1.3304,
|
|
"step": 351
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998669446437241e-05,
|
|
"loss": 1.3411,
|
|
"step": 352
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9986475496270245e-05,
|
|
"loss": 1.219,
|
|
"step": 353
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998625474229224e-05,
|
|
"loss": 1.2371,
|
|
"step": 354
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998603220247787e-05,
|
|
"loss": 1.2929,
|
|
"step": 355
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998580787686693e-05,
|
|
"loss": 1.2961,
|
|
"step": 356
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998558176549953e-05,
|
|
"loss": 1.2641,
|
|
"step": 357
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9985353868416116e-05,
|
|
"loss": 1.3009,
|
|
"step": 358
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998512418565744e-05,
|
|
"loss": 1.2714,
|
|
"step": 359
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9984892717264573e-05,
|
|
"loss": 1.2607,
|
|
"step": 360
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9984659463278907e-05,
|
|
"loss": 1.2706,
|
|
"step": 361
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998442442374216e-05,
|
|
"loss": 1.2328,
|
|
"step": 362
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9984187598696364e-05,
|
|
"loss": 1.2996,
|
|
"step": 363
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9983948988183864e-05,
|
|
"loss": 1.3323,
|
|
"step": 364
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9983708592247332e-05,
|
|
"loss": 1.2726,
|
|
"step": 365
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9983466410929764e-05,
|
|
"loss": 1.3515,
|
|
"step": 366
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9983222444274465e-05,
|
|
"loss": 1.3568,
|
|
"step": 367
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998297669232506e-05,
|
|
"loss": 1.2347,
|
|
"step": 368
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.9982729155125506e-05,
|
|
"loss": 1.2818,
|
|
"step": 369
|
|
},
|
|
{
|
|
"epoch": 0.14,
|
|
"learning_rate": 1.998247983272006e-05,
|
|
"loss": 1.3248,
|
|
"step": 370
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9982228725153315e-05,
|
|
"loss": 1.2622,
|
|
"step": 371
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9981975832470172e-05,
|
|
"loss": 1.3256,
|
|
"step": 372
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.998172115471586e-05,
|
|
"loss": 1.2842,
|
|
"step": 373
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9981464691935915e-05,
|
|
"loss": 1.2826,
|
|
"step": 374
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.998120644417621e-05,
|
|
"loss": 1.3457,
|
|
"step": 375
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9980946411482923e-05,
|
|
"loss": 1.2702,
|
|
"step": 376
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9980684593902555e-05,
|
|
"loss": 1.3298,
|
|
"step": 377
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.998042099148193e-05,
|
|
"loss": 1.2733,
|
|
"step": 378
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9980155604268177e-05,
|
|
"loss": 1.2484,
|
|
"step": 379
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9979888432308766e-05,
|
|
"loss": 1.2581,
|
|
"step": 380
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9979619475651476e-05,
|
|
"loss": 1.2636,
|
|
"step": 381
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.99793487343444e-05,
|
|
"loss": 1.2436,
|
|
"step": 382
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.997907620843595e-05,
|
|
"loss": 1.2909,
|
|
"step": 383
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9978801897974873e-05,
|
|
"loss": 1.2654,
|
|
"step": 384
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9978525803010213e-05,
|
|
"loss": 1.2872,
|
|
"step": 385
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.997824792359135e-05,
|
|
"loss": 1.3196,
|
|
"step": 386
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9977968259767976e-05,
|
|
"loss": 1.3127,
|
|
"step": 387
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.99776868115901e-05,
|
|
"loss": 1.2515,
|
|
"step": 388
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.997740357910806e-05,
|
|
"loss": 1.2444,
|
|
"step": 389
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9977118562372497e-05,
|
|
"loss": 1.3072,
|
|
"step": 390
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.997683176143439e-05,
|
|
"loss": 1.193,
|
|
"step": 391
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.997654317634502e-05,
|
|
"loss": 1.1984,
|
|
"step": 392
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.9976252807155997e-05,
|
|
"loss": 1.2833,
|
|
"step": 393
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.997596065391925e-05,
|
|
"loss": 1.2416,
|
|
"step": 394
|
|
},
|
|
{
|
|
"epoch": 0.15,
|
|
"learning_rate": 1.997566671668702e-05,
|
|
"loss": 1.2436,
|
|
"step": 395
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9975370995511877e-05,
|
|
"loss": 1.236,
|
|
"step": 396
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9975073490446704e-05,
|
|
"loss": 1.2347,
|
|
"step": 397
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9974774201544696e-05,
|
|
"loss": 1.2636,
|
|
"step": 398
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9974473128859383e-05,
|
|
"loss": 1.3049,
|
|
"step": 399
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9974170272444604e-05,
|
|
"loss": 1.2662,
|
|
"step": 400
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9973865632354516e-05,
|
|
"loss": 1.2366,
|
|
"step": 401
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9973559208643598e-05,
|
|
"loss": 1.2795,
|
|
"step": 402
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.997325100136665e-05,
|
|
"loss": 1.236,
|
|
"step": 403
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.997294101057879e-05,
|
|
"loss": 1.3734,
|
|
"step": 404
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9972629236335445e-05,
|
|
"loss": 1.273,
|
|
"step": 405
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.997231567869238e-05,
|
|
"loss": 1.2794,
|
|
"step": 406
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.997200033770566e-05,
|
|
"loss": 1.2686,
|
|
"step": 407
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9971683213431687e-05,
|
|
"loss": 1.3156,
|
|
"step": 408
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.997136430592716e-05,
|
|
"loss": 1.3333,
|
|
"step": 409
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9971043615249113e-05,
|
|
"loss": 1.3581,
|
|
"step": 410
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9970721141454905e-05,
|
|
"loss": 1.2716,
|
|
"step": 411
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.997039688460219e-05,
|
|
"loss": 1.2966,
|
|
"step": 412
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.997007084474896e-05,
|
|
"loss": 1.2573,
|
|
"step": 413
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.996974302195352e-05,
|
|
"loss": 1.339,
|
|
"step": 414
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9969413416274496e-05,
|
|
"loss": 1.2206,
|
|
"step": 415
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.996908202777083e-05,
|
|
"loss": 1.3241,
|
|
"step": 416
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.996874885650179e-05,
|
|
"loss": 1.2714,
|
|
"step": 417
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9968413902526947e-05,
|
|
"loss": 1.2768,
|
|
"step": 418
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9968077165906205e-05,
|
|
"loss": 1.3866,
|
|
"step": 419
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.996773864669978e-05,
|
|
"loss": 1.3672,
|
|
"step": 420
|
|
},
|
|
{
|
|
"epoch": 0.16,
|
|
"learning_rate": 1.9967398344968217e-05,
|
|
"loss": 1.4013,
|
|
"step": 421
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9967056260772364e-05,
|
|
"loss": 1.2759,
|
|
"step": 422
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9966712394173398e-05,
|
|
"loss": 1.2257,
|
|
"step": 423
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9966366745232814e-05,
|
|
"loss": 1.2825,
|
|
"step": 424
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.996601931401242e-05,
|
|
"loss": 1.3091,
|
|
"step": 425
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9965670100574353e-05,
|
|
"loss": 1.3071,
|
|
"step": 426
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9965319104981056e-05,
|
|
"loss": 1.2724,
|
|
"step": 427
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9964966327295303e-05,
|
|
"loss": 1.3367,
|
|
"step": 428
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.996461176758018e-05,
|
|
"loss": 1.3034,
|
|
"step": 429
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.996425542589909e-05,
|
|
"loss": 1.3113,
|
|
"step": 430
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.996389730231576e-05,
|
|
"loss": 1.3502,
|
|
"step": 431
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9963537396894232e-05,
|
|
"loss": 1.2882,
|
|
"step": 432
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9963175709698864e-05,
|
|
"loss": 1.2982,
|
|
"step": 433
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9962812240794344e-05,
|
|
"loss": 1.2233,
|
|
"step": 434
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9962446990245666e-05,
|
|
"loss": 1.2863,
|
|
"step": 435
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.996207995811815e-05,
|
|
"loss": 1.1289,
|
|
"step": 436
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.996171114447743e-05,
|
|
"loss": 1.2994,
|
|
"step": 437
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.996134054938946e-05,
|
|
"loss": 1.368,
|
|
"step": 438
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9960968172920516e-05,
|
|
"loss": 1.3182,
|
|
"step": 439
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9960594015137187e-05,
|
|
"loss": 1.3237,
|
|
"step": 440
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9960218076106386e-05,
|
|
"loss": 1.308,
|
|
"step": 441
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.995984035589534e-05,
|
|
"loss": 1.2681,
|
|
"step": 442
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9959460854571603e-05,
|
|
"loss": 1.2313,
|
|
"step": 443
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.9959079572203027e-05,
|
|
"loss": 1.2622,
|
|
"step": 444
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.995869650885781e-05,
|
|
"loss": 1.2936,
|
|
"step": 445
|
|
},
|
|
{
|
|
"epoch": 0.17,
|
|
"learning_rate": 1.995831166460445e-05,
|
|
"loss": 1.2178,
|
|
"step": 446
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.995792503951177e-05,
|
|
"loss": 1.3215,
|
|
"step": 447
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9957536633648904e-05,
|
|
"loss": 1.2995,
|
|
"step": 448
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9957146447085316e-05,
|
|
"loss": 1.2041,
|
|
"step": 449
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9956754479890785e-05,
|
|
"loss": 1.2951,
|
|
"step": 450
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.99563607321354e-05,
|
|
"loss": 1.3182,
|
|
"step": 451
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.995596520388958e-05,
|
|
"loss": 1.2324,
|
|
"step": 452
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9955567895224052e-05,
|
|
"loss": 1.1744,
|
|
"step": 453
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9955168806209866e-05,
|
|
"loss": 1.2893,
|
|
"step": 454
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.99547679369184e-05,
|
|
"loss": 1.3059,
|
|
"step": 455
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9954365287421336e-05,
|
|
"loss": 1.2827,
|
|
"step": 456
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9953960857790675e-05,
|
|
"loss": 1.2659,
|
|
"step": 457
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9953554648098748e-05,
|
|
"loss": 1.2722,
|
|
"step": 458
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.995314665841819e-05,
|
|
"loss": 1.2601,
|
|
"step": 459
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.995273688882197e-05,
|
|
"loss": 1.2184,
|
|
"step": 460
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.995232533938336e-05,
|
|
"loss": 1.3312,
|
|
"step": 461
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9951912010175963e-05,
|
|
"loss": 1.2395,
|
|
"step": 462
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9951496901273687e-05,
|
|
"loss": 1.2166,
|
|
"step": 463
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.995108001275077e-05,
|
|
"loss": 1.3446,
|
|
"step": 464
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9950661344681767e-05,
|
|
"loss": 1.2949,
|
|
"step": 465
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9950240897141543e-05,
|
|
"loss": 1.2045,
|
|
"step": 466
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.994981867020529e-05,
|
|
"loss": 1.277,
|
|
"step": 467
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.994939466394851e-05,
|
|
"loss": 1.1556,
|
|
"step": 468
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9948968878447035e-05,
|
|
"loss": 1.3327,
|
|
"step": 469
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9948541313777002e-05,
|
|
"loss": 1.2002,
|
|
"step": 470
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.994811197001487e-05,
|
|
"loss": 1.2592,
|
|
"step": 471
|
|
},
|
|
{
|
|
"epoch": 0.18,
|
|
"learning_rate": 1.9947680847237425e-05,
|
|
"loss": 1.2298,
|
|
"step": 472
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9947247945521764e-05,
|
|
"loss": 1.2332,
|
|
"step": 473
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9946813264945298e-05,
|
|
"loss": 1.2608,
|
|
"step": 474
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9946376805585762e-05,
|
|
"loss": 1.2367,
|
|
"step": 475
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.994593856752121e-05,
|
|
"loss": 1.2995,
|
|
"step": 476
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.994549855083001e-05,
|
|
"loss": 1.2546,
|
|
"step": 477
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.994505675559085e-05,
|
|
"loss": 1.2851,
|
|
"step": 478
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.994461318188274e-05,
|
|
"loss": 1.2499,
|
|
"step": 479
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9944167829784996e-05,
|
|
"loss": 1.2547,
|
|
"step": 480
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9943720699377268e-05,
|
|
"loss": 1.2574,
|
|
"step": 481
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.994327179073951e-05,
|
|
"loss": 1.1927,
|
|
"step": 482
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9942821103952002e-05,
|
|
"loss": 1.2795,
|
|
"step": 483
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9942368639095344e-05,
|
|
"loss": 1.253,
|
|
"step": 484
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9941914396250447e-05,
|
|
"loss": 1.2652,
|
|
"step": 485
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9941458375498542e-05,
|
|
"loss": 1.2359,
|
|
"step": 486
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9941000576921175e-05,
|
|
"loss": 1.3256,
|
|
"step": 487
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9940541000600223e-05,
|
|
"loss": 1.2361,
|
|
"step": 488
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.994007964661787e-05,
|
|
"loss": 1.268,
|
|
"step": 489
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9939616515056614e-05,
|
|
"loss": 1.244,
|
|
"step": 490
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.993915160599928e-05,
|
|
"loss": 1.3335,
|
|
"step": 491
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.993868491952901e-05,
|
|
"loss": 1.2387,
|
|
"step": 492
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.993821645572925e-05,
|
|
"loss": 1.254,
|
|
"step": 493
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9937746214683793e-05,
|
|
"loss": 1.2573,
|
|
"step": 494
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9937274196476717e-05,
|
|
"loss": 1.2576,
|
|
"step": 495
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.993680040119244e-05,
|
|
"loss": 1.2111,
|
|
"step": 496
|
|
},
|
|
{
|
|
"epoch": 0.19,
|
|
"learning_rate": 1.9936324828915688e-05,
|
|
"loss": 1.2812,
|
|
"step": 497
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.993584747973151e-05,
|
|
"loss": 1.3009,
|
|
"step": 498
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9935368353725266e-05,
|
|
"loss": 1.2968,
|
|
"step": 499
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9934887450982637e-05,
|
|
"loss": 1.3402,
|
|
"step": 500
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.993440477158963e-05,
|
|
"loss": 1.2149,
|
|
"step": 501
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9933920315632557e-05,
|
|
"loss": 1.3056,
|
|
"step": 502
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9933434083198057e-05,
|
|
"loss": 1.144,
|
|
"step": 503
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9932946074373073e-05,
|
|
"loss": 1.2482,
|
|
"step": 504
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9932456289244888e-05,
|
|
"loss": 1.2402,
|
|
"step": 505
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.993196472790108e-05,
|
|
"loss": 1.2704,
|
|
"step": 506
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.993147139042956e-05,
|
|
"loss": 1.2337,
|
|
"step": 507
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.993097627691855e-05,
|
|
"loss": 1.2656,
|
|
"step": 508
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9930479387456588e-05,
|
|
"loss": 1.2648,
|
|
"step": 509
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9929980722132538e-05,
|
|
"loss": 1.2945,
|
|
"step": 510
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9929480281035572e-05,
|
|
"loss": 1.2551,
|
|
"step": 511
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9928978064255186e-05,
|
|
"loss": 1.2684,
|
|
"step": 512
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.992847407188119e-05,
|
|
"loss": 1.2613,
|
|
"step": 513
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.992796830400371e-05,
|
|
"loss": 1.3767,
|
|
"step": 514
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9927460760713198e-05,
|
|
"loss": 1.2976,
|
|
"step": 515
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9926951442100414e-05,
|
|
"loss": 1.3562,
|
|
"step": 516
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9926440348256436e-05,
|
|
"loss": 1.3285,
|
|
"step": 517
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9925927479272667e-05,
|
|
"loss": 1.3297,
|
|
"step": 518
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9925412835240826e-05,
|
|
"loss": 1.2296,
|
|
"step": 519
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9924896416252938e-05,
|
|
"loss": 1.2151,
|
|
"step": 520
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9924378222401363e-05,
|
|
"loss": 1.2871,
|
|
"step": 521
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9923858253778765e-05,
|
|
"loss": 1.2431,
|
|
"step": 522
|
|
},
|
|
{
|
|
"epoch": 0.2,
|
|
"learning_rate": 1.9923336510478126e-05,
|
|
"loss": 1.3285,
|
|
"step": 523
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.992281299259276e-05,
|
|
"loss": 1.3267,
|
|
"step": 524
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.992228770021627e-05,
|
|
"loss": 1.277,
|
|
"step": 525
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9921760633442612e-05,
|
|
"loss": 1.3288,
|
|
"step": 526
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.992123179236603e-05,
|
|
"loss": 1.2247,
|
|
"step": 527
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9920701177081098e-05,
|
|
"loss": 1.2638,
|
|
"step": 528
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.992016878768271e-05,
|
|
"loss": 1.2906,
|
|
"step": 529
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.991963462426607e-05,
|
|
"loss": 1.2231,
|
|
"step": 530
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.99190986869267e-05,
|
|
"loss": 1.2697,
|
|
"step": 531
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9918560975760444e-05,
|
|
"loss": 1.2607,
|
|
"step": 532
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.991802149086346e-05,
|
|
"loss": 1.3566,
|
|
"step": 533
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9917480232332226e-05,
|
|
"loss": 1.2741,
|
|
"step": 534
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9916937200263526e-05,
|
|
"loss": 1.2257,
|
|
"step": 535
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9916392394754483e-05,
|
|
"loss": 1.3138,
|
|
"step": 536
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9915845815902518e-05,
|
|
"loss": 1.284,
|
|
"step": 537
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9915297463805373e-05,
|
|
"loss": 1.2344,
|
|
"step": 538
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9914747338561115e-05,
|
|
"loss": 1.3639,
|
|
"step": 539
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.991419544026812e-05,
|
|
"loss": 1.2909,
|
|
"step": 540
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.991364176902508e-05,
|
|
"loss": 1.2843,
|
|
"step": 541
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.991308632493101e-05,
|
|
"loss": 1.3351,
|
|
"step": 542
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9912529108085245e-05,
|
|
"loss": 1.2641,
|
|
"step": 543
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.991197011858742e-05,
|
|
"loss": 1.1984,
|
|
"step": 544
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9911409356537518e-05,
|
|
"loss": 1.1801,
|
|
"step": 545
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9910846822035797e-05,
|
|
"loss": 1.305,
|
|
"step": 546
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9910282515182873e-05,
|
|
"loss": 1.29,
|
|
"step": 547
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.990971643607965e-05,
|
|
"loss": 1.3234,
|
|
"step": 548
|
|
},
|
|
{
|
|
"epoch": 0.21,
|
|
"learning_rate": 1.9909148584827366e-05,
|
|
"loss": 1.2457,
|
|
"step": 549
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9908578961527564e-05,
|
|
"loss": 1.2711,
|
|
"step": 550
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9908007566282114e-05,
|
|
"loss": 1.365,
|
|
"step": 551
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9907434399193196e-05,
|
|
"loss": 1.2785,
|
|
"step": 552
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9906859460363307e-05,
|
|
"loss": 1.2249,
|
|
"step": 553
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.990628274989527e-05,
|
|
"loss": 1.2636,
|
|
"step": 554
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.990570426789221e-05,
|
|
"loss": 1.2979,
|
|
"step": 555
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.990512401445758e-05,
|
|
"loss": 1.3222,
|
|
"step": 556
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.990454198969515e-05,
|
|
"loss": 1.2136,
|
|
"step": 557
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9903958193708998e-05,
|
|
"loss": 1.2813,
|
|
"step": 558
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9903372626603522e-05,
|
|
"loss": 1.2671,
|
|
"step": 559
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9902785288483444e-05,
|
|
"loss": 1.2446,
|
|
"step": 560
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9902196179453796e-05,
|
|
"loss": 1.3042,
|
|
"step": 561
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9901605299619927e-05,
|
|
"loss": 1.2542,
|
|
"step": 562
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9901012649087505e-05,
|
|
"loss": 1.4078,
|
|
"step": 563
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9900418227962512e-05,
|
|
"loss": 1.3082,
|
|
"step": 564
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9899822036351245e-05,
|
|
"loss": 1.1579,
|
|
"step": 565
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9899224074360328e-05,
|
|
"loss": 1.2266,
|
|
"step": 566
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9898624342096687e-05,
|
|
"loss": 1.3046,
|
|
"step": 567
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9898022839667577e-05,
|
|
"loss": 1.3446,
|
|
"step": 568
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9897419567180565e-05,
|
|
"loss": 1.3339,
|
|
"step": 569
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.989681452474353e-05,
|
|
"loss": 1.2115,
|
|
"step": 570
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.989620771246467e-05,
|
|
"loss": 1.3036,
|
|
"step": 571
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9895599130452507e-05,
|
|
"loss": 1.1767,
|
|
"step": 572
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.989498877881587e-05,
|
|
"loss": 1.3184,
|
|
"step": 573
|
|
},
|
|
{
|
|
"epoch": 0.22,
|
|
"learning_rate": 1.9894376657663904e-05,
|
|
"loss": 1.2535,
|
|
"step": 574
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.989376276710608e-05,
|
|
"loss": 1.2522,
|
|
"step": 575
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.989314710725218e-05,
|
|
"loss": 1.3376,
|
|
"step": 576
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9892529678212297e-05,
|
|
"loss": 1.2434,
|
|
"step": 577
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.989191048009685e-05,
|
|
"loss": 1.259,
|
|
"step": 578
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9891289513016567e-05,
|
|
"loss": 1.3064,
|
|
"step": 579
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.98906667770825e-05,
|
|
"loss": 1.246,
|
|
"step": 580
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.989004227240601e-05,
|
|
"loss": 1.2378,
|
|
"step": 581
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988941599909877e-05,
|
|
"loss": 1.246,
|
|
"step": 582
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988878795727278e-05,
|
|
"loss": 1.2284,
|
|
"step": 583
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988815814704036e-05,
|
|
"loss": 1.263,
|
|
"step": 584
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988752656851413e-05,
|
|
"loss": 1.2346,
|
|
"step": 585
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9886893221807045e-05,
|
|
"loss": 1.2699,
|
|
"step": 586
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9886258107032352e-05,
|
|
"loss": 1.2533,
|
|
"step": 587
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9885621224303634e-05,
|
|
"loss": 1.1944,
|
|
"step": 588
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988498257373479e-05,
|
|
"loss": 1.2872,
|
|
"step": 589
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9884342155440023e-05,
|
|
"loss": 1.2273,
|
|
"step": 590
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988369996953386e-05,
|
|
"loss": 1.1729,
|
|
"step": 591
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9883056016131142e-05,
|
|
"loss": 1.2758,
|
|
"step": 592
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988241029534703e-05,
|
|
"loss": 1.2195,
|
|
"step": 593
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9881762807296995e-05,
|
|
"loss": 1.2767,
|
|
"step": 594
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988111355209683e-05,
|
|
"loss": 1.2565,
|
|
"step": 595
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.988046252986264e-05,
|
|
"loss": 1.309,
|
|
"step": 596
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.987980974071084e-05,
|
|
"loss": 1.2925,
|
|
"step": 597
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9879155184758175e-05,
|
|
"loss": 1.1759,
|
|
"step": 598
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.98784988621217e-05,
|
|
"loss": 1.2027,
|
|
"step": 599
|
|
},
|
|
{
|
|
"epoch": 0.23,
|
|
"learning_rate": 1.9877840772918785e-05,
|
|
"loss": 1.2723,
|
|
"step": 600
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.987718091726711e-05,
|
|
"loss": 1.2888,
|
|
"step": 601
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.987651929528468e-05,
|
|
"loss": 1.1905,
|
|
"step": 602
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9875855907089813e-05,
|
|
"loss": 1.2737,
|
|
"step": 603
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.987519075280114e-05,
|
|
"loss": 1.2324,
|
|
"step": 604
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9874523832537615e-05,
|
|
"loss": 1.2993,
|
|
"step": 605
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9873855146418498e-05,
|
|
"loss": 1.2841,
|
|
"step": 606
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9873184694563372e-05,
|
|
"loss": 1.2671,
|
|
"step": 607
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.987251247709213e-05,
|
|
"loss": 1.2751,
|
|
"step": 608
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.987183849412499e-05,
|
|
"loss": 1.2984,
|
|
"step": 609
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.987116274578248e-05,
|
|
"loss": 1.2662,
|
|
"step": 610
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.987048523218544e-05,
|
|
"loss": 1.2541,
|
|
"step": 611
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9869805953455025e-05,
|
|
"loss": 1.3044,
|
|
"step": 612
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.986912490971272e-05,
|
|
"loss": 1.2577,
|
|
"step": 613
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9868442101080313e-05,
|
|
"loss": 1.2939,
|
|
"step": 614
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9867757527679905e-05,
|
|
"loss": 1.2217,
|
|
"step": 615
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.986707118963392e-05,
|
|
"loss": 1.2709,
|
|
"step": 616
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.98663830870651e-05,
|
|
"loss": 1.3029,
|
|
"step": 617
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9865693220096493e-05,
|
|
"loss": 1.2887,
|
|
"step": 618
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.986500158885147e-05,
|
|
"loss": 1.2178,
|
|
"step": 619
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9864308193453713e-05,
|
|
"loss": 1.2619,
|
|
"step": 620
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9863613034027224e-05,
|
|
"loss": 1.1629,
|
|
"step": 621
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.986291611069632e-05,
|
|
"loss": 1.2733,
|
|
"step": 622
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9862217423585623e-05,
|
|
"loss": 1.2517,
|
|
"step": 623
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.9861516972820088e-05,
|
|
"loss": 1.1942,
|
|
"step": 624
|
|
},
|
|
{
|
|
"epoch": 0.24,
|
|
"learning_rate": 1.986081475852497e-05,
|
|
"loss": 1.2436,
|
|
"step": 625
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9860110780825844e-05,
|
|
"loss": 1.2397,
|
|
"step": 626
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9859405039848613e-05,
|
|
"loss": 1.2937,
|
|
"step": 627
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9858697535719474e-05,
|
|
"loss": 1.1873,
|
|
"step": 628
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9857988268564955e-05,
|
|
"loss": 1.1704,
|
|
"step": 629
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.985727723851189e-05,
|
|
"loss": 1.2598,
|
|
"step": 630
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.985656444568743e-05,
|
|
"loss": 1.2862,
|
|
"step": 631
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.985584989021905e-05,
|
|
"loss": 1.2308,
|
|
"step": 632
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.985513357223453e-05,
|
|
"loss": 1.2267,
|
|
"step": 633
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9854415491861968e-05,
|
|
"loss": 1.2425,
|
|
"step": 634
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.985369564922978e-05,
|
|
"loss": 1.2216,
|
|
"step": 635
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9852974044466695e-05,
|
|
"loss": 1.2656,
|
|
"step": 636
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9852250677701757e-05,
|
|
"loss": 1.1649,
|
|
"step": 637
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9851525549064324e-05,
|
|
"loss": 1.2514,
|
|
"step": 638
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9850798658684073e-05,
|
|
"loss": 1.3065,
|
|
"step": 639
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9850070006690987e-05,
|
|
"loss": 1.2477,
|
|
"step": 640
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9849339593215377e-05,
|
|
"loss": 1.2377,
|
|
"step": 641
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.984860741838786e-05,
|
|
"loss": 1.2207,
|
|
"step": 642
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.984787348233937e-05,
|
|
"loss": 1.1635,
|
|
"step": 643
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9847137785201158e-05,
|
|
"loss": 1.2052,
|
|
"step": 644
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9846400327104785e-05,
|
|
"loss": 1.2893,
|
|
"step": 645
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.984566110818214e-05,
|
|
"loss": 1.2502,
|
|
"step": 646
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.98449201285654e-05,
|
|
"loss": 1.2837,
|
|
"step": 647
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9844177388387093e-05,
|
|
"loss": 1.2554,
|
|
"step": 648
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9843432887780024e-05,
|
|
"loss": 1.2548,
|
|
"step": 649
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.984268662687735e-05,
|
|
"loss": 1.3569,
|
|
"step": 650
|
|
},
|
|
{
|
|
"epoch": 0.25,
|
|
"learning_rate": 1.9841938605812507e-05,
|
|
"loss": 1.1802,
|
|
"step": 651
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9841188824719277e-05,
|
|
"loss": 1.1955,
|
|
"step": 652
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9840437283731736e-05,
|
|
"loss": 1.2369,
|
|
"step": 653
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9839683982984286e-05,
|
|
"loss": 1.2257,
|
|
"step": 654
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9838928922611634e-05,
|
|
"loss": 1.2716,
|
|
"step": 655
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.983817210274881e-05,
|
|
"loss": 1.2323,
|
|
"step": 656
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9837413523531156e-05,
|
|
"loss": 1.2384,
|
|
"step": 657
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9836653185094325e-05,
|
|
"loss": 1.1973,
|
|
"step": 658
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9835891087574296e-05,
|
|
"loss": 1.2514,
|
|
"step": 659
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9835127231107342e-05,
|
|
"loss": 1.3341,
|
|
"step": 660
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9834361615830077e-05,
|
|
"loss": 1.3048,
|
|
"step": 661
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9833594241879406e-05,
|
|
"loss": 1.1719,
|
|
"step": 662
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.983282510939256e-05,
|
|
"loss": 1.2527,
|
|
"step": 663
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.983205421850708e-05,
|
|
"loss": 1.2582,
|
|
"step": 664
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.983128156936083e-05,
|
|
"loss": 1.207,
|
|
"step": 665
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9830507162091984e-05,
|
|
"loss": 1.1819,
|
|
"step": 666
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.982973099683902e-05,
|
|
"loss": 1.273,
|
|
"step": 667
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9828953073740744e-05,
|
|
"loss": 1.2188,
|
|
"step": 668
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.982817339293627e-05,
|
|
"loss": 1.1583,
|
|
"step": 669
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.982739195456503e-05,
|
|
"loss": 1.272,
|
|
"step": 670
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9826608758766767e-05,
|
|
"loss": 1.2997,
|
|
"step": 671
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9825823805681543e-05,
|
|
"loss": 1.3041,
|
|
"step": 672
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9825037095449722e-05,
|
|
"loss": 1.2588,
|
|
"step": 673
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9824248628212e-05,
|
|
"loss": 1.1868,
|
|
"step": 674
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.9823458404109376e-05,
|
|
"loss": 1.3453,
|
|
"step": 675
|
|
},
|
|
{
|
|
"epoch": 0.26,
|
|
"learning_rate": 1.982266642328316e-05,
|
|
"loss": 1.2792,
|
|
"step": 676
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.982187268587499e-05,
|
|
"loss": 1.2182,
|
|
"step": 677
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.98210771920268e-05,
|
|
"loss": 1.2057,
|
|
"step": 678
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9820279941880854e-05,
|
|
"loss": 1.2251,
|
|
"step": 679
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9819480935579723e-05,
|
|
"loss": 1.2987,
|
|
"step": 680
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.981868017326629e-05,
|
|
"loss": 1.2689,
|
|
"step": 681
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.981787765508376e-05,
|
|
"loss": 1.2621,
|
|
"step": 682
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9817073381175644e-05,
|
|
"loss": 1.2373,
|
|
"step": 683
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.981626735168577e-05,
|
|
"loss": 1.1532,
|
|
"step": 684
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.981545956675828e-05,
|
|
"loss": 1.2395,
|
|
"step": 685
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9814650026537632e-05,
|
|
"loss": 1.2142,
|
|
"step": 686
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9813838731168588e-05,
|
|
"loss": 1.2328,
|
|
"step": 687
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.981302568079624e-05,
|
|
"loss": 1.2079,
|
|
"step": 688
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.981221087556598e-05,
|
|
"loss": 1.2714,
|
|
"step": 689
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9811394315623523e-05,
|
|
"loss": 1.3035,
|
|
"step": 690
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.981057600111489e-05,
|
|
"loss": 1.2731,
|
|
"step": 691
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9809755932186425e-05,
|
|
"loss": 1.1727,
|
|
"step": 692
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.980893410898477e-05,
|
|
"loss": 1.3174,
|
|
"step": 693
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9808110531656907e-05,
|
|
"loss": 1.228,
|
|
"step": 694
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9807285200350104e-05,
|
|
"loss": 1.303,
|
|
"step": 695
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9806458115211957e-05,
|
|
"loss": 1.1939,
|
|
"step": 696
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9805629276390376e-05,
|
|
"loss": 1.2661,
|
|
"step": 697
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9804798684033578e-05,
|
|
"loss": 1.2887,
|
|
"step": 698
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9803966338290098e-05,
|
|
"loss": 1.2495,
|
|
"step": 699
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9803132239308787e-05,
|
|
"loss": 1.1779,
|
|
"step": 700
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9802296387238806e-05,
|
|
"loss": 1.2011,
|
|
"step": 701
|
|
},
|
|
{
|
|
"epoch": 0.27,
|
|
"learning_rate": 1.9801458782229626e-05,
|
|
"loss": 1.2204,
|
|
"step": 702
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.980061942443104e-05,
|
|
"loss": 1.2589,
|
|
"step": 703
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9799778313993148e-05,
|
|
"loss": 1.3396,
|
|
"step": 704
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9798935451066363e-05,
|
|
"loss": 1.2359,
|
|
"step": 705
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9798090835801418e-05,
|
|
"loss": 1.2752,
|
|
"step": 706
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.979724446834935e-05,
|
|
"loss": 1.2233,
|
|
"step": 707
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9796396348861522e-05,
|
|
"loss": 1.2449,
|
|
"step": 708
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9795546477489592e-05,
|
|
"loss": 1.2745,
|
|
"step": 709
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9794694854385555e-05,
|
|
"loss": 1.2153,
|
|
"step": 710
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9793841479701696e-05,
|
|
"loss": 1.2296,
|
|
"step": 711
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9792986353590628e-05,
|
|
"loss": 1.2601,
|
|
"step": 712
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.979212947620527e-05,
|
|
"loss": 1.2183,
|
|
"step": 713
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.979127084769886e-05,
|
|
"loss": 1.226,
|
|
"step": 714
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9790410468224943e-05,
|
|
"loss": 1.2472,
|
|
"step": 715
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9789548337937383e-05,
|
|
"loss": 1.215,
|
|
"step": 716
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9788684456990354e-05,
|
|
"loss": 1.1718,
|
|
"step": 717
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.978781882553834e-05,
|
|
"loss": 1.1258,
|
|
"step": 718
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9786951443736148e-05,
|
|
"loss": 1.2615,
|
|
"step": 719
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.978608231173888e-05,
|
|
"loss": 1.2992,
|
|
"step": 720
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9785211429701977e-05,
|
|
"loss": 1.2389,
|
|
"step": 721
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9784338797781166e-05,
|
|
"loss": 1.2415,
|
|
"step": 722
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.9783464416132507e-05,
|
|
"loss": 1.2418,
|
|
"step": 723
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.978258828491236e-05,
|
|
"loss": 1.2442,
|
|
"step": 724
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.978171040427741e-05,
|
|
"loss": 1.1888,
|
|
"step": 725
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.978083077438464e-05,
|
|
"loss": 1.291,
|
|
"step": 726
|
|
},
|
|
{
|
|
"epoch": 0.28,
|
|
"learning_rate": 1.977994939539136e-05,
|
|
"loss": 1.1739,
|
|
"step": 727
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.977906626745518e-05,
|
|
"loss": 1.1915,
|
|
"step": 728
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9778181390734033e-05,
|
|
"loss": 1.2419,
|
|
"step": 729
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9777294765386162e-05,
|
|
"loss": 1.2253,
|
|
"step": 730
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9776406391570124e-05,
|
|
"loss": 1.2871,
|
|
"step": 731
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.977551626944478e-05,
|
|
"loss": 1.2278,
|
|
"step": 732
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9774624399169312e-05,
|
|
"loss": 1.2024,
|
|
"step": 733
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9773730780903214e-05,
|
|
"loss": 1.2913,
|
|
"step": 734
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9772835414806292e-05,
|
|
"loss": 1.268,
|
|
"step": 735
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9771938301038664e-05,
|
|
"loss": 1.2196,
|
|
"step": 736
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9771039439760757e-05,
|
|
"loss": 1.232,
|
|
"step": 737
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9770138831133315e-05,
|
|
"loss": 1.2477,
|
|
"step": 738
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9769236475317396e-05,
|
|
"loss": 1.2494,
|
|
"step": 739
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.976833237247437e-05,
|
|
"loss": 1.2248,
|
|
"step": 740
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9767426522765907e-05,
|
|
"loss": 1.2441,
|
|
"step": 741
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.976651892635401e-05,
|
|
"loss": 1.2606,
|
|
"step": 742
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9765609583400977e-05,
|
|
"loss": 1.2987,
|
|
"step": 743
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.976469849406943e-05,
|
|
"loss": 1.3158,
|
|
"step": 744
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9763785658522296e-05,
|
|
"loss": 1.1686,
|
|
"step": 745
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.976287107692282e-05,
|
|
"loss": 1.207,
|
|
"step": 746
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.976195474943455e-05,
|
|
"loss": 1.2754,
|
|
"step": 747
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9761036676221362e-05,
|
|
"loss": 1.2368,
|
|
"step": 748
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9760116857447422e-05,
|
|
"loss": 1.2441,
|
|
"step": 749
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.975919529327723e-05,
|
|
"loss": 1.2777,
|
|
"step": 750
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9758271983875586e-05,
|
|
"loss": 1.2093,
|
|
"step": 751
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9757346929407607e-05,
|
|
"loss": 1.247,
|
|
"step": 752
|
|
},
|
|
{
|
|
"epoch": 0.29,
|
|
"learning_rate": 1.9756420130038715e-05,
|
|
"loss": 1.3066,
|
|
"step": 753
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9755491585934658e-05,
|
|
"loss": 1.2581,
|
|
"step": 754
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9754561297261475e-05,
|
|
"loss": 1.2585,
|
|
"step": 755
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9753629264185537e-05,
|
|
"loss": 1.2013,
|
|
"step": 756
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9752695486873516e-05,
|
|
"loss": 1.2036,
|
|
"step": 757
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.97517599654924e-05,
|
|
"loss": 1.1789,
|
|
"step": 758
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.975082270020949e-05,
|
|
"loss": 1.2547,
|
|
"step": 759
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9749883691192392e-05,
|
|
"loss": 1.2242,
|
|
"step": 760
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.974894293860903e-05,
|
|
"loss": 1.3077,
|
|
"step": 761
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.974800044262764e-05,
|
|
"loss": 1.2585,
|
|
"step": 762
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.974705620341677e-05,
|
|
"loss": 1.2847,
|
|
"step": 763
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.974611022114527e-05,
|
|
"loss": 1.1657,
|
|
"step": 764
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.974516249598232e-05,
|
|
"loss": 1.2137,
|
|
"step": 765
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.974421302809739e-05,
|
|
"loss": 1.1938,
|
|
"step": 766
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.974326181766028e-05,
|
|
"loss": 1.3231,
|
|
"step": 767
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9742308864841094e-05,
|
|
"loss": 1.177,
|
|
"step": 768
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9741354169810246e-05,
|
|
"loss": 1.1869,
|
|
"step": 769
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9740397732738462e-05,
|
|
"loss": 1.1992,
|
|
"step": 770
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9739439553796784e-05,
|
|
"loss": 1.3038,
|
|
"step": 771
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9738479633156566e-05,
|
|
"loss": 1.1866,
|
|
"step": 772
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.973751797098946e-05,
|
|
"loss": 1.2003,
|
|
"step": 773
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.973655456746745e-05,
|
|
"loss": 1.2065,
|
|
"step": 774
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.973558942276282e-05,
|
|
"loss": 1.2958,
|
|
"step": 775
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9734622537048154e-05,
|
|
"loss": 1.2062,
|
|
"step": 776
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9733653910496374e-05,
|
|
"loss": 1.2527,
|
|
"step": 777
|
|
},
|
|
{
|
|
"epoch": 0.3,
|
|
"learning_rate": 1.9732683543280693e-05,
|
|
"loss": 1.2427,
|
|
"step": 778
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.973171143557464e-05,
|
|
"loss": 1.2055,
|
|
"step": 779
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.973073758755206e-05,
|
|
"loss": 1.2672,
|
|
"step": 780
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9729761999387102e-05,
|
|
"loss": 1.2088,
|
|
"step": 781
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9728784671254237e-05,
|
|
"loss": 1.2557,
|
|
"step": 782
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9727805603328227e-05,
|
|
"loss": 1.2354,
|
|
"step": 783
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.972682479578417e-05,
|
|
"loss": 1.2384,
|
|
"step": 784
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9725842248797458e-05,
|
|
"loss": 1.301,
|
|
"step": 785
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.97248579625438e-05,
|
|
"loss": 1.2833,
|
|
"step": 786
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9723871937199218e-05,
|
|
"loss": 1.3037,
|
|
"step": 787
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9722884172940038e-05,
|
|
"loss": 1.1934,
|
|
"step": 788
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9721894669942904e-05,
|
|
"loss": 1.2946,
|
|
"step": 789
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.972090342838477e-05,
|
|
"loss": 1.1245,
|
|
"step": 790
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9719910448442893e-05,
|
|
"loss": 1.2498,
|
|
"step": 791
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9718915730294856e-05,
|
|
"loss": 1.2921,
|
|
"step": 792
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9717919274118532e-05,
|
|
"loss": 1.1679,
|
|
"step": 793
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9716921080092127e-05,
|
|
"loss": 1.2261,
|
|
"step": 794
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9715921148394148e-05,
|
|
"loss": 1.2768,
|
|
"step": 795
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9714919479203406e-05,
|
|
"loss": 1.2147,
|
|
"step": 796
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.971391607269903e-05,
|
|
"loss": 1.1885,
|
|
"step": 797
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.971291092906046e-05,
|
|
"loss": 1.2342,
|
|
"step": 798
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9711904048467447e-05,
|
|
"loss": 1.2314,
|
|
"step": 799
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.971089543110005e-05,
|
|
"loss": 1.2235,
|
|
"step": 800
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9709885077138635e-05,
|
|
"loss": 1.2403,
|
|
"step": 801
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9708872986763883e-05,
|
|
"loss": 1.2517,
|
|
"step": 802
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9707859160156796e-05,
|
|
"loss": 1.2748,
|
|
"step": 803
|
|
},
|
|
{
|
|
"epoch": 0.31,
|
|
"learning_rate": 1.9706843597498667e-05,
|
|
"loss": 1.1614,
|
|
"step": 804
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9705826298971112e-05,
|
|
"loss": 1.3295,
|
|
"step": 805
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9704807264756052e-05,
|
|
"loss": 1.2399,
|
|
"step": 806
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9703786495035722e-05,
|
|
"loss": 1.315,
|
|
"step": 807
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.970276398999266e-05,
|
|
"loss": 1.207,
|
|
"step": 808
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.970173974980973e-05,
|
|
"loss": 1.3096,
|
|
"step": 809
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9700713774670087e-05,
|
|
"loss": 1.2919,
|
|
"step": 810
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.969968606475721e-05,
|
|
"loss": 1.2521,
|
|
"step": 811
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9698656620254886e-05,
|
|
"loss": 1.3114,
|
|
"step": 812
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.969762544134721e-05,
|
|
"loss": 1.1423,
|
|
"step": 813
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9696592528218577e-05,
|
|
"loss": 1.2361,
|
|
"step": 814
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9695557881053715e-05,
|
|
"loss": 1.1929,
|
|
"step": 815
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.969452150003764e-05,
|
|
"loss": 1.3089,
|
|
"step": 816
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9693483385355695e-05,
|
|
"loss": 1.2304,
|
|
"step": 817
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.969244353719352e-05,
|
|
"loss": 1.2895,
|
|
"step": 818
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.969140195573707e-05,
|
|
"loss": 1.199,
|
|
"step": 819
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9690358641172617e-05,
|
|
"loss": 1.1619,
|
|
"step": 820
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.968931359368673e-05,
|
|
"loss": 1.2566,
|
|
"step": 821
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.96882668134663e-05,
|
|
"loss": 1.2346,
|
|
"step": 822
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9687218300698515e-05,
|
|
"loss": 1.196,
|
|
"step": 823
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9686168055570888e-05,
|
|
"loss": 1.1957,
|
|
"step": 824
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9685116078271224e-05,
|
|
"loss": 1.2962,
|
|
"step": 825
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9684062368987657e-05,
|
|
"loss": 1.2869,
|
|
"step": 826
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9683006927908613e-05,
|
|
"loss": 1.2224,
|
|
"step": 827
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.9681949755222846e-05,
|
|
"loss": 1.2443,
|
|
"step": 828
|
|
},
|
|
{
|
|
"epoch": 0.32,
|
|
"learning_rate": 1.96808908511194e-05,
|
|
"loss": 1.1982,
|
|
"step": 829
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.967983021578764e-05,
|
|
"loss": 1.2828,
|
|
"step": 830
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9678767849417242e-05,
|
|
"loss": 1.2152,
|
|
"step": 831
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9677703752198188e-05,
|
|
"loss": 1.2452,
|
|
"step": 832
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.967663792432077e-05,
|
|
"loss": 1.2376,
|
|
"step": 833
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9675570365975587e-05,
|
|
"loss": 1.2471,
|
|
"step": 834
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9674501077353548e-05,
|
|
"loss": 1.1682,
|
|
"step": 835
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.967343005864588e-05,
|
|
"loss": 1.2878,
|
|
"step": 836
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.967235731004411e-05,
|
|
"loss": 1.0843,
|
|
"step": 837
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9671282831740076e-05,
|
|
"loss": 1.1395,
|
|
"step": 838
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9670206623925926e-05,
|
|
"loss": 1.2857,
|
|
"step": 839
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.966912868679412e-05,
|
|
"loss": 1.2775,
|
|
"step": 840
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9668049020537418e-05,
|
|
"loss": 1.261,
|
|
"step": 841
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9666967625348907e-05,
|
|
"loss": 1.2234,
|
|
"step": 842
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9665884501421963e-05,
|
|
"loss": 1.2906,
|
|
"step": 843
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9664799648950287e-05,
|
|
"loss": 1.2038,
|
|
"step": 844
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9663713068127878e-05,
|
|
"loss": 1.2063,
|
|
"step": 845
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.966262475914905e-05,
|
|
"loss": 1.2405,
|
|
"step": 846
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9661534722208424e-05,
|
|
"loss": 1.2569,
|
|
"step": 847
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.966044295750093e-05,
|
|
"loss": 1.1625,
|
|
"step": 848
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9659349465221815e-05,
|
|
"loss": 1.1876,
|
|
"step": 849
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9658254245566617e-05,
|
|
"loss": 1.1971,
|
|
"step": 850
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.96571572987312e-05,
|
|
"loss": 1.2267,
|
|
"step": 851
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9656058624911728e-05,
|
|
"loss": 1.2552,
|
|
"step": 852
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.965495822430468e-05,
|
|
"loss": 1.1857,
|
|
"step": 853
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.9653856097106833e-05,
|
|
"loss": 1.2571,
|
|
"step": 854
|
|
},
|
|
{
|
|
"epoch": 0.33,
|
|
"learning_rate": 1.965275224351529e-05,
|
|
"loss": 1.1443,
|
|
"step": 855
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.965164666372744e-05,
|
|
"loss": 1.1875,
|
|
"step": 856
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9650539357941003e-05,
|
|
"loss": 1.2654,
|
|
"step": 857
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9649430326353992e-05,
|
|
"loss": 1.298,
|
|
"step": 858
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.964831956916474e-05,
|
|
"loss": 1.2064,
|
|
"step": 859
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9647207086571876e-05,
|
|
"loss": 1.1859,
|
|
"step": 860
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.964609287877435e-05,
|
|
"loss": 1.2817,
|
|
"step": 861
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9644976945971417e-05,
|
|
"loss": 1.2608,
|
|
"step": 862
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.964385928836263e-05,
|
|
"loss": 1.2153,
|
|
"step": 863
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9642739906147872e-05,
|
|
"loss": 1.1951,
|
|
"step": 864
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.964161879952731e-05,
|
|
"loss": 1.2176,
|
|
"step": 865
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9640495968701433e-05,
|
|
"loss": 1.2029,
|
|
"step": 866
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.963937141387104e-05,
|
|
"loss": 1.2389,
|
|
"step": 867
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9638245135237234e-05,
|
|
"loss": 1.2834,
|
|
"step": 868
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9637117133001424e-05,
|
|
"loss": 1.163,
|
|
"step": 869
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9635987407365335e-05,
|
|
"loss": 1.1468,
|
|
"step": 870
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9634855958530985e-05,
|
|
"loss": 1.1867,
|
|
"step": 871
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9633722786700723e-05,
|
|
"loss": 1.2259,
|
|
"step": 872
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.963258789207718e-05,
|
|
"loss": 1.2467,
|
|
"step": 873
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9631451274863326e-05,
|
|
"loss": 1.2197,
|
|
"step": 874
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9630312935262406e-05,
|
|
"loss": 1.2651,
|
|
"step": 875
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9629172873477995e-05,
|
|
"loss": 1.2123,
|
|
"step": 876
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.962803108971397e-05,
|
|
"loss": 1.274,
|
|
"step": 877
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.962688758417451e-05,
|
|
"loss": 1.2031,
|
|
"step": 878
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.962574235706412e-05,
|
|
"loss": 1.2062,
|
|
"step": 879
|
|
},
|
|
{
|
|
"epoch": 0.34,
|
|
"learning_rate": 1.9624595408587587e-05,
|
|
"loss": 1.1829,
|
|
"step": 880
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9623446738950024e-05,
|
|
"loss": 1.3248,
|
|
"step": 881
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9622296348356847e-05,
|
|
"loss": 1.2671,
|
|
"step": 882
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9621144237013786e-05,
|
|
"loss": 1.2592,
|
|
"step": 883
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.961999040512686e-05,
|
|
"loss": 1.3245,
|
|
"step": 884
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9618834852902417e-05,
|
|
"loss": 1.2504,
|
|
"step": 885
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9617677580547103e-05,
|
|
"loss": 1.1506,
|
|
"step": 886
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.961651858826787e-05,
|
|
"loss": 1.3078,
|
|
"step": 887
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.961535787627198e-05,
|
|
"loss": 1.2529,
|
|
"step": 888
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9614195444767005e-05,
|
|
"loss": 1.2203,
|
|
"step": 889
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9613031293960817e-05,
|
|
"loss": 1.2777,
|
|
"step": 890
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9611865424061606e-05,
|
|
"loss": 1.2604,
|
|
"step": 891
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.961069783527786e-05,
|
|
"loss": 1.3047,
|
|
"step": 892
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.960952852781838e-05,
|
|
"loss": 1.3356,
|
|
"step": 893
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9608357501892272e-05,
|
|
"loss": 1.1926,
|
|
"step": 894
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9607184757708953e-05,
|
|
"loss": 1.2834,
|
|
"step": 895
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.960601029547814e-05,
|
|
"loss": 1.1851,
|
|
"step": 896
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.960483411540986e-05,
|
|
"loss": 1.2686,
|
|
"step": 897
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9603656217714454e-05,
|
|
"loss": 1.2389,
|
|
"step": 898
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.960247660260256e-05,
|
|
"loss": 1.2391,
|
|
"step": 899
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9601295270285138e-05,
|
|
"loss": 1.1587,
|
|
"step": 900
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.960011222097343e-05,
|
|
"loss": 1.2571,
|
|
"step": 901
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9598927454879007e-05,
|
|
"loss": 1.2214,
|
|
"step": 902
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9597740972213744e-05,
|
|
"loss": 1.2527,
|
|
"step": 903
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.9596552773189814e-05,
|
|
"loss": 1.2302,
|
|
"step": 904
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.95953628580197e-05,
|
|
"loss": 1.2174,
|
|
"step": 905
|
|
},
|
|
{
|
|
"epoch": 0.35,
|
|
"learning_rate": 1.95941712269162e-05,
|
|
"loss": 1.1868,
|
|
"step": 906
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.959297788009241e-05,
|
|
"loss": 1.2819,
|
|
"step": 907
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9591782817761737e-05,
|
|
"loss": 1.2512,
|
|
"step": 908
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.959058604013789e-05,
|
|
"loss": 1.2032,
|
|
"step": 909
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.958938754743489e-05,
|
|
"loss": 1.2005,
|
|
"step": 910
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9588187339867062e-05,
|
|
"loss": 1.2364,
|
|
"step": 911
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.958698541764904e-05,
|
|
"loss": 1.1953,
|
|
"step": 912
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9585781780995764e-05,
|
|
"loss": 1.2121,
|
|
"step": 913
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9584576430122473e-05,
|
|
"loss": 1.2602,
|
|
"step": 914
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9583369365244728e-05,
|
|
"loss": 1.2,
|
|
"step": 915
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9582160586578384e-05,
|
|
"loss": 1.3111,
|
|
"step": 916
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9580950094339603e-05,
|
|
"loss": 1.2264,
|
|
"step": 917
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9579737888744864e-05,
|
|
"loss": 1.2126,
|
|
"step": 918
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9578523970010935e-05,
|
|
"loss": 1.2279,
|
|
"step": 919
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.957730833835491e-05,
|
|
"loss": 1.2899,
|
|
"step": 920
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9576090993994173e-05,
|
|
"loss": 1.2374,
|
|
"step": 921
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9574871937146426e-05,
|
|
"loss": 1.1472,
|
|
"step": 922
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.957365116802967e-05,
|
|
"loss": 1.2568,
|
|
"step": 923
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9572428686862218e-05,
|
|
"loss": 1.2289,
|
|
"step": 924
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9571204493862676e-05,
|
|
"loss": 1.2417,
|
|
"step": 925
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9569978589249976e-05,
|
|
"loss": 1.2228,
|
|
"step": 926
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.956875097324334e-05,
|
|
"loss": 1.1628,
|
|
"step": 927
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9567521646062305e-05,
|
|
"loss": 1.1927,
|
|
"step": 928
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.956629060792671e-05,
|
|
"loss": 1.2223,
|
|
"step": 929
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.95650578590567e-05,
|
|
"loss": 1.271,
|
|
"step": 930
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9563823399672723e-05,
|
|
"loss": 1.1593,
|
|
"step": 931
|
|
},
|
|
{
|
|
"epoch": 0.36,
|
|
"learning_rate": 1.9562587229995546e-05,
|
|
"loss": 1.1556,
|
|
"step": 932
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9561349350246226e-05,
|
|
"loss": 1.2268,
|
|
"step": 933
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9560109760646134e-05,
|
|
"loss": 1.2364,
|
|
"step": 934
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9558868461416945e-05,
|
|
"loss": 1.2572,
|
|
"step": 935
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9557625452780638e-05,
|
|
"loss": 1.2434,
|
|
"step": 936
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9556380734959504e-05,
|
|
"loss": 1.2126,
|
|
"step": 937
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.955513430817613e-05,
|
|
"loss": 1.2748,
|
|
"step": 938
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9553886172653417e-05,
|
|
"loss": 1.211,
|
|
"step": 939
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9552636328614572e-05,
|
|
"loss": 1.2457,
|
|
"step": 940
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9551384776283094e-05,
|
|
"loss": 1.2292,
|
|
"step": 941
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9550131515882803e-05,
|
|
"loss": 1.2551,
|
|
"step": 942
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.954887654763782e-05,
|
|
"loss": 1.2158,
|
|
"step": 943
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9547619871772575e-05,
|
|
"loss": 1.2188,
|
|
"step": 944
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9546361488511785e-05,
|
|
"loss": 1.2039,
|
|
"step": 945
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9545101398080497e-05,
|
|
"loss": 1.181,
|
|
"step": 946
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9543839600704047e-05,
|
|
"loss": 1.1919,
|
|
"step": 947
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9542576096608088e-05,
|
|
"loss": 1.2149,
|
|
"step": 948
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9541310886018563e-05,
|
|
"loss": 1.2388,
|
|
"step": 949
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9540043969161736e-05,
|
|
"loss": 1.183,
|
|
"step": 950
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9538775346264163e-05,
|
|
"loss": 1.197,
|
|
"step": 951
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9537505017552716e-05,
|
|
"loss": 1.2362,
|
|
"step": 952
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9536232983254567e-05,
|
|
"loss": 1.1553,
|
|
"step": 953
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.953495924359719e-05,
|
|
"loss": 1.2002,
|
|
"step": 954
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.953368379880837e-05,
|
|
"loss": 1.2644,
|
|
"step": 955
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.953240664911619e-05,
|
|
"loss": 1.1491,
|
|
"step": 956
|
|
},
|
|
{
|
|
"epoch": 0.37,
|
|
"learning_rate": 1.9531127794749046e-05,
|
|
"loss": 1.2536,
|
|
"step": 957
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9529847235935636e-05,
|
|
"loss": 1.2414,
|
|
"step": 958
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9528564972904957e-05,
|
|
"loss": 1.2799,
|
|
"step": 959
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.952728100588632e-05,
|
|
"loss": 1.2208,
|
|
"step": 960
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9525995335109333e-05,
|
|
"loss": 1.2391,
|
|
"step": 961
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.952470796080391e-05,
|
|
"loss": 1.2457,
|
|
"step": 962
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.952341888320028e-05,
|
|
"loss": 1.22,
|
|
"step": 963
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9522128102528958e-05,
|
|
"loss": 1.2815,
|
|
"step": 964
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.952083561902078e-05,
|
|
"loss": 1.1611,
|
|
"step": 965
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.951954143290687e-05,
|
|
"loss": 1.1535,
|
|
"step": 966
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9518245544418682e-05,
|
|
"loss": 1.2007,
|
|
"step": 967
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9516947953787953e-05,
|
|
"loss": 1.233,
|
|
"step": 968
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9515648661246724e-05,
|
|
"loss": 1.2652,
|
|
"step": 969
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9514347667027352e-05,
|
|
"loss": 1.2055,
|
|
"step": 970
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9513044971362494e-05,
|
|
"loss": 1.1322,
|
|
"step": 971
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.951174057448511e-05,
|
|
"loss": 1.1714,
|
|
"step": 972
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.951043447662846e-05,
|
|
"loss": 1.1713,
|
|
"step": 973
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9509126678026117e-05,
|
|
"loss": 1.1963,
|
|
"step": 974
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9507817178911956e-05,
|
|
"loss": 1.1938,
|
|
"step": 975
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.950650597952015e-05,
|
|
"loss": 1.1654,
|
|
"step": 976
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9505193080085178e-05,
|
|
"loss": 1.2034,
|
|
"step": 977
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9503878480841832e-05,
|
|
"loss": 1.1318,
|
|
"step": 978
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9502562182025196e-05,
|
|
"loss": 1.2019,
|
|
"step": 979
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9501244183870666e-05,
|
|
"loss": 1.284,
|
|
"step": 980
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9499924486613935e-05,
|
|
"loss": 1.266,
|
|
"step": 981
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9498603090491008e-05,
|
|
"loss": 1.2185,
|
|
"step": 982
|
|
},
|
|
{
|
|
"epoch": 0.38,
|
|
"learning_rate": 1.9497279995738187e-05,
|
|
"loss": 1.2817,
|
|
"step": 983
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9495955202592078e-05,
|
|
"loss": 1.1562,
|
|
"step": 984
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.94946287112896e-05,
|
|
"loss": 1.1832,
|
|
"step": 985
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9493300522067963e-05,
|
|
"loss": 1.1633,
|
|
"step": 986
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9491970635164685e-05,
|
|
"loss": 1.1966,
|
|
"step": 987
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9490639050817592e-05,
|
|
"loss": 1.2219,
|
|
"step": 988
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9489305769264813e-05,
|
|
"loss": 1.1946,
|
|
"step": 989
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9487970790744774e-05,
|
|
"loss": 1.1601,
|
|
"step": 990
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.948663411549621e-05,
|
|
"loss": 1.1826,
|
|
"step": 991
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9485295743758158e-05,
|
|
"loss": 1.2129,
|
|
"step": 992
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9483955675769955e-05,
|
|
"loss": 1.2334,
|
|
"step": 993
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.948261391177125e-05,
|
|
"loss": 1.1962,
|
|
"step": 994
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.948127045200199e-05,
|
|
"loss": 1.2137,
|
|
"step": 995
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.947992529670242e-05,
|
|
"loss": 1.2034,
|
|
"step": 996
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9478578446113093e-05,
|
|
"loss": 1.2311,
|
|
"step": 997
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.947722990047487e-05,
|
|
"loss": 1.2198,
|
|
"step": 998
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.947587966002891e-05,
|
|
"loss": 1.1452,
|
|
"step": 999
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9474527725016674e-05,
|
|
"loss": 1.2225,
|
|
"step": 1000
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.947317409567993e-05,
|
|
"loss": 1.1888,
|
|
"step": 1001
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9471818772260746e-05,
|
|
"loss": 1.1346,
|
|
"step": 1002
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9470461755001492e-05,
|
|
"loss": 1.1821,
|
|
"step": 1003
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9469103044144844e-05,
|
|
"loss": 1.231,
|
|
"step": 1004
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.946774263993378e-05,
|
|
"loss": 1.2387,
|
|
"step": 1005
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.946638054261158e-05,
|
|
"loss": 1.29,
|
|
"step": 1006
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9465016752421825e-05,
|
|
"loss": 1.1785,
|
|
"step": 1007
|
|
},
|
|
{
|
|
"epoch": 0.39,
|
|
"learning_rate": 1.9463651269608402e-05,
|
|
"loss": 1.2244,
|
|
"step": 1008
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.94622840944155e-05,
|
|
"loss": 1.2876,
|
|
"step": 1009
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9460915227087612e-05,
|
|
"loss": 1.2687,
|
|
"step": 1010
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9459544667869527e-05,
|
|
"loss": 1.1957,
|
|
"step": 1011
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9458172417006347e-05,
|
|
"loss": 1.1669,
|
|
"step": 1012
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.945679847474347e-05,
|
|
"loss": 1.2504,
|
|
"step": 1013
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9455422841326595e-05,
|
|
"loss": 1.2499,
|
|
"step": 1014
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.945404551700172e-05,
|
|
"loss": 1.2374,
|
|
"step": 1015
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9452666502015165e-05,
|
|
"loss": 1.271,
|
|
"step": 1016
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9451285796613528e-05,
|
|
"loss": 1.2314,
|
|
"step": 1017
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9449903401043724e-05,
|
|
"loss": 1.2072,
|
|
"step": 1018
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9448519315552964e-05,
|
|
"loss": 1.1898,
|
|
"step": 1019
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9447133540388764e-05,
|
|
"loss": 1.2507,
|
|
"step": 1020
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.944574607579894e-05,
|
|
"loss": 1.2816,
|
|
"step": 1021
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9444356922031615e-05,
|
|
"loss": 1.1418,
|
|
"step": 1022
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9442966079335207e-05,
|
|
"loss": 1.1871,
|
|
"step": 1023
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.944157354795844e-05,
|
|
"loss": 1.1611,
|
|
"step": 1024
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9440179328150345e-05,
|
|
"loss": 1.164,
|
|
"step": 1025
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9438783420160242e-05,
|
|
"loss": 1.1793,
|
|
"step": 1026
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9437385824237763e-05,
|
|
"loss": 1.2127,
|
|
"step": 1027
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9435986540632843e-05,
|
|
"loss": 1.2881,
|
|
"step": 1028
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.943458556959571e-05,
|
|
"loss": 1.2986,
|
|
"step": 1029
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9433182911376903e-05,
|
|
"loss": 1.1965,
|
|
"step": 1030
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9431778566227253e-05,
|
|
"loss": 1.2211,
|
|
"step": 1031
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9430372534397904e-05,
|
|
"loss": 1.1761,
|
|
"step": 1032
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9428964816140295e-05,
|
|
"loss": 1.1862,
|
|
"step": 1033
|
|
},
|
|
{
|
|
"epoch": 0.4,
|
|
"learning_rate": 1.9427555411706165e-05,
|
|
"loss": 1.2357,
|
|
"step": 1034
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9426144321347557e-05,
|
|
"loss": 1.259,
|
|
"step": 1035
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.942473154531682e-05,
|
|
"loss": 1.2083,
|
|
"step": 1036
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9423317083866593e-05,
|
|
"loss": 1.2675,
|
|
"step": 1037
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9421900937249828e-05,
|
|
"loss": 1.2088,
|
|
"step": 1038
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9420483105719773e-05,
|
|
"loss": 1.1224,
|
|
"step": 1039
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.941906358952998e-05,
|
|
"loss": 1.1601,
|
|
"step": 1040
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9417642388934294e-05,
|
|
"loss": 1.1519,
|
|
"step": 1041
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9416219504186876e-05,
|
|
"loss": 1.254,
|
|
"step": 1042
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.941479493554217e-05,
|
|
"loss": 1.1706,
|
|
"step": 1043
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.941336868325494e-05,
|
|
"loss": 1.242,
|
|
"step": 1044
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9411940747580236e-05,
|
|
"loss": 1.2406,
|
|
"step": 1045
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.941051112877342e-05,
|
|
"loss": 1.2097,
|
|
"step": 1046
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9409079827090145e-05,
|
|
"loss": 1.2302,
|
|
"step": 1047
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.940764684278637e-05,
|
|
"loss": 1.1964,
|
|
"step": 1048
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.940621217611836e-05,
|
|
"loss": 1.1934,
|
|
"step": 1049
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.940477582734267e-05,
|
|
"loss": 1.1133,
|
|
"step": 1050
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.940333779671616e-05,
|
|
"loss": 1.226,
|
|
"step": 1051
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9401898084496e-05,
|
|
"loss": 1.2524,
|
|
"step": 1052
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9400456690939647e-05,
|
|
"loss": 1.2283,
|
|
"step": 1053
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.939901361630487e-05,
|
|
"loss": 1.2257,
|
|
"step": 1054
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9397568860849726e-05,
|
|
"loss": 1.2481,
|
|
"step": 1055
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9396122424832586e-05,
|
|
"loss": 1.1394,
|
|
"step": 1056
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9394674308512108e-05,
|
|
"loss": 1.1564,
|
|
"step": 1057
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.939322451214727e-05,
|
|
"loss": 1.1565,
|
|
"step": 1058
|
|
},
|
|
{
|
|
"epoch": 0.41,
|
|
"learning_rate": 1.9391773035997327e-05,
|
|
"loss": 1.2075,
|
|
"step": 1059
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9390319880321845e-05,
|
|
"loss": 1.2377,
|
|
"step": 1060
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9388865045380702e-05,
|
|
"loss": 1.181,
|
|
"step": 1061
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9387408531434057e-05,
|
|
"loss": 1.1567,
|
|
"step": 1062
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9385950338742377e-05,
|
|
"loss": 1.2282,
|
|
"step": 1063
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9384490467566435e-05,
|
|
"loss": 1.1653,
|
|
"step": 1064
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9383028918167296e-05,
|
|
"loss": 1.2175,
|
|
"step": 1065
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9381565690806328e-05,
|
|
"loss": 1.1985,
|
|
"step": 1066
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.93801007857452e-05,
|
|
"loss": 1.1506,
|
|
"step": 1067
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.937863420324588e-05,
|
|
"loss": 1.0919,
|
|
"step": 1068
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9377165943570633e-05,
|
|
"loss": 1.227,
|
|
"step": 1069
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.937569600698203e-05,
|
|
"loss": 1.1115,
|
|
"step": 1070
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9374224393742937e-05,
|
|
"loss": 1.2401,
|
|
"step": 1071
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.937275110411653e-05,
|
|
"loss": 1.2041,
|
|
"step": 1072
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9371276138366264e-05,
|
|
"loss": 1.1397,
|
|
"step": 1073
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9369799496755912e-05,
|
|
"loss": 1.1903,
|
|
"step": 1074
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9368321179549544e-05,
|
|
"loss": 1.2225,
|
|
"step": 1075
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9366841187011523e-05,
|
|
"loss": 1.1659,
|
|
"step": 1076
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.936535951940652e-05,
|
|
"loss": 1.2361,
|
|
"step": 1077
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.936387617699949e-05,
|
|
"loss": 1.2154,
|
|
"step": 1078
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.936239116005571e-05,
|
|
"loss": 1.1597,
|
|
"step": 1079
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.936090446884074e-05,
|
|
"loss": 1.2124,
|
|
"step": 1080
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9359416103620444e-05,
|
|
"loss": 1.2223,
|
|
"step": 1081
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.935792606466098e-05,
|
|
"loss": 1.2654,
|
|
"step": 1082
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9356434352228825e-05,
|
|
"loss": 1.2222,
|
|
"step": 1083
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.935494096659073e-05,
|
|
"loss": 1.2839,
|
|
"step": 1084
|
|
},
|
|
{
|
|
"epoch": 0.42,
|
|
"learning_rate": 1.9353445908013756e-05,
|
|
"loss": 1.2428,
|
|
"step": 1085
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9351949176765267e-05,
|
|
"loss": 1.1379,
|
|
"step": 1086
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9350450773112924e-05,
|
|
"loss": 1.1979,
|
|
"step": 1087
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9348950697324685e-05,
|
|
"loss": 1.158,
|
|
"step": 1088
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9347448949668802e-05,
|
|
"loss": 1.2093,
|
|
"step": 1089
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9345945530413838e-05,
|
|
"loss": 1.1852,
|
|
"step": 1090
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9344440439828648e-05,
|
|
"loss": 1.2593,
|
|
"step": 1091
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9342933678182385e-05,
|
|
"loss": 1.1747,
|
|
"step": 1092
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.93414252457445e-05,
|
|
"loss": 1.1872,
|
|
"step": 1093
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9339915142784753e-05,
|
|
"loss": 1.1877,
|
|
"step": 1094
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9338403369573186e-05,
|
|
"loss": 1.2474,
|
|
"step": 1095
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9336889926380152e-05,
|
|
"loss": 1.1558,
|
|
"step": 1096
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.93353748134763e-05,
|
|
"loss": 1.245,
|
|
"step": 1097
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9333858031132577e-05,
|
|
"loss": 1.1361,
|
|
"step": 1098
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.933233957962023e-05,
|
|
"loss": 1.2172,
|
|
"step": 1099
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9330819459210794e-05,
|
|
"loss": 1.1471,
|
|
"step": 1100
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9329297670176124e-05,
|
|
"loss": 1.1422,
|
|
"step": 1101
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9327774212788348e-05,
|
|
"loss": 1.1076,
|
|
"step": 1102
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9326249087319917e-05,
|
|
"loss": 1.1281,
|
|
"step": 1103
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.932472229404356e-05,
|
|
"loss": 1.2491,
|
|
"step": 1104
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9323193833232316e-05,
|
|
"loss": 1.2031,
|
|
"step": 1105
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.932166370515952e-05,
|
|
"loss": 1.1893,
|
|
"step": 1106
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9320131910098795e-05,
|
|
"loss": 1.1601,
|
|
"step": 1107
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9318598448324083e-05,
|
|
"loss": 1.1986,
|
|
"step": 1108
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9317063320109608e-05,
|
|
"loss": 1.1203,
|
|
"step": 1109
|
|
},
|
|
{
|
|
"epoch": 0.43,
|
|
"learning_rate": 1.9315526525729894e-05,
|
|
"loss": 1.2749,
|
|
"step": 1110
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.931398806545977e-05,
|
|
"loss": 1.1742,
|
|
"step": 1111
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.931244793957435e-05,
|
|
"loss": 1.24,
|
|
"step": 1112
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.931090614834906e-05,
|
|
"loss": 1.0981,
|
|
"step": 1113
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9309362692059617e-05,
|
|
"loss": 1.1858,
|
|
"step": 1114
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9307817570982035e-05,
|
|
"loss": 1.1202,
|
|
"step": 1115
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9306270785392625e-05,
|
|
"loss": 1.2486,
|
|
"step": 1116
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9304722335568e-05,
|
|
"loss": 1.1768,
|
|
"step": 1117
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9303172221785073e-05,
|
|
"loss": 1.223,
|
|
"step": 1118
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9301620444321042e-05,
|
|
"loss": 1.1755,
|
|
"step": 1119
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9300067003453416e-05,
|
|
"loss": 1.2345,
|
|
"step": 1120
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9298511899459992e-05,
|
|
"loss": 1.202,
|
|
"step": 1121
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9296955132618877e-05,
|
|
"loss": 1.2684,
|
|
"step": 1122
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9295396703208454e-05,
|
|
"loss": 1.319,
|
|
"step": 1123
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9293836611507423e-05,
|
|
"loss": 1.2021,
|
|
"step": 1124
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9292274857794772e-05,
|
|
"loss": 1.1759,
|
|
"step": 1125
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9290711442349795e-05,
|
|
"loss": 1.2062,
|
|
"step": 1126
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.928914636545207e-05,
|
|
"loss": 1.1778,
|
|
"step": 1127
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9287579627381482e-05,
|
|
"loss": 1.2119,
|
|
"step": 1128
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9286011228418204e-05,
|
|
"loss": 1.177,
|
|
"step": 1129
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.928444116884272e-05,
|
|
"loss": 1.1806,
|
|
"step": 1130
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.92828694489358e-05,
|
|
"loss": 1.1068,
|
|
"step": 1131
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9281296068978515e-05,
|
|
"loss": 1.1905,
|
|
"step": 1132
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9279721029252224e-05,
|
|
"loss": 1.1669,
|
|
"step": 1133
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.9278144330038603e-05,
|
|
"loss": 1.2427,
|
|
"step": 1134
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.92765659716196e-05,
|
|
"loss": 1.1733,
|
|
"step": 1135
|
|
},
|
|
{
|
|
"epoch": 0.44,
|
|
"learning_rate": 1.927498595427748e-05,
|
|
"loss": 1.1588,
|
|
"step": 1136
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9273404278294798e-05,
|
|
"loss": 1.2473,
|
|
"step": 1137
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9271820943954393e-05,
|
|
"loss": 1.1824,
|
|
"step": 1138
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9270235951539425e-05,
|
|
"loss": 1.1552,
|
|
"step": 1139
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9268649301333327e-05,
|
|
"loss": 1.1287,
|
|
"step": 1140
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9267060993619847e-05,
|
|
"loss": 1.2473,
|
|
"step": 1141
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9265471028683017e-05,
|
|
"loss": 1.279,
|
|
"step": 1142
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.926387940680717e-05,
|
|
"loss": 1.1829,
|
|
"step": 1143
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.926228612827693e-05,
|
|
"loss": 1.1904,
|
|
"step": 1144
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9260691193377234e-05,
|
|
"loss": 1.154,
|
|
"step": 1145
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9259094602393295e-05,
|
|
"loss": 1.202,
|
|
"step": 1146
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.925749635561063e-05,
|
|
"loss": 1.2185,
|
|
"step": 1147
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9255896453315054e-05,
|
|
"loss": 1.1356,
|
|
"step": 1148
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9254294895792677e-05,
|
|
"loss": 1.1207,
|
|
"step": 1149
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9252691683329907e-05,
|
|
"loss": 1.2138,
|
|
"step": 1150
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.925108681621344e-05,
|
|
"loss": 1.1669,
|
|
"step": 1151
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.924948029473028e-05,
|
|
"loss": 1.2173,
|
|
"step": 1152
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9247872119167716e-05,
|
|
"loss": 1.2344,
|
|
"step": 1153
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9246262289813334e-05,
|
|
"loss": 1.1802,
|
|
"step": 1154
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.924465080695503e-05,
|
|
"loss": 1.1752,
|
|
"step": 1155
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9243037670880973e-05,
|
|
"loss": 1.1821,
|
|
"step": 1156
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9241422881879644e-05,
|
|
"loss": 1.209,
|
|
"step": 1157
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.923980644023982e-05,
|
|
"loss": 1.2241,
|
|
"step": 1158
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.923818834625056e-05,
|
|
"loss": 1.2581,
|
|
"step": 1159
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9236568600201225e-05,
|
|
"loss": 1.1449,
|
|
"step": 1160
|
|
},
|
|
{
|
|
"epoch": 0.45,
|
|
"learning_rate": 1.9234947202381487e-05,
|
|
"loss": 1.1713,
|
|
"step": 1161
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9233324153081283e-05,
|
|
"loss": 1.1653,
|
|
"step": 1162
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9231699452590874e-05,
|
|
"loss": 1.1311,
|
|
"step": 1163
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.92300731012008e-05,
|
|
"loss": 1.1175,
|
|
"step": 1164
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.92284450992019e-05,
|
|
"loss": 1.2389,
|
|
"step": 1165
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.922681544688531e-05,
|
|
"loss": 1.1882,
|
|
"step": 1166
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.922518414454246e-05,
|
|
"loss": 1.189,
|
|
"step": 1167
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9223551192465072e-05,
|
|
"loss": 1.1778,
|
|
"step": 1168
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9221916590945167e-05,
|
|
"loss": 1.2691,
|
|
"step": 1169
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9220280340275064e-05,
|
|
"loss": 1.1575,
|
|
"step": 1170
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9218642440747368e-05,
|
|
"loss": 1.2499,
|
|
"step": 1171
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9217002892654987e-05,
|
|
"loss": 1.2179,
|
|
"step": 1172
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9215361696291114e-05,
|
|
"loss": 1.158,
|
|
"step": 1173
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9213718851949253e-05,
|
|
"loss": 1.2297,
|
|
"step": 1174
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.921207435992319e-05,
|
|
"loss": 1.257,
|
|
"step": 1175
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9210428220507e-05,
|
|
"loss": 1.1986,
|
|
"step": 1176
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9208780433995067e-05,
|
|
"loss": 1.1833,
|
|
"step": 1177
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9207131000682067e-05,
|
|
"loss": 1.2289,
|
|
"step": 1178
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.920547992086297e-05,
|
|
"loss": 1.2892,
|
|
"step": 1179
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9203827194833026e-05,
|
|
"loss": 1.1716,
|
|
"step": 1180
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9202172822887803e-05,
|
|
"loss": 1.1553,
|
|
"step": 1181
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.920051680532314e-05,
|
|
"loss": 1.2003,
|
|
"step": 1182
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.919885914243519e-05,
|
|
"loss": 1.1958,
|
|
"step": 1183
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.919719983452039e-05,
|
|
"loss": 1.1898,
|
|
"step": 1184
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.919553888187547e-05,
|
|
"loss": 1.2004,
|
|
"step": 1185
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.9193876284797465e-05,
|
|
"loss": 1.2159,
|
|
"step": 1186
|
|
},
|
|
{
|
|
"epoch": 0.46,
|
|
"learning_rate": 1.919221204358369e-05,
|
|
"loss": 1.1917,
|
|
"step": 1187
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9190546158531758e-05,
|
|
"loss": 1.1993,
|
|
"step": 1188
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9188878629939584e-05,
|
|
"loss": 1.1922,
|
|
"step": 1189
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.918720945810537e-05,
|
|
"loss": 1.1905,
|
|
"step": 1190
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9185538643327607e-05,
|
|
"loss": 1.1717,
|
|
"step": 1191
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9183866185905095e-05,
|
|
"loss": 1.2168,
|
|
"step": 1192
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9182192086136914e-05,
|
|
"loss": 1.2437,
|
|
"step": 1193
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9180516344322445e-05,
|
|
"loss": 1.2303,
|
|
"step": 1194
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9178838960761355e-05,
|
|
"loss": 1.1638,
|
|
"step": 1195
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9177159935753612e-05,
|
|
"loss": 1.1534,
|
|
"step": 1196
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.917547926959948e-05,
|
|
"loss": 1.2478,
|
|
"step": 1197
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9173796962599502e-05,
|
|
"loss": 1.1656,
|
|
"step": 1198
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.917211301505453e-05,
|
|
"loss": 1.2435,
|
|
"step": 1199
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9170427427265705e-05,
|
|
"loss": 1.1985,
|
|
"step": 1200
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9168740199534458e-05,
|
|
"loss": 1.1703,
|
|
"step": 1201
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.916705133216251e-05,
|
|
"loss": 1.2477,
|
|
"step": 1202
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9165360825451888e-05,
|
|
"loss": 1.1116,
|
|
"step": 1203
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.91636686797049e-05,
|
|
"loss": 1.102,
|
|
"step": 1204
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.916197489522415e-05,
|
|
"loss": 1.2537,
|
|
"step": 1205
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.916027947231254e-05,
|
|
"loss": 1.2494,
|
|
"step": 1206
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9158582411273264e-05,
|
|
"loss": 1.2339,
|
|
"step": 1207
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9156883712409803e-05,
|
|
"loss": 1.1773,
|
|
"step": 1208
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9155183376025928e-05,
|
|
"loss": 1.165,
|
|
"step": 1209
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.915348140242572e-05,
|
|
"loss": 1.174,
|
|
"step": 1210
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9151777791913537e-05,
|
|
"loss": 1.2597,
|
|
"step": 1211
|
|
},
|
|
{
|
|
"epoch": 0.47,
|
|
"learning_rate": 1.9150072544794033e-05,
|
|
"loss": 1.1241,
|
|
"step": 1212
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.914836566137216e-05,
|
|
"loss": 1.1637,
|
|
"step": 1213
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9146657141953157e-05,
|
|
"loss": 1.176,
|
|
"step": 1214
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9144946986842562e-05,
|
|
"loss": 1.2084,
|
|
"step": 1215
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9143235196346194e-05,
|
|
"loss": 1.1774,
|
|
"step": 1216
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9141521770770174e-05,
|
|
"loss": 1.1721,
|
|
"step": 1217
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9139806710420914e-05,
|
|
"loss": 1.2923,
|
|
"step": 1218
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9138090015605117e-05,
|
|
"loss": 1.1206,
|
|
"step": 1219
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.913637168662978e-05,
|
|
"loss": 1.1821,
|
|
"step": 1220
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.913465172380219e-05,
|
|
"loss": 1.2645,
|
|
"step": 1221
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9132930127429927e-05,
|
|
"loss": 1.1947,
|
|
"step": 1222
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9131206897820857e-05,
|
|
"loss": 1.1853,
|
|
"step": 1223
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9129482035283152e-05,
|
|
"loss": 1.1992,
|
|
"step": 1224
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9127755540125267e-05,
|
|
"loss": 1.1438,
|
|
"step": 1225
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.912602741265595e-05,
|
|
"loss": 1.2524,
|
|
"step": 1226
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9124297653184237e-05,
|
|
"loss": 1.2125,
|
|
"step": 1227
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.912256626201947e-05,
|
|
"loss": 1.1651,
|
|
"step": 1228
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.912083323947126e-05,
|
|
"loss": 1.2313,
|
|
"step": 1229
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.911909858584953e-05,
|
|
"loss": 1.2336,
|
|
"step": 1230
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9117362301464485e-05,
|
|
"loss": 1.1303,
|
|
"step": 1231
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9115624386626623e-05,
|
|
"loss": 1.1925,
|
|
"step": 1232
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9113884841646736e-05,
|
|
"loss": 1.292,
|
|
"step": 1233
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9112143666835904e-05,
|
|
"loss": 1.1936,
|
|
"step": 1234
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9110400862505504e-05,
|
|
"loss": 1.2082,
|
|
"step": 1235
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9108656428967197e-05,
|
|
"loss": 1.1691,
|
|
"step": 1236
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.910691036653294e-05,
|
|
"loss": 1.1846,
|
|
"step": 1237
|
|
},
|
|
{
|
|
"epoch": 0.48,
|
|
"learning_rate": 1.9105162675514987e-05,
|
|
"loss": 1.2906,
|
|
"step": 1238
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9103413356225862e-05,
|
|
"loss": 1.1568,
|
|
"step": 1239
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9101662408978406e-05,
|
|
"loss": 1.1884,
|
|
"step": 1240
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9099909834085738e-05,
|
|
"loss": 1.153,
|
|
"step": 1241
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9098155631861272e-05,
|
|
"loss": 1.1392,
|
|
"step": 1242
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9096399802618703e-05,
|
|
"loss": 1.2237,
|
|
"step": 1243
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9094642346672035e-05,
|
|
"loss": 1.2172,
|
|
"step": 1244
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9092883264335542e-05,
|
|
"loss": 1.3139,
|
|
"step": 1245
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.909112255592381e-05,
|
|
"loss": 1.2679,
|
|
"step": 1246
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9089360221751702e-05,
|
|
"loss": 1.2154,
|
|
"step": 1247
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9087596262134372e-05,
|
|
"loss": 1.1784,
|
|
"step": 1248
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.908583067738727e-05,
|
|
"loss": 1.2206,
|
|
"step": 1249
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9084063467826137e-05,
|
|
"loss": 1.2823,
|
|
"step": 1250
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9082294633766997e-05,
|
|
"loss": 1.1692,
|
|
"step": 1251
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9080524175526176e-05,
|
|
"loss": 1.1896,
|
|
"step": 1252
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.907875209342028e-05,
|
|
"loss": 1.1704,
|
|
"step": 1253
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.907697838776621e-05,
|
|
"loss": 1.1898,
|
|
"step": 1254
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9075203058881155e-05,
|
|
"loss": 1.2165,
|
|
"step": 1255
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.90734261070826e-05,
|
|
"loss": 1.0827,
|
|
"step": 1256
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9071647532688315e-05,
|
|
"loss": 1.1423,
|
|
"step": 1257
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9069867336016356e-05,
|
|
"loss": 1.171,
|
|
"step": 1258
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9068085517385082e-05,
|
|
"loss": 1.1654,
|
|
"step": 1259
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9066302077113134e-05,
|
|
"loss": 1.3114,
|
|
"step": 1260
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.906451701551944e-05,
|
|
"loss": 1.1332,
|
|
"step": 1261
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9062730332923222e-05,
|
|
"loss": 1.2447,
|
|
"step": 1262
|
|
},
|
|
{
|
|
"epoch": 0.49,
|
|
"learning_rate": 1.9060942029643996e-05,
|
|
"loss": 1.1621,
|
|
"step": 1263
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9059152106001556e-05,
|
|
"loss": 1.2331,
|
|
"step": 1264
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9057360562315997e-05,
|
|
"loss": 1.1754,
|
|
"step": 1265
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9055567398907704e-05,
|
|
"loss": 1.1586,
|
|
"step": 1266
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.905377261609734e-05,
|
|
"loss": 1.2137,
|
|
"step": 1267
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9051976214205868e-05,
|
|
"loss": 1.2442,
|
|
"step": 1268
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.905017819355454e-05,
|
|
"loss": 1.124,
|
|
"step": 1269
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.904837855446489e-05,
|
|
"loss": 1.1655,
|
|
"step": 1270
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9046577297258747e-05,
|
|
"loss": 1.1741,
|
|
"step": 1271
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9044774422258237e-05,
|
|
"loss": 1.1588,
|
|
"step": 1272
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9042969929785754e-05,
|
|
"loss": 1.2757,
|
|
"step": 1273
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9041163820164007e-05,
|
|
"loss": 1.1935,
|
|
"step": 1274
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9039356093715975e-05,
|
|
"loss": 1.2314,
|
|
"step": 1275
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9037546750764933e-05,
|
|
"loss": 1.1813,
|
|
"step": 1276
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9035735791634445e-05,
|
|
"loss": 1.181,
|
|
"step": 1277
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9033923216648365e-05,
|
|
"loss": 1.2077,
|
|
"step": 1278
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9032109026130836e-05,
|
|
"loss": 1.1062,
|
|
"step": 1279
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9030293220406285e-05,
|
|
"loss": 1.0804,
|
|
"step": 1280
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9028475799799436e-05,
|
|
"loss": 1.2554,
|
|
"step": 1281
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.902665676463529e-05,
|
|
"loss": 1.2073,
|
|
"step": 1282
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9024836115239154e-05,
|
|
"loss": 1.1617,
|
|
"step": 1283
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9023013851936603e-05,
|
|
"loss": 1.1978,
|
|
"step": 1284
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.902118997505352e-05,
|
|
"loss": 1.2222,
|
|
"step": 1285
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9019364484916063e-05,
|
|
"loss": 1.1807,
|
|
"step": 1286
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9017537381850686e-05,
|
|
"loss": 1.1815,
|
|
"step": 1287
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.901570866618413e-05,
|
|
"loss": 1.1616,
|
|
"step": 1288
|
|
},
|
|
{
|
|
"epoch": 0.5,
|
|
"learning_rate": 1.9013878338243418e-05,
|
|
"loss": 1.1883,
|
|
"step": 1289
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.9012046398355868e-05,
|
|
"loss": 1.2504,
|
|
"step": 1290
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.901021284684909e-05,
|
|
"loss": 1.1217,
|
|
"step": 1291
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.900837768405097e-05,
|
|
"loss": 1.1546,
|
|
"step": 1292
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.9006540910289698e-05,
|
|
"loss": 1.1341,
|
|
"step": 1293
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.900470252589373e-05,
|
|
"loss": 1.1286,
|
|
"step": 1294
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.900286253119184e-05,
|
|
"loss": 1.1299,
|
|
"step": 1295
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.900102092651306e-05,
|
|
"loss": 1.275,
|
|
"step": 1296
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8999177712186725e-05,
|
|
"loss": 1.2328,
|
|
"step": 1297
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.899733288854246e-05,
|
|
"loss": 1.1897,
|
|
"step": 1298
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.899548645591017e-05,
|
|
"loss": 1.2164,
|
|
"step": 1299
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8993638414620057e-05,
|
|
"loss": 1.2743,
|
|
"step": 1300
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.89917887650026e-05,
|
|
"loss": 1.2135,
|
|
"step": 1301
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8989937507388575e-05,
|
|
"loss": 1.1498,
|
|
"step": 1302
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8988084642109035e-05,
|
|
"loss": 1.2744,
|
|
"step": 1303
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8986230169495335e-05,
|
|
"loss": 1.1561,
|
|
"step": 1304
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8984374089879103e-05,
|
|
"loss": 1.1151,
|
|
"step": 1305
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8982516403592263e-05,
|
|
"loss": 1.2328,
|
|
"step": 1306
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8980657110967024e-05,
|
|
"loss": 1.1968,
|
|
"step": 1307
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8978796212335884e-05,
|
|
"loss": 1.2667,
|
|
"step": 1308
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.897693370803162e-05,
|
|
"loss": 1.2499,
|
|
"step": 1309
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8975069598387312e-05,
|
|
"loss": 1.1473,
|
|
"step": 1310
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.897320388373631e-05,
|
|
"loss": 1.1648,
|
|
"step": 1311
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.897133656441226e-05,
|
|
"loss": 1.1283,
|
|
"step": 1312
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.89694676407491e-05,
|
|
"loss": 1.2381,
|
|
"step": 1313
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.896759711308104e-05,
|
|
"loss": 1.2541,
|
|
"step": 1314
|
|
},
|
|
{
|
|
"epoch": 0.51,
|
|
"learning_rate": 1.8965724981742593e-05,
|
|
"loss": 1.1792,
|
|
"step": 1315
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8963851247068542e-05,
|
|
"loss": 1.1946,
|
|
"step": 1316
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8961975909393977e-05,
|
|
"loss": 1.1312,
|
|
"step": 1317
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8960098969054253e-05,
|
|
"loss": 1.1263,
|
|
"step": 1318
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.895822042638503e-05,
|
|
"loss": 1.2388,
|
|
"step": 1319
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8956340281722246e-05,
|
|
"loss": 1.2512,
|
|
"step": 1320
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8954458535402124e-05,
|
|
"loss": 1.2359,
|
|
"step": 1321
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8952575187761174e-05,
|
|
"loss": 1.1594,
|
|
"step": 1322
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8950690239136192e-05,
|
|
"loss": 1.2324,
|
|
"step": 1323
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.894880368986427e-05,
|
|
"loss": 1.1942,
|
|
"step": 1324
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8946915540282775e-05,
|
|
"loss": 1.2521,
|
|
"step": 1325
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.894502579072936e-05,
|
|
"loss": 1.1299,
|
|
"step": 1326
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8943134441541976e-05,
|
|
"loss": 1.1607,
|
|
"step": 1327
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8941241493058844e-05,
|
|
"loss": 1.1183,
|
|
"step": 1328
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8939346945618484e-05,
|
|
"loss": 1.189,
|
|
"step": 1329
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8937450799559694e-05,
|
|
"loss": 1.1807,
|
|
"step": 1330
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8935553055221564e-05,
|
|
"loss": 1.107,
|
|
"step": 1331
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8933653712943457e-05,
|
|
"loss": 1.1689,
|
|
"step": 1332
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8931752773065046e-05,
|
|
"loss": 1.1373,
|
|
"step": 1333
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8929850235926265e-05,
|
|
"loss": 1.1272,
|
|
"step": 1334
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8927946101867348e-05,
|
|
"loss": 1.1693,
|
|
"step": 1335
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8926040371228805e-05,
|
|
"loss": 1.1445,
|
|
"step": 1336
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8924133044351442e-05,
|
|
"loss": 1.2685,
|
|
"step": 1337
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8922224121576345e-05,
|
|
"loss": 1.2078,
|
|
"step": 1338
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.892031360324488e-05,
|
|
"loss": 1.2908,
|
|
"step": 1339
|
|
},
|
|
{
|
|
"epoch": 0.52,
|
|
"learning_rate": 1.8918401489698713e-05,
|
|
"loss": 1.201,
|
|
"step": 1340
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8916487781279777e-05,
|
|
"loss": 1.0848,
|
|
"step": 1341
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8914572478330307e-05,
|
|
"loss": 1.1781,
|
|
"step": 1342
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.891265558119281e-05,
|
|
"loss": 1.1761,
|
|
"step": 1343
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8910737090210086e-05,
|
|
"loss": 1.2239,
|
|
"step": 1344
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.890881700572522e-05,
|
|
"loss": 1.2037,
|
|
"step": 1345
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8906895328081573e-05,
|
|
"loss": 1.1389,
|
|
"step": 1346
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.89049720576228e-05,
|
|
"loss": 1.1488,
|
|
"step": 1347
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8903047194692844e-05,
|
|
"loss": 1.1782,
|
|
"step": 1348
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8901120739635925e-05,
|
|
"loss": 1.1378,
|
|
"step": 1349
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8899192692796543e-05,
|
|
"loss": 1.1667,
|
|
"step": 1350
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8897263054519498e-05,
|
|
"loss": 1.1727,
|
|
"step": 1351
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.889533182514986e-05,
|
|
"loss": 1.2576,
|
|
"step": 1352
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8893399005032994e-05,
|
|
"loss": 1.1625,
|
|
"step": 1353
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8891464594514545e-05,
|
|
"loss": 1.2181,
|
|
"step": 1354
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.888952859394044e-05,
|
|
"loss": 1.2107,
|
|
"step": 1355
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8887591003656896e-05,
|
|
"loss": 1.1711,
|
|
"step": 1356
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8885651824010406e-05,
|
|
"loss": 1.21,
|
|
"step": 1357
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8883711055347755e-05,
|
|
"loss": 1.1571,
|
|
"step": 1358
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.888176869801601e-05,
|
|
"loss": 1.1967,
|
|
"step": 1359
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.887982475236253e-05,
|
|
"loss": 1.1819,
|
|
"step": 1360
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8877879218734932e-05,
|
|
"loss": 1.2307,
|
|
"step": 1361
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8875932097481147e-05,
|
|
"loss": 1.0852,
|
|
"step": 1362
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8873983388949374e-05,
|
|
"loss": 1.2435,
|
|
"step": 1363
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.8872033093488102e-05,
|
|
"loss": 1.1703,
|
|
"step": 1364
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.88700812114461e-05,
|
|
"loss": 1.189,
|
|
"step": 1365
|
|
},
|
|
{
|
|
"epoch": 0.53,
|
|
"learning_rate": 1.886812774317242e-05,
|
|
"loss": 1.124,
|
|
"step": 1366
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8866172689016406e-05,
|
|
"loss": 1.2337,
|
|
"step": 1367
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8864216049327667e-05,
|
|
"loss": 1.1483,
|
|
"step": 1368
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.886225782445612e-05,
|
|
"loss": 1.1643,
|
|
"step": 1369
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8860298014751947e-05,
|
|
"loss": 1.239,
|
|
"step": 1370
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.885833662056562e-05,
|
|
"loss": 1.1906,
|
|
"step": 1371
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.88563736422479e-05,
|
|
"loss": 1.1905,
|
|
"step": 1372
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8854409080149817e-05,
|
|
"loss": 1.0954,
|
|
"step": 1373
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8852442934622694e-05,
|
|
"loss": 1.2895,
|
|
"step": 1374
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.885047520601814e-05,
|
|
"loss": 1.1455,
|
|
"step": 1375
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8848505894688042e-05,
|
|
"loss": 1.1222,
|
|
"step": 1376
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8846535000984566e-05,
|
|
"loss": 1.2132,
|
|
"step": 1377
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8844562525260172e-05,
|
|
"loss": 1.1907,
|
|
"step": 1378
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8842588467867592e-05,
|
|
"loss": 1.1512,
|
|
"step": 1379
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.884061282915985e-05,
|
|
"loss": 1.1538,
|
|
"step": 1380
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8838635609490247e-05,
|
|
"loss": 1.2076,
|
|
"step": 1381
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8836656809212362e-05,
|
|
"loss": 1.2022,
|
|
"step": 1382
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.883467642868007e-05,
|
|
"loss": 1.1342,
|
|
"step": 1383
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.883269446824752e-05,
|
|
"loss": 1.2032,
|
|
"step": 1384
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8830710928269146e-05,
|
|
"loss": 1.2591,
|
|
"step": 1385
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8828725809099657e-05,
|
|
"loss": 1.1479,
|
|
"step": 1386
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.882673911109406e-05,
|
|
"loss": 1.2006,
|
|
"step": 1387
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8824750834607624e-05,
|
|
"loss": 1.2146,
|
|
"step": 1388
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8822760979995923e-05,
|
|
"loss": 1.2048,
|
|
"step": 1389
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8820769547614796e-05,
|
|
"loss": 1.211,
|
|
"step": 1390
|
|
},
|
|
{
|
|
"epoch": 0.54,
|
|
"learning_rate": 1.8818776537820366e-05,
|
|
"loss": 1.2415,
|
|
"step": 1391
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8816781950969044e-05,
|
|
"loss": 1.1714,
|
|
"step": 1392
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.881478578741753e-05,
|
|
"loss": 1.1795,
|
|
"step": 1393
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8812788047522785e-05,
|
|
"loss": 1.2271,
|
|
"step": 1394
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.881078873164207e-05,
|
|
"loss": 1.1628,
|
|
"step": 1395
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8808787840132918e-05,
|
|
"loss": 1.2319,
|
|
"step": 1396
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8806785373353146e-05,
|
|
"loss": 1.193,
|
|
"step": 1397
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8804781331660865e-05,
|
|
"loss": 1.1825,
|
|
"step": 1398
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.880277571541444e-05,
|
|
"loss": 1.1074,
|
|
"step": 1399
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.880076852497255e-05,
|
|
"loss": 1.2083,
|
|
"step": 1400
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8798759760694128e-05,
|
|
"loss": 1.1586,
|
|
"step": 1401
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8796749422938407e-05,
|
|
"loss": 1.178,
|
|
"step": 1402
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.879473751206489e-05,
|
|
"loss": 1.1949,
|
|
"step": 1403
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8792724028433374e-05,
|
|
"loss": 1.27,
|
|
"step": 1404
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8790708972403918e-05,
|
|
"loss": 1.0803,
|
|
"step": 1405
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.878869234433688e-05,
|
|
"loss": 1.2286,
|
|
"step": 1406
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8786674144592895e-05,
|
|
"loss": 1.1769,
|
|
"step": 1407
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8784654373532867e-05,
|
|
"loss": 1.2575,
|
|
"step": 1408
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8782633031517997e-05,
|
|
"loss": 1.2184,
|
|
"step": 1409
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8780610118909763e-05,
|
|
"loss": 1.1465,
|
|
"step": 1410
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8778585636069918e-05,
|
|
"loss": 1.1873,
|
|
"step": 1411
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8776559583360495e-05,
|
|
"loss": 1.1391,
|
|
"step": 1412
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8774531961143823e-05,
|
|
"loss": 1.2193,
|
|
"step": 1413
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.877250276978249e-05,
|
|
"loss": 1.1898,
|
|
"step": 1414
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.8770472009639377e-05,
|
|
"loss": 1.137,
|
|
"step": 1415
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.876843968107765e-05,
|
|
"loss": 1.1801,
|
|
"step": 1416
|
|
},
|
|
{
|
|
"epoch": 0.55,
|
|
"learning_rate": 1.876640578446074e-05,
|
|
"loss": 1.1572,
|
|
"step": 1417
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8764370320152373e-05,
|
|
"loss": 1.1727,
|
|
"step": 1418
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8762333288516554e-05,
|
|
"loss": 1.1474,
|
|
"step": 1419
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8760294689917556e-05,
|
|
"loss": 1.203,
|
|
"step": 1420
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8758254524719947e-05,
|
|
"loss": 1.2781,
|
|
"step": 1421
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.875621279328856e-05,
|
|
"loss": 1.16,
|
|
"step": 1422
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8754169495988527e-05,
|
|
"loss": 1.1739,
|
|
"step": 1423
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.875212463318524e-05,
|
|
"loss": 1.153,
|
|
"step": 1424
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.875007820524439e-05,
|
|
"loss": 1.2288,
|
|
"step": 1425
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8748030212531936e-05,
|
|
"loss": 1.084,
|
|
"step": 1426
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8745980655414113e-05,
|
|
"loss": 1.1856,
|
|
"step": 1427
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.874392953425745e-05,
|
|
"loss": 1.2274,
|
|
"step": 1428
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.874187684942874e-05,
|
|
"loss": 1.2637,
|
|
"step": 1429
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8739822601295072e-05,
|
|
"loss": 1.178,
|
|
"step": 1430
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8737766790223805e-05,
|
|
"loss": 1.107,
|
|
"step": 1431
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8735709416582573e-05,
|
|
"loss": 1.1803,
|
|
"step": 1432
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8733650480739295e-05,
|
|
"loss": 1.2078,
|
|
"step": 1433
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8731589983062175e-05,
|
|
"loss": 1.1146,
|
|
"step": 1434
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8729527923919684e-05,
|
|
"loss": 1.2097,
|
|
"step": 1435
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8727464303680588e-05,
|
|
"loss": 1.1473,
|
|
"step": 1436
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8725399122713914e-05,
|
|
"loss": 1.1308,
|
|
"step": 1437
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8723332381388983e-05,
|
|
"loss": 1.1655,
|
|
"step": 1438
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8721264080075386e-05,
|
|
"loss": 1.2041,
|
|
"step": 1439
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8719194219142998e-05,
|
|
"loss": 1.2129,
|
|
"step": 1440
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.8717122798961965e-05,
|
|
"loss": 1.1317,
|
|
"step": 1441
|
|
},
|
|
{
|
|
"epoch": 0.56,
|
|
"learning_rate": 1.871504981990273e-05,
|
|
"loss": 1.2057,
|
|
"step": 1442
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8712975282335994e-05,
|
|
"loss": 1.1548,
|
|
"step": 1443
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8710899186632744e-05,
|
|
"loss": 1.1762,
|
|
"step": 1444
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8708821533164255e-05,
|
|
"loss": 1.1061,
|
|
"step": 1445
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8706742322302064e-05,
|
|
"loss": 1.1853,
|
|
"step": 1446
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8704661554418003e-05,
|
|
"loss": 1.2144,
|
|
"step": 1447
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8702579229884168e-05,
|
|
"loss": 1.1607,
|
|
"step": 1448
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.870049534907294e-05,
|
|
"loss": 1.2033,
|
|
"step": 1449
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8698409912356984e-05,
|
|
"loss": 1.2514,
|
|
"step": 1450
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8696322920109237e-05,
|
|
"loss": 1.2168,
|
|
"step": 1451
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8694234372702907e-05,
|
|
"loss": 1.1744,
|
|
"step": 1452
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8692144270511495e-05,
|
|
"loss": 1.1402,
|
|
"step": 1453
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.869005261390877e-05,
|
|
"loss": 1.1727,
|
|
"step": 1454
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8687959403268784e-05,
|
|
"loss": 1.144,
|
|
"step": 1455
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8685864638965858e-05,
|
|
"loss": 1.1695,
|
|
"step": 1456
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8683768321374608e-05,
|
|
"loss": 1.1649,
|
|
"step": 1457
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8681670450869907e-05,
|
|
"loss": 1.1974,
|
|
"step": 1458
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8679571027826923e-05,
|
|
"loss": 1.1723,
|
|
"step": 1459
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8677470052621095e-05,
|
|
"loss": 1.1713,
|
|
"step": 1460
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8675367525628136e-05,
|
|
"loss": 1.1485,
|
|
"step": 1461
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8673263447224036e-05,
|
|
"loss": 1.2053,
|
|
"step": 1462
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8671157817785077e-05,
|
|
"loss": 1.1664,
|
|
"step": 1463
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8669050637687796e-05,
|
|
"loss": 1.1141,
|
|
"step": 1464
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8666941907309026e-05,
|
|
"loss": 1.2397,
|
|
"step": 1465
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.866483162702587e-05,
|
|
"loss": 1.0912,
|
|
"step": 1466
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.86627197972157e-05,
|
|
"loss": 1.1723,
|
|
"step": 1467
|
|
},
|
|
{
|
|
"epoch": 0.57,
|
|
"learning_rate": 1.8660606418256184e-05,
|
|
"loss": 1.2263,
|
|
"step": 1468
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.865849149052525e-05,
|
|
"loss": 1.1732,
|
|
"step": 1469
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8656375014401113e-05,
|
|
"loss": 1.1506,
|
|
"step": 1470
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.865425699026226e-05,
|
|
"loss": 1.1974,
|
|
"step": 1471
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8652137418487454e-05,
|
|
"loss": 1.217,
|
|
"step": 1472
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.865001629945574e-05,
|
|
"loss": 1.2224,
|
|
"step": 1473
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8647893633546434e-05,
|
|
"loss": 1.1605,
|
|
"step": 1474
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8645769421139136e-05,
|
|
"loss": 1.2086,
|
|
"step": 1475
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8643643662613708e-05,
|
|
"loss": 1.1776,
|
|
"step": 1476
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.864151635835031e-05,
|
|
"loss": 1.1801,
|
|
"step": 1477
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8639387508729358e-05,
|
|
"loss": 1.288,
|
|
"step": 1478
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.863725711413155e-05,
|
|
"loss": 1.1083,
|
|
"step": 1479
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.863512517493788e-05,
|
|
"loss": 1.2136,
|
|
"step": 1480
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8632991691529585e-05,
|
|
"loss": 1.1579,
|
|
"step": 1481
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.86308566642882e-05,
|
|
"loss": 1.1192,
|
|
"step": 1482
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8628720093595533e-05,
|
|
"loss": 1.2206,
|
|
"step": 1483
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.862658197983366e-05,
|
|
"loss": 1.1012,
|
|
"step": 1484
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8624442323384944e-05,
|
|
"loss": 1.1146,
|
|
"step": 1485
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8622301124632017e-05,
|
|
"loss": 1.217,
|
|
"step": 1486
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8620158383957785e-05,
|
|
"loss": 1.1375,
|
|
"step": 1487
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8618014101745444e-05,
|
|
"loss": 1.2606,
|
|
"step": 1488
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.861586827837844e-05,
|
|
"loss": 1.2251,
|
|
"step": 1489
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.861372091424052e-05,
|
|
"loss": 1.1333,
|
|
"step": 1490
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.8611572009715695e-05,
|
|
"loss": 1.1449,
|
|
"step": 1491
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.860942156518825e-05,
|
|
"loss": 1.2165,
|
|
"step": 1492
|
|
},
|
|
{
|
|
"epoch": 0.58,
|
|
"learning_rate": 1.860726958104274e-05,
|
|
"loss": 1.1918,
|
|
"step": 1493
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8605116057664017e-05,
|
|
"loss": 1.1505,
|
|
"step": 1494
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8602960995437187e-05,
|
|
"loss": 1.1989,
|
|
"step": 1495
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.860080439474764e-05,
|
|
"loss": 1.2521,
|
|
"step": 1496
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8598646255981042e-05,
|
|
"loss": 1.1825,
|
|
"step": 1497
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8596486579523326e-05,
|
|
"loss": 1.2668,
|
|
"step": 1498
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.859432536576071e-05,
|
|
"loss": 1.2713,
|
|
"step": 1499
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.859216261507968e-05,
|
|
"loss": 1.1284,
|
|
"step": 1500
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8589998327867004e-05,
|
|
"loss": 1.1529,
|
|
"step": 1501
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8587832504509716e-05,
|
|
"loss": 1.2651,
|
|
"step": 1502
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.858566514539513e-05,
|
|
"loss": 1.0833,
|
|
"step": 1503
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8583496250910835e-05,
|
|
"loss": 1.1968,
|
|
"step": 1504
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.858132582144469e-05,
|
|
"loss": 1.2015,
|
|
"step": 1505
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8579153857384836e-05,
|
|
"loss": 1.2375,
|
|
"step": 1506
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8576980359119683e-05,
|
|
"loss": 1.2109,
|
|
"step": 1507
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8574805327037914e-05,
|
|
"loss": 1.0328,
|
|
"step": 1508
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.857262876152849e-05,
|
|
"loss": 1.1036,
|
|
"step": 1509
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8570450662980648e-05,
|
|
"loss": 1.178,
|
|
"step": 1510
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8568271031783892e-05,
|
|
"loss": 1.1015,
|
|
"step": 1511
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8566089868328004e-05,
|
|
"loss": 1.1231,
|
|
"step": 1512
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8563907173003043e-05,
|
|
"loss": 1.193,
|
|
"step": 1513
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.856172294619934e-05,
|
|
"loss": 1.1631,
|
|
"step": 1514
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.855953718830749e-05,
|
|
"loss": 1.256,
|
|
"step": 1515
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8557349899718383e-05,
|
|
"loss": 1.1688,
|
|
"step": 1516
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8555161080823167e-05,
|
|
"loss": 1.1829,
|
|
"step": 1517
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.8552970732013267e-05,
|
|
"loss": 1.1423,
|
|
"step": 1518
|
|
},
|
|
{
|
|
"epoch": 0.59,
|
|
"learning_rate": 1.855077885368038e-05,
|
|
"loss": 1.1689,
|
|
"step": 1519
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8548585446216475e-05,
|
|
"loss": 1.1985,
|
|
"step": 1520
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8546390510013806e-05,
|
|
"loss": 1.1465,
|
|
"step": 1521
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8544194045464888e-05,
|
|
"loss": 1.2276,
|
|
"step": 1522
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8541996052962513e-05,
|
|
"loss": 1.2269,
|
|
"step": 1523
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8539796532899744e-05,
|
|
"loss": 1.2359,
|
|
"step": 1524
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8537595485669932e-05,
|
|
"loss": 1.1454,
|
|
"step": 1525
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.853539291166667e-05,
|
|
"loss": 1.2172,
|
|
"step": 1526
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.853318881128386e-05,
|
|
"loss": 1.1234,
|
|
"step": 1527
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.853098318491565e-05,
|
|
"loss": 1.1685,
|
|
"step": 1528
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.852877603295648e-05,
|
|
"loss": 1.2025,
|
|
"step": 1529
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8526567355801044e-05,
|
|
"loss": 1.1218,
|
|
"step": 1530
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8524357153844325e-05,
|
|
"loss": 1.2255,
|
|
"step": 1531
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8522145427481567e-05,
|
|
"loss": 1.0958,
|
|
"step": 1532
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8519932177108297e-05,
|
|
"loss": 1.189,
|
|
"step": 1533
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8517717403120307e-05,
|
|
"loss": 1.1727,
|
|
"step": 1534
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8515501105913666e-05,
|
|
"loss": 1.1864,
|
|
"step": 1535
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.851328328588471e-05,
|
|
"loss": 1.2598,
|
|
"step": 1536
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.851106394343005e-05,
|
|
"loss": 1.1961,
|
|
"step": 1537
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8508843078946566e-05,
|
|
"loss": 1.0796,
|
|
"step": 1538
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8506620692831427e-05,
|
|
"loss": 1.2293,
|
|
"step": 1539
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8504396785482052e-05,
|
|
"loss": 1.251,
|
|
"step": 1540
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8502171357296144e-05,
|
|
"loss": 1.1674,
|
|
"step": 1541
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.849994440867167e-05,
|
|
"loss": 1.1748,
|
|
"step": 1542
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.849771594000688e-05,
|
|
"loss": 1.2072,
|
|
"step": 1543
|
|
},
|
|
{
|
|
"epoch": 0.6,
|
|
"learning_rate": 1.8495485951700285e-05,
|
|
"loss": 1.1753,
|
|
"step": 1544
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.849325444415068e-05,
|
|
"loss": 1.0947,
|
|
"step": 1545
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8491021417757114e-05,
|
|
"loss": 1.2139,
|
|
"step": 1546
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8488786872918925e-05,
|
|
"loss": 1.1595,
|
|
"step": 1547
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8486550810035714e-05,
|
|
"loss": 1.1237,
|
|
"step": 1548
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8484313229507352e-05,
|
|
"loss": 1.1807,
|
|
"step": 1549
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8482074131733992e-05,
|
|
"loss": 1.2594,
|
|
"step": 1550
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.847983351711604e-05,
|
|
"loss": 1.2009,
|
|
"step": 1551
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.847759138605419e-05,
|
|
"loss": 1.1877,
|
|
"step": 1552
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8475347738949403e-05,
|
|
"loss": 1.2303,
|
|
"step": 1553
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8473102576202904e-05,
|
|
"loss": 1.2542,
|
|
"step": 1554
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8470855898216197e-05,
|
|
"loss": 1.1373,
|
|
"step": 1555
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.846860770539105e-05,
|
|
"loss": 1.1592,
|
|
"step": 1556
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8466357998129516e-05,
|
|
"loss": 1.0976,
|
|
"step": 1557
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.84641067768339e-05,
|
|
"loss": 1.2,
|
|
"step": 1558
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8461854041906787e-05,
|
|
"loss": 1.1156,
|
|
"step": 1559
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.845959979375104e-05,
|
|
"loss": 1.131,
|
|
"step": 1560
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8457344032769774e-05,
|
|
"loss": 1.1777,
|
|
"step": 1561
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8455086759366393e-05,
|
|
"loss": 1.1847,
|
|
"step": 1562
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8452827973944566e-05,
|
|
"loss": 1.2121,
|
|
"step": 1563
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8450567676908222e-05,
|
|
"loss": 1.1565,
|
|
"step": 1564
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8448305868661572e-05,
|
|
"loss": 1.1376,
|
|
"step": 1565
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.84460425496091e-05,
|
|
"loss": 1.1616,
|
|
"step": 1566
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.844377772015554e-05,
|
|
"loss": 1.2039,
|
|
"step": 1567
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.8441511380705927e-05,
|
|
"loss": 1.1874,
|
|
"step": 1568
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.843924353166554e-05,
|
|
"loss": 1.1647,
|
|
"step": 1569
|
|
},
|
|
{
|
|
"epoch": 0.61,
|
|
"learning_rate": 1.843697417343994e-05,
|
|
"loss": 1.1837,
|
|
"step": 1570
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8434703306434948e-05,
|
|
"loss": 1.1867,
|
|
"step": 1571
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.843243093105667e-05,
|
|
"loss": 1.1396,
|
|
"step": 1572
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8430157047711473e-05,
|
|
"loss": 1.2064,
|
|
"step": 1573
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.842788165680599e-05,
|
|
"loss": 1.1481,
|
|
"step": 1574
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8425604758747136e-05,
|
|
"loss": 1.2667,
|
|
"step": 1575
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8423326353942076e-05,
|
|
"loss": 1.175,
|
|
"step": 1576
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8421046442798265e-05,
|
|
"loss": 1.2048,
|
|
"step": 1577
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8418765025723413e-05,
|
|
"loss": 1.2006,
|
|
"step": 1578
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8416482103125505e-05,
|
|
"loss": 1.2094,
|
|
"step": 1579
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8414197675412796e-05,
|
|
"loss": 1.202,
|
|
"step": 1580
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.841191174299381e-05,
|
|
"loss": 1.1497,
|
|
"step": 1581
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8409624306277342e-05,
|
|
"loss": 1.1854,
|
|
"step": 1582
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8407335365672446e-05,
|
|
"loss": 1.2599,
|
|
"step": 1583
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8405044921588456e-05,
|
|
"loss": 1.1651,
|
|
"step": 1584
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8402752974434968e-05,
|
|
"loss": 1.2436,
|
|
"step": 1585
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8400459524621852e-05,
|
|
"loss": 1.1746,
|
|
"step": 1586
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8398164572559243e-05,
|
|
"loss": 1.1628,
|
|
"step": 1587
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.839586811865755e-05,
|
|
"loss": 1.1311,
|
|
"step": 1588
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.839357016332744e-05,
|
|
"loss": 1.1885,
|
|
"step": 1589
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8391270706979864e-05,
|
|
"loss": 1.3217,
|
|
"step": 1590
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8388969750026023e-05,
|
|
"loss": 1.1211,
|
|
"step": 1591
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.83866672928774e-05,
|
|
"loss": 1.1061,
|
|
"step": 1592
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.838436333594574e-05,
|
|
"loss": 1.1723,
|
|
"step": 1593
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.8382057879643062e-05,
|
|
"loss": 1.3199,
|
|
"step": 1594
|
|
},
|
|
{
|
|
"epoch": 0.62,
|
|
"learning_rate": 1.837975092438165e-05,
|
|
"loss": 1.174,
|
|
"step": 1595
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8377442470574053e-05,
|
|
"loss": 1.1467,
|
|
"step": 1596
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.837513251863309e-05,
|
|
"loss": 1.135,
|
|
"step": 1597
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.837282106897185e-05,
|
|
"loss": 1.1643,
|
|
"step": 1598
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8370508122003686e-05,
|
|
"loss": 1.1349,
|
|
"step": 1599
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8368193678142223e-05,
|
|
"loss": 1.1852,
|
|
"step": 1600
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.836587773780135e-05,
|
|
"loss": 1.104,
|
|
"step": 1601
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.836356030139523e-05,
|
|
"loss": 1.2843,
|
|
"step": 1602
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8361241369338283e-05,
|
|
"loss": 1.1454,
|
|
"step": 1603
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8358920942045205e-05,
|
|
"loss": 1.2036,
|
|
"step": 1604
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8356599019930957e-05,
|
|
"loss": 1.1669,
|
|
"step": 1605
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8354275603410764e-05,
|
|
"loss": 1.1793,
|
|
"step": 1606
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8351950692900127e-05,
|
|
"loss": 1.2123,
|
|
"step": 1607
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8349624288814803e-05,
|
|
"loss": 1.0948,
|
|
"step": 1608
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8347296391570828e-05,
|
|
"loss": 1.0939,
|
|
"step": 1609
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.834496700158449e-05,
|
|
"loss": 1.0799,
|
|
"step": 1610
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.834263611927236e-05,
|
|
"loss": 1.1705,
|
|
"step": 1611
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8340303745051264e-05,
|
|
"loss": 1.1976,
|
|
"step": 1612
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.83379698793383e-05,
|
|
"loss": 1.1004,
|
|
"step": 1613
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.833563452255083e-05,
|
|
"loss": 1.1928,
|
|
"step": 1614
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.833329767510649e-05,
|
|
"loss": 1.1133,
|
|
"step": 1615
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8330959337423176e-05,
|
|
"loss": 1.1205,
|
|
"step": 1616
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8328619509919047e-05,
|
|
"loss": 1.1461,
|
|
"step": 1617
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8326278193012535e-05,
|
|
"loss": 1.1626,
|
|
"step": 1618
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.832393538712234e-05,
|
|
"loss": 1.1499,
|
|
"step": 1619
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8321591092667417e-05,
|
|
"loss": 1.1232,
|
|
"step": 1620
|
|
},
|
|
{
|
|
"epoch": 0.63,
|
|
"learning_rate": 1.8319245310067e-05,
|
|
"loss": 1.1654,
|
|
"step": 1621
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8316898039740586e-05,
|
|
"loss": 1.1952,
|
|
"step": 1622
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8314549282107934e-05,
|
|
"loss": 1.2448,
|
|
"step": 1623
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.831219903758907e-05,
|
|
"loss": 1.2128,
|
|
"step": 1624
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8309847306604288e-05,
|
|
"loss": 1.1772,
|
|
"step": 1625
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8307494089574148e-05,
|
|
"loss": 1.1743,
|
|
"step": 1626
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.830513938691947e-05,
|
|
"loss": 1.2245,
|
|
"step": 1627
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8302783199061346e-05,
|
|
"loss": 1.0852,
|
|
"step": 1628
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8300425526421138e-05,
|
|
"loss": 1.2992,
|
|
"step": 1629
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8298066369420454e-05,
|
|
"loss": 1.1262,
|
|
"step": 1630
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8295705728481198e-05,
|
|
"loss": 1.1449,
|
|
"step": 1631
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8293343604025504e-05,
|
|
"loss": 1.0919,
|
|
"step": 1632
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8290979996475804e-05,
|
|
"loss": 1.1367,
|
|
"step": 1633
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8288614906254775e-05,
|
|
"loss": 1.1671,
|
|
"step": 1634
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.828624833378536e-05,
|
|
"loss": 1.0641,
|
|
"step": 1635
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.828388027949078e-05,
|
|
"loss": 1.1869,
|
|
"step": 1636
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8281510743794507e-05,
|
|
"loss": 1.2094,
|
|
"step": 1637
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.827913972712029e-05,
|
|
"loss": 1.1507,
|
|
"step": 1638
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8276767229892132e-05,
|
|
"loss": 1.2551,
|
|
"step": 1639
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8274393252534305e-05,
|
|
"loss": 1.1424,
|
|
"step": 1640
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8272017795471345e-05,
|
|
"loss": 1.1677,
|
|
"step": 1641
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8269640859128057e-05,
|
|
"loss": 1.1694,
|
|
"step": 1642
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8267262443929507e-05,
|
|
"loss": 1.256,
|
|
"step": 1643
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8264882550301025e-05,
|
|
"loss": 1.1835,
|
|
"step": 1644
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8262501178668203e-05,
|
|
"loss": 1.1238,
|
|
"step": 1645
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8260118329456905e-05,
|
|
"loss": 1.2233,
|
|
"step": 1646
|
|
},
|
|
{
|
|
"epoch": 0.64,
|
|
"learning_rate": 1.8257734003093252e-05,
|
|
"loss": 1.0827,
|
|
"step": 1647
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8255348200003634e-05,
|
|
"loss": 1.2062,
|
|
"step": 1648
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.82529609206147e-05,
|
|
"loss": 1.2625,
|
|
"step": 1649
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8250572165353365e-05,
|
|
"loss": 1.1595,
|
|
"step": 1650
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8248181934646814e-05,
|
|
"loss": 1.1312,
|
|
"step": 1651
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8245790228922485e-05,
|
|
"loss": 1.136,
|
|
"step": 1652
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8243397048608087e-05,
|
|
"loss": 1.1688,
|
|
"step": 1653
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.82410023941316e-05,
|
|
"loss": 1.2087,
|
|
"step": 1654
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.823860626592124e-05,
|
|
"loss": 1.1638,
|
|
"step": 1655
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.823620866440552e-05,
|
|
"loss": 1.1554,
|
|
"step": 1656
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.82338095900132e-05,
|
|
"loss": 1.2336,
|
|
"step": 1657
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.82314090431733e-05,
|
|
"loss": 1.2209,
|
|
"step": 1658
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.822900702431511e-05,
|
|
"loss": 1.1999,
|
|
"step": 1659
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8226603533868184e-05,
|
|
"loss": 1.1962,
|
|
"step": 1660
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8224198572262336e-05,
|
|
"loss": 1.1511,
|
|
"step": 1661
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8221792139927648e-05,
|
|
"loss": 1.1388,
|
|
"step": 1662
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8219384237294454e-05,
|
|
"loss": 1.2057,
|
|
"step": 1663
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.821697486479336e-05,
|
|
"loss": 1.1913,
|
|
"step": 1664
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.821456402285523e-05,
|
|
"loss": 1.1406,
|
|
"step": 1665
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.82121517119112e-05,
|
|
"loss": 1.199,
|
|
"step": 1666
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.820973793239266e-05,
|
|
"loss": 1.2105,
|
|
"step": 1667
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.820732268473126e-05,
|
|
"loss": 1.1071,
|
|
"step": 1668
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8204905969358924e-05,
|
|
"loss": 1.136,
|
|
"step": 1669
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.820248778670783e-05,
|
|
"loss": 1.2059,
|
|
"step": 1670
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.820006813721042e-05,
|
|
"loss": 1.1124,
|
|
"step": 1671
|
|
},
|
|
{
|
|
"epoch": 0.65,
|
|
"learning_rate": 1.8197647021299398e-05,
|
|
"loss": 1.1072,
|
|
"step": 1672
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8195224439407727e-05,
|
|
"loss": 1.2148,
|
|
"step": 1673
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8192800391968643e-05,
|
|
"loss": 1.1697,
|
|
"step": 1674
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8190374879415634e-05,
|
|
"loss": 1.2282,
|
|
"step": 1675
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8187947902182454e-05,
|
|
"loss": 1.0998,
|
|
"step": 1676
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8185519460703113e-05,
|
|
"loss": 1.163,
|
|
"step": 1677
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8183089555411893e-05,
|
|
"loss": 1.1303,
|
|
"step": 1678
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8180658186743333e-05,
|
|
"loss": 1.2517,
|
|
"step": 1679
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8178225355132227e-05,
|
|
"loss": 1.1956,
|
|
"step": 1680
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8175791061013645e-05,
|
|
"loss": 1.2381,
|
|
"step": 1681
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8173355304822907e-05,
|
|
"loss": 1.0966,
|
|
"step": 1682
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8170918086995598e-05,
|
|
"loss": 1.1565,
|
|
"step": 1683
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8168479407967564e-05,
|
|
"loss": 1.2438,
|
|
"step": 1684
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8166039268174916e-05,
|
|
"loss": 1.113,
|
|
"step": 1685
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8163597668054013e-05,
|
|
"loss": 1.197,
|
|
"step": 1686
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8161154608041496e-05,
|
|
"loss": 1.2278,
|
|
"step": 1687
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8158710088574252e-05,
|
|
"loss": 1.2393,
|
|
"step": 1688
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.815626411008943e-05,
|
|
"loss": 1.178,
|
|
"step": 1689
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8153816673024447e-05,
|
|
"loss": 1.1489,
|
|
"step": 1690
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8151367777816976e-05,
|
|
"loss": 1.2082,
|
|
"step": 1691
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8148917424904952e-05,
|
|
"loss": 1.1486,
|
|
"step": 1692
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8146465614726566e-05,
|
|
"loss": 1.2476,
|
|
"step": 1693
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.814401234772028e-05,
|
|
"loss": 1.1109,
|
|
"step": 1694
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.814155762432481e-05,
|
|
"loss": 1.1513,
|
|
"step": 1695
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.8139101444979123e-05,
|
|
"loss": 1.2105,
|
|
"step": 1696
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.813664381012247e-05,
|
|
"loss": 1.1296,
|
|
"step": 1697
|
|
},
|
|
{
|
|
"epoch": 0.66,
|
|
"learning_rate": 1.813418472019434e-05,
|
|
"loss": 1.2681,
|
|
"step": 1698
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.813172417563449e-05,
|
|
"loss": 1.132,
|
|
"step": 1699
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8129262176882948e-05,
|
|
"loss": 1.1266,
|
|
"step": 1700
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8126798724379976e-05,
|
|
"loss": 1.1145,
|
|
"step": 1701
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8124333818566124e-05,
|
|
"loss": 1.134,
|
|
"step": 1702
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8121867459882186e-05,
|
|
"loss": 1.205,
|
|
"step": 1703
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8119399648769218e-05,
|
|
"loss": 1.0913,
|
|
"step": 1704
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.811693038566854e-05,
|
|
"loss": 1.1987,
|
|
"step": 1705
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.811445967102173e-05,
|
|
"loss": 1.1759,
|
|
"step": 1706
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.811198750527062e-05,
|
|
"loss": 1.1328,
|
|
"step": 1707
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8109513888857305e-05,
|
|
"loss": 1.2094,
|
|
"step": 1708
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.810703882222415e-05,
|
|
"loss": 1.1144,
|
|
"step": 1709
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.810456230581376e-05,
|
|
"loss": 1.1859,
|
|
"step": 1710
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8102084340069014e-05,
|
|
"loss": 1.1496,
|
|
"step": 1711
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8099604925433042e-05,
|
|
"loss": 1.125,
|
|
"step": 1712
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8097124062349237e-05,
|
|
"loss": 1.1464,
|
|
"step": 1713
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8094641751261256e-05,
|
|
"loss": 1.1769,
|
|
"step": 1714
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8092157992613e-05,
|
|
"loss": 1.1618,
|
|
"step": 1715
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8089672786848648e-05,
|
|
"loss": 1.1007,
|
|
"step": 1716
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8087186134412616e-05,
|
|
"loss": 1.2127,
|
|
"step": 1717
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8084698035749603e-05,
|
|
"loss": 1.1195,
|
|
"step": 1718
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.808220849130455e-05,
|
|
"loss": 1.1189,
|
|
"step": 1719
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8079717501522657e-05,
|
|
"loss": 1.2188,
|
|
"step": 1720
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8077225066849387e-05,
|
|
"loss": 1.1531,
|
|
"step": 1721
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8074731187730466e-05,
|
|
"loss": 1.2077,
|
|
"step": 1722
|
|
},
|
|
{
|
|
"epoch": 0.67,
|
|
"learning_rate": 1.8072235864611868e-05,
|
|
"loss": 1.1457,
|
|
"step": 1723
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8069739097939832e-05,
|
|
"loss": 1.2406,
|
|
"step": 1724
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8067240888160854e-05,
|
|
"loss": 1.1383,
|
|
"step": 1725
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.806474123572169e-05,
|
|
"loss": 1.3256,
|
|
"step": 1726
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8062240141069343e-05,
|
|
"loss": 1.0686,
|
|
"step": 1727
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8059737604651092e-05,
|
|
"loss": 1.0939,
|
|
"step": 1728
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8057233626914455e-05,
|
|
"loss": 1.0955,
|
|
"step": 1729
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8054728208307226e-05,
|
|
"loss": 1.14,
|
|
"step": 1730
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8052221349277445e-05,
|
|
"loss": 1.1336,
|
|
"step": 1731
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8049713050273404e-05,
|
|
"loss": 1.1108,
|
|
"step": 1732
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8047203311743673e-05,
|
|
"loss": 1.1233,
|
|
"step": 1733
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8044692134137057e-05,
|
|
"loss": 1.1479,
|
|
"step": 1734
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8042179517902636e-05,
|
|
"loss": 1.1445,
|
|
"step": 1735
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8039665463489738e-05,
|
|
"loss": 1.1887,
|
|
"step": 1736
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8037149971347946e-05,
|
|
"loss": 1.1637,
|
|
"step": 1737
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8034633041927106e-05,
|
|
"loss": 1.163,
|
|
"step": 1738
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8032114675677323e-05,
|
|
"loss": 1.1808,
|
|
"step": 1739
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8029594873048952e-05,
|
|
"loss": 1.1354,
|
|
"step": 1740
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8027073634492606e-05,
|
|
"loss": 1.1625,
|
|
"step": 1741
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8024550960459156e-05,
|
|
"loss": 1.1164,
|
|
"step": 1742
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8022026851399737e-05,
|
|
"loss": 1.179,
|
|
"step": 1743
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8019501307765727e-05,
|
|
"loss": 1.1555,
|
|
"step": 1744
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8016974330008772e-05,
|
|
"loss": 1.1545,
|
|
"step": 1745
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.801444591858077e-05,
|
|
"loss": 1.165,
|
|
"step": 1746
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.801191607393387e-05,
|
|
"loss": 1.0872,
|
|
"step": 1747
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.8009384796520487e-05,
|
|
"loss": 1.2499,
|
|
"step": 1748
|
|
},
|
|
{
|
|
"epoch": 0.68,
|
|
"learning_rate": 1.800685208679329e-05,
|
|
"loss": 1.1235,
|
|
"step": 1749
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.8004317945205197e-05,
|
|
"loss": 1.1754,
|
|
"step": 1750
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.8001782372209393e-05,
|
|
"loss": 1.1319,
|
|
"step": 1751
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.799924536825931e-05,
|
|
"loss": 1.1392,
|
|
"step": 1752
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7996706933808633e-05,
|
|
"loss": 1.1911,
|
|
"step": 1753
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7994167069311318e-05,
|
|
"loss": 1.1131,
|
|
"step": 1754
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7991625775221562e-05,
|
|
"loss": 1.1514,
|
|
"step": 1755
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.798908305199383e-05,
|
|
"loss": 1.2007,
|
|
"step": 1756
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.798653890008283e-05,
|
|
"loss": 1.1222,
|
|
"step": 1757
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7983993319943528e-05,
|
|
"loss": 1.1027,
|
|
"step": 1758
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7981446312031155e-05,
|
|
"loss": 1.0933,
|
|
"step": 1759
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7978897876801192e-05,
|
|
"loss": 1.1618,
|
|
"step": 1760
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7976348014709368e-05,
|
|
"loss": 1.1989,
|
|
"step": 1761
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.797379672621168e-05,
|
|
"loss": 1.1904,
|
|
"step": 1762
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7971244011764368e-05,
|
|
"loss": 1.1295,
|
|
"step": 1763
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7968689871823938e-05,
|
|
"loss": 1.1321,
|
|
"step": 1764
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7966134306847135e-05,
|
|
"loss": 1.1743,
|
|
"step": 1765
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7963577317290982e-05,
|
|
"loss": 1.0516,
|
|
"step": 1766
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7961018903612738e-05,
|
|
"loss": 1.1534,
|
|
"step": 1767
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7958459066269922e-05,
|
|
"loss": 1.1527,
|
|
"step": 1768
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7955897805720308e-05,
|
|
"loss": 1.1926,
|
|
"step": 1769
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7953335122421928e-05,
|
|
"loss": 1.163,
|
|
"step": 1770
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7950771016833064e-05,
|
|
"loss": 1.1615,
|
|
"step": 1771
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.794820548941225e-05,
|
|
"loss": 1.1859,
|
|
"step": 1772
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.7945638540618288e-05,
|
|
"loss": 1.1421,
|
|
"step": 1773
|
|
},
|
|
{
|
|
"epoch": 0.69,
|
|
"learning_rate": 1.794307017091021e-05,
|
|
"loss": 1.1854,
|
|
"step": 1774
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7940500380747322e-05,
|
|
"loss": 1.2454,
|
|
"step": 1775
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7937929170589185e-05,
|
|
"loss": 1.1678,
|
|
"step": 1776
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.79353565408956e-05,
|
|
"loss": 1.2299,
|
|
"step": 1777
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7932782492126626e-05,
|
|
"loss": 1.3141,
|
|
"step": 1778
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7930207024742584e-05,
|
|
"loss": 1.1936,
|
|
"step": 1779
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7927630139204044e-05,
|
|
"loss": 1.2135,
|
|
"step": 1780
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7925051835971827e-05,
|
|
"loss": 1.0661,
|
|
"step": 1781
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7922472115507006e-05,
|
|
"loss": 1.0942,
|
|
"step": 1782
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.791989097827092e-05,
|
|
"loss": 1.2176,
|
|
"step": 1783
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7917308424725146e-05,
|
|
"loss": 1.1965,
|
|
"step": 1784
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.791472445533152e-05,
|
|
"loss": 1.0592,
|
|
"step": 1785
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7912139070552136e-05,
|
|
"loss": 1.0946,
|
|
"step": 1786
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7909552270849333e-05,
|
|
"loss": 1.1033,
|
|
"step": 1787
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7906964056685706e-05,
|
|
"loss": 1.0978,
|
|
"step": 1788
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.790437442852411e-05,
|
|
"loss": 1.1084,
|
|
"step": 1789
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7901783386827643e-05,
|
|
"loss": 1.1979,
|
|
"step": 1790
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7899190932059655e-05,
|
|
"loss": 1.2472,
|
|
"step": 1791
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.789659706468376e-05,
|
|
"loss": 1.1987,
|
|
"step": 1792
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.789400178516382e-05,
|
|
"loss": 1.2224,
|
|
"step": 1793
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.789140509396394e-05,
|
|
"loss": 1.1367,
|
|
"step": 1794
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7888806991548488e-05,
|
|
"loss": 1.1907,
|
|
"step": 1795
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7886207478382078e-05,
|
|
"loss": 1.1423,
|
|
"step": 1796
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.7883606554929582e-05,
|
|
"loss": 1.1774,
|
|
"step": 1797
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.788100422165613e-05,
|
|
"loss": 1.1439,
|
|
"step": 1798
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.787840047902708e-05,
|
|
"loss": 1.1324,
|
|
"step": 1799
|
|
},
|
|
{
|
|
"epoch": 0.7,
|
|
"learning_rate": 1.787579532750807e-05,
|
|
"loss": 1.1013,
|
|
"step": 1800
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7873188767564974e-05,
|
|
"loss": 1.1357,
|
|
"step": 1801
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.787058079966392e-05,
|
|
"loss": 1.1754,
|
|
"step": 1802
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7867971424271288e-05,
|
|
"loss": 1.2037,
|
|
"step": 1803
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7865360641853718e-05,
|
|
"loss": 1.1439,
|
|
"step": 1804
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.786274845287809e-05,
|
|
"loss": 1.185,
|
|
"step": 1805
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.786013485781154e-05,
|
|
"loss": 1.2105,
|
|
"step": 1806
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.785751985712146e-05,
|
|
"loss": 1.1272,
|
|
"step": 1807
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7854903451275484e-05,
|
|
"loss": 1.073,
|
|
"step": 1808
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7852285640741503e-05,
|
|
"loss": 1.1549,
|
|
"step": 1809
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7849666425987658e-05,
|
|
"loss": 1.2445,
|
|
"step": 1810
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7847045807482347e-05,
|
|
"loss": 1.1857,
|
|
"step": 1811
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.784442378569421e-05,
|
|
"loss": 1.1371,
|
|
"step": 1812
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.784180036109214e-05,
|
|
"loss": 1.1574,
|
|
"step": 1813
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7839175534145287e-05,
|
|
"loss": 1.1934,
|
|
"step": 1814
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7836549305323046e-05,
|
|
"loss": 1.1793,
|
|
"step": 1815
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.783392167509506e-05,
|
|
"loss": 1.1517,
|
|
"step": 1816
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.783129264393123e-05,
|
|
"loss": 1.2465,
|
|
"step": 1817
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7828662212301704e-05,
|
|
"loss": 1.1845,
|
|
"step": 1818
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.782603038067688e-05,
|
|
"loss": 1.2697,
|
|
"step": 1819
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.782339714952741e-05,
|
|
"loss": 1.1474,
|
|
"step": 1820
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7820762519324187e-05,
|
|
"loss": 1.1649,
|
|
"step": 1821
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7818126490538365e-05,
|
|
"loss": 1.1551,
|
|
"step": 1822
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7815489063641345e-05,
|
|
"loss": 1.1702,
|
|
"step": 1823
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.7812850239104773e-05,
|
|
"loss": 1.1535,
|
|
"step": 1824
|
|
},
|
|
{
|
|
"epoch": 0.71,
|
|
"learning_rate": 1.781021001740055e-05,
|
|
"loss": 1.1917,
|
|
"step": 1825
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7807568399000824e-05,
|
|
"loss": 1.1415,
|
|
"step": 1826
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7804925384377996e-05,
|
|
"loss": 1.1247,
|
|
"step": 1827
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7802280974004717e-05,
|
|
"loss": 1.1268,
|
|
"step": 1828
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7799635168353877e-05,
|
|
"loss": 1.2343,
|
|
"step": 1829
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7796987967898633e-05,
|
|
"loss": 1.2128,
|
|
"step": 1830
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7794339373112375e-05,
|
|
"loss": 1.127,
|
|
"step": 1831
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7791689384468757e-05,
|
|
"loss": 1.1259,
|
|
"step": 1832
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7789038002441673e-05,
|
|
"loss": 1.1932,
|
|
"step": 1833
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7786385227505266e-05,
|
|
"loss": 1.2373,
|
|
"step": 1834
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.778373106013393e-05,
|
|
"loss": 1.1554,
|
|
"step": 1835
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7781075500802307e-05,
|
|
"loss": 1.1565,
|
|
"step": 1836
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7778418549985294e-05,
|
|
"loss": 1.1323,
|
|
"step": 1837
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.777576020815803e-05,
|
|
"loss": 1.1897,
|
|
"step": 1838
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7773100475795904e-05,
|
|
"loss": 1.1509,
|
|
"step": 1839
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7770439353374555e-05,
|
|
"loss": 1.0866,
|
|
"step": 1840
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.776777684136987e-05,
|
|
"loss": 1.1629,
|
|
"step": 1841
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7765112940257983e-05,
|
|
"loss": 1.1659,
|
|
"step": 1842
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.776244765051528e-05,
|
|
"loss": 1.1577,
|
|
"step": 1843
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.775978097261839e-05,
|
|
"loss": 1.1857,
|
|
"step": 1844
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.77571129070442e-05,
|
|
"loss": 1.107,
|
|
"step": 1845
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7754443454269833e-05,
|
|
"loss": 1.1445,
|
|
"step": 1846
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.775177261477267e-05,
|
|
"loss": 1.1944,
|
|
"step": 1847
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7749100389030326e-05,
|
|
"loss": 1.1925,
|
|
"step": 1848
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.7746426777520688e-05,
|
|
"loss": 1.1901,
|
|
"step": 1849
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.774375178072187e-05,
|
|
"loss": 1.2261,
|
|
"step": 1850
|
|
},
|
|
{
|
|
"epoch": 0.72,
|
|
"learning_rate": 1.774107539911223e-05,
|
|
"loss": 1.2204,
|
|
"step": 1851
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7738397633170404e-05,
|
|
"loss": 1.1507,
|
|
"step": 1852
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.773571848337524e-05,
|
|
"loss": 1.2267,
|
|
"step": 1853
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7733037950205855e-05,
|
|
"loss": 1.1285,
|
|
"step": 1854
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7730356034141606e-05,
|
|
"loss": 1.1866,
|
|
"step": 1855
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7727672735662097e-05,
|
|
"loss": 1.1206,
|
|
"step": 1856
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7724988055247183e-05,
|
|
"loss": 1.2187,
|
|
"step": 1857
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7722301993376965e-05,
|
|
"loss": 1.2343,
|
|
"step": 1858
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7719614550531785e-05,
|
|
"loss": 1.1205,
|
|
"step": 1859
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7716925727192244e-05,
|
|
"loss": 1.184,
|
|
"step": 1860
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7714235523839174e-05,
|
|
"loss": 1.1483,
|
|
"step": 1861
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7711543940953667e-05,
|
|
"loss": 1.0993,
|
|
"step": 1862
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.770885097901706e-05,
|
|
"loss": 1.1389,
|
|
"step": 1863
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.770615663851093e-05,
|
|
"loss": 1.232,
|
|
"step": 1864
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.770346091991711e-05,
|
|
"loss": 1.0842,
|
|
"step": 1865
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7700763823717668e-05,
|
|
"loss": 1.153,
|
|
"step": 1866
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7698065350394925e-05,
|
|
"loss": 1.0676,
|
|
"step": 1867
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7695365500431447e-05,
|
|
"loss": 1.1203,
|
|
"step": 1868
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.769266427431005e-05,
|
|
"loss": 1.0424,
|
|
"step": 1869
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.768996167251379e-05,
|
|
"loss": 1.0792,
|
|
"step": 1870
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.768725769552597e-05,
|
|
"loss": 1.188,
|
|
"step": 1871
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7684552343830147e-05,
|
|
"loss": 1.1328,
|
|
"step": 1872
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7681845617910115e-05,
|
|
"loss": 1.1481,
|
|
"step": 1873
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7679137518249912e-05,
|
|
"loss": 1.1509,
|
|
"step": 1874
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.7676428045333832e-05,
|
|
"loss": 1.048,
|
|
"step": 1875
|
|
},
|
|
{
|
|
"epoch": 0.73,
|
|
"learning_rate": 1.76737171996464e-05,
|
|
"loss": 1.141,
|
|
"step": 1876
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7671004981672404e-05,
|
|
"loss": 1.1719,
|
|
"step": 1877
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7668291391896866e-05,
|
|
"loss": 1.1543,
|
|
"step": 1878
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7665576430805053e-05,
|
|
"loss": 1.0629,
|
|
"step": 1879
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.766286009888248e-05,
|
|
"loss": 1.1786,
|
|
"step": 1880
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7660142396614916e-05,
|
|
"loss": 1.1825,
|
|
"step": 1881
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7657423324488352e-05,
|
|
"loss": 1.17,
|
|
"step": 1882
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.765470288298905e-05,
|
|
"loss": 1.2086,
|
|
"step": 1883
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7651981072603498e-05,
|
|
"loss": 1.2245,
|
|
"step": 1884
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7649257893818436e-05,
|
|
"loss": 1.1274,
|
|
"step": 1885
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7646533347120852e-05,
|
|
"loss": 1.1287,
|
|
"step": 1886
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7643807432997975e-05,
|
|
"loss": 1.1202,
|
|
"step": 1887
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.764108015193728e-05,
|
|
"loss": 1.1767,
|
|
"step": 1888
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.763835150442648e-05,
|
|
"loss": 1.1909,
|
|
"step": 1889
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.763562149095354e-05,
|
|
"loss": 1.2025,
|
|
"step": 1890
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7632890112006668e-05,
|
|
"loss": 1.2039,
|
|
"step": 1891
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.763015736807431e-05,
|
|
"loss": 1.1352,
|
|
"step": 1892
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7627423259645166e-05,
|
|
"loss": 1.254,
|
|
"step": 1893
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7624687787208178e-05,
|
|
"loss": 1.1441,
|
|
"step": 1894
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7621950951252523e-05,
|
|
"loss": 1.2458,
|
|
"step": 1895
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7619212752267628e-05,
|
|
"loss": 1.17,
|
|
"step": 1896
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.761647319074317e-05,
|
|
"loss": 1.1468,
|
|
"step": 1897
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7613732267169057e-05,
|
|
"loss": 1.136,
|
|
"step": 1898
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7610989982035448e-05,
|
|
"loss": 1.0299,
|
|
"step": 1899
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7608246335832742e-05,
|
|
"loss": 1.0686,
|
|
"step": 1900
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.7605501329051594e-05,
|
|
"loss": 1.1751,
|
|
"step": 1901
|
|
},
|
|
{
|
|
"epoch": 0.74,
|
|
"learning_rate": 1.760275496218288e-05,
|
|
"loss": 1.1554,
|
|
"step": 1902
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7600007235717736e-05,
|
|
"loss": 1.2201,
|
|
"step": 1903
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7597258150147537e-05,
|
|
"loss": 1.2101,
|
|
"step": 1904
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7594507705963897e-05,
|
|
"loss": 1.1593,
|
|
"step": 1905
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7591755903658682e-05,
|
|
"loss": 1.1248,
|
|
"step": 1906
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7589002743723994e-05,
|
|
"loss": 1.1894,
|
|
"step": 1907
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7586248226652175e-05,
|
|
"loss": 1.2348,
|
|
"step": 1908
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7583492352935817e-05,
|
|
"loss": 1.0667,
|
|
"step": 1909
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.758073512306775e-05,
|
|
"loss": 1.1813,
|
|
"step": 1910
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7577976537541048e-05,
|
|
"loss": 1.1736,
|
|
"step": 1911
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7575216596849028e-05,
|
|
"loss": 1.1545,
|
|
"step": 1912
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.757245530148525e-05,
|
|
"loss": 1.208,
|
|
"step": 1913
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7569692651943513e-05,
|
|
"loss": 1.0653,
|
|
"step": 1914
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.756692864871786e-05,
|
|
"loss": 1.0697,
|
|
"step": 1915
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7564163292302575e-05,
|
|
"loss": 1.2119,
|
|
"step": 1916
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.756139658319219e-05,
|
|
"loss": 1.1882,
|
|
"step": 1917
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.755862852188147e-05,
|
|
"loss": 1.0971,
|
|
"step": 1918
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7555859108865428e-05,
|
|
"loss": 1.1123,
|
|
"step": 1919
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7553088344639312e-05,
|
|
"loss": 1.0973,
|
|
"step": 1920
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.755031622969862e-05,
|
|
"loss": 1.1309,
|
|
"step": 1921
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.754754276453909e-05,
|
|
"loss": 1.0897,
|
|
"step": 1922
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7544767949656693e-05,
|
|
"loss": 1.062,
|
|
"step": 1923
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7541991785547653e-05,
|
|
"loss": 1.1651,
|
|
"step": 1924
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.753921427270843e-05,
|
|
"loss": 1.1908,
|
|
"step": 1925
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.7536435411635718e-05,
|
|
"loss": 1.1875,
|
|
"step": 1926
|
|
},
|
|
{
|
|
"epoch": 0.75,
|
|
"learning_rate": 1.753365520282647e-05,
|
|
"loss": 1.1871,
|
|
"step": 1927
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7530873646777856e-05,
|
|
"loss": 1.0939,
|
|
"step": 1928
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7528090743987315e-05,
|
|
"loss": 1.1207,
|
|
"step": 1929
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7525306494952498e-05,
|
|
"loss": 1.1815,
|
|
"step": 1930
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7522520900171324e-05,
|
|
"loss": 1.1929,
|
|
"step": 1931
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7519733960141927e-05,
|
|
"loss": 1.1518,
|
|
"step": 1932
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.75169456753627e-05,
|
|
"loss": 1.2296,
|
|
"step": 1933
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7514156046332274e-05,
|
|
"loss": 1.1148,
|
|
"step": 1934
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7511365073549508e-05,
|
|
"loss": 1.1053,
|
|
"step": 1935
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.750857275751352e-05,
|
|
"loss": 1.1414,
|
|
"step": 1936
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.750577909872365e-05,
|
|
"loss": 1.1438,
|
|
"step": 1937
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.750298409767949e-05,
|
|
"loss": 1.1778,
|
|
"step": 1938
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7500187754880865e-05,
|
|
"loss": 1.1374,
|
|
"step": 1939
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.749739007082785e-05,
|
|
"loss": 1.1738,
|
|
"step": 1940
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7494591046020746e-05,
|
|
"loss": 1.1379,
|
|
"step": 1941
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.749179068096011e-05,
|
|
"loss": 1.0969,
|
|
"step": 1942
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.748898897614672e-05,
|
|
"loss": 1.109,
|
|
"step": 1943
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.748618593208161e-05,
|
|
"loss": 1.1979,
|
|
"step": 1944
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7483381549266044e-05,
|
|
"loss": 1.1874,
|
|
"step": 1945
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.748057582820153e-05,
|
|
"loss": 1.1299,
|
|
"step": 1946
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.747776876938981e-05,
|
|
"loss": 1.1401,
|
|
"step": 1947
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7474960373332873e-05,
|
|
"loss": 1.1644,
|
|
"step": 1948
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.747215064053294e-05,
|
|
"loss": 1.1134,
|
|
"step": 1949
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7469339571492478e-05,
|
|
"loss": 1.0845,
|
|
"step": 1950
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7466527166714184e-05,
|
|
"loss": 1.1556,
|
|
"step": 1951
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7463713426700998e-05,
|
|
"loss": 1.1846,
|
|
"step": 1952
|
|
},
|
|
{
|
|
"epoch": 0.76,
|
|
"learning_rate": 1.7460898351956107e-05,
|
|
"loss": 1.1394,
|
|
"step": 1953
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.745808194298292e-05,
|
|
"loss": 1.225,
|
|
"step": 1954
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.74552642002851e-05,
|
|
"loss": 1.1554,
|
|
"step": 1955
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7452445124366543e-05,
|
|
"loss": 1.0526,
|
|
"step": 1956
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7449624715731377e-05,
|
|
"loss": 1.1624,
|
|
"step": 1957
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7446802974883978e-05,
|
|
"loss": 1.1414,
|
|
"step": 1958
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.744397990232896e-05,
|
|
"loss": 1.2614,
|
|
"step": 1959
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7441155498571164e-05,
|
|
"loss": 1.1154,
|
|
"step": 1960
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7438329764115677e-05,
|
|
"loss": 1.1907,
|
|
"step": 1961
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.743550269946783e-05,
|
|
"loss": 1.2392,
|
|
"step": 1962
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7432674305133183e-05,
|
|
"loss": 1.1033,
|
|
"step": 1963
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7429844581617532e-05,
|
|
"loss": 1.1844,
|
|
"step": 1964
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.742701352942692e-05,
|
|
"loss": 1.2417,
|
|
"step": 1965
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7424181149067617e-05,
|
|
"loss": 1.1258,
|
|
"step": 1966
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7421347441046146e-05,
|
|
"loss": 1.1227,
|
|
"step": 1967
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7418512405869246e-05,
|
|
"loss": 1.159,
|
|
"step": 1968
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.741567604404391e-05,
|
|
"loss": 1.1373,
|
|
"step": 1969
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.741283835607736e-05,
|
|
"loss": 1.2125,
|
|
"step": 1970
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7409999342477066e-05,
|
|
"loss": 1.1549,
|
|
"step": 1971
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.740715900375072e-05,
|
|
"loss": 1.2086,
|
|
"step": 1972
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7404317340406257e-05,
|
|
"loss": 1.1385,
|
|
"step": 1973
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7401474352951858e-05,
|
|
"loss": 1.1547,
|
|
"step": 1974
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7398630041895925e-05,
|
|
"loss": 1.2316,
|
|
"step": 1975
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.739578440774711e-05,
|
|
"loss": 1.1277,
|
|
"step": 1976
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7392937451014292e-05,
|
|
"loss": 1.1512,
|
|
"step": 1977
|
|
},
|
|
{
|
|
"epoch": 0.77,
|
|
"learning_rate": 1.7390089172206594e-05,
|
|
"loss": 1.1976,
|
|
"step": 1978
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.738723957183337e-05,
|
|
"loss": 1.1109,
|
|
"step": 1979
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7384388650404213e-05,
|
|
"loss": 1.1089,
|
|
"step": 1980
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7381536408428948e-05,
|
|
"loss": 1.1122,
|
|
"step": 1981
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7378682846417647e-05,
|
|
"loss": 1.1679,
|
|
"step": 1982
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.737582796488061e-05,
|
|
"loss": 1.2236,
|
|
"step": 1983
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7372971764328368e-05,
|
|
"loss": 1.2019,
|
|
"step": 1984
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7370114245271695e-05,
|
|
"loss": 1.2476,
|
|
"step": 1985
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7367255408221606e-05,
|
|
"loss": 1.1508,
|
|
"step": 1986
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7364395253689338e-05,
|
|
"loss": 1.1946,
|
|
"step": 1987
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7361533782186378e-05,
|
|
"loss": 1.0997,
|
|
"step": 1988
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7358670994224432e-05,
|
|
"loss": 1.1046,
|
|
"step": 1989
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7355806890315457e-05,
|
|
"loss": 1.1152,
|
|
"step": 1990
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.735294147097164e-05,
|
|
"loss": 1.0932,
|
|
"step": 1991
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.73500747367054e-05,
|
|
"loss": 1.203,
|
|
"step": 1992
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.734720668802939e-05,
|
|
"loss": 1.0858,
|
|
"step": 1993
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.734433732545651e-05,
|
|
"loss": 1.2929,
|
|
"step": 1994
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.734146664949988e-05,
|
|
"loss": 1.1796,
|
|
"step": 1995
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7338594660672865e-05,
|
|
"loss": 1.1684,
|
|
"step": 1996
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7335721359489058e-05,
|
|
"loss": 1.1382,
|
|
"step": 1997
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.733284674646229e-05,
|
|
"loss": 1.1634,
|
|
"step": 1998
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7329970822106627e-05,
|
|
"loss": 1.1899,
|
|
"step": 1999
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.732709358693637e-05,
|
|
"loss": 1.2005,
|
|
"step": 2000
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7324215041466054e-05,
|
|
"loss": 1.1698,
|
|
"step": 2001
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7321335186210448e-05,
|
|
"loss": 1.0995,
|
|
"step": 2002
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7318454021684553e-05,
|
|
"loss": 1.0678,
|
|
"step": 2003
|
|
},
|
|
{
|
|
"epoch": 0.78,
|
|
"learning_rate": 1.7315571548403606e-05,
|
|
"loss": 1.2014,
|
|
"step": 2004
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7312687766883077e-05,
|
|
"loss": 1.1191,
|
|
"step": 2005
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7309802677638674e-05,
|
|
"loss": 1.1413,
|
|
"step": 2006
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7306916281186338e-05,
|
|
"loss": 1.2116,
|
|
"step": 2007
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7304028578042235e-05,
|
|
"loss": 1.1373,
|
|
"step": 2008
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7301139568722772e-05,
|
|
"loss": 1.1932,
|
|
"step": 2009
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7298249253744597e-05,
|
|
"loss": 1.1792,
|
|
"step": 2010
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7295357633624575e-05,
|
|
"loss": 1.1912,
|
|
"step": 2011
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7292464708879816e-05,
|
|
"loss": 1.1065,
|
|
"step": 2012
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7289570480027658e-05,
|
|
"loss": 1.1461,
|
|
"step": 2013
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7286674947585683e-05,
|
|
"loss": 1.1824,
|
|
"step": 2014
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7283778112071683e-05,
|
|
"loss": 1.1369,
|
|
"step": 2015
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.728087997400371e-05,
|
|
"loss": 1.1594,
|
|
"step": 2016
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.727798053390003e-05,
|
|
"loss": 1.1053,
|
|
"step": 2017
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.727507979227915e-05,
|
|
"loss": 1.1124,
|
|
"step": 2018
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.727217774965981e-05,
|
|
"loss": 1.1672,
|
|
"step": 2019
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7269274406560974e-05,
|
|
"loss": 1.1973,
|
|
"step": 2020
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.726636976350185e-05,
|
|
"loss": 1.2011,
|
|
"step": 2021
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7263463821001877e-05,
|
|
"loss": 1.0951,
|
|
"step": 2022
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7260556579580717e-05,
|
|
"loss": 1.232,
|
|
"step": 2023
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7257648039758273e-05,
|
|
"loss": 1.1063,
|
|
"step": 2024
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7254738202054678e-05,
|
|
"loss": 1.1955,
|
|
"step": 2025
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7251827066990298e-05,
|
|
"loss": 1.183,
|
|
"step": 2026
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7248914635085726e-05,
|
|
"loss": 1.1594,
|
|
"step": 2027
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7246000906861792e-05,
|
|
"loss": 1.1309,
|
|
"step": 2028
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7243085882839557e-05,
|
|
"loss": 1.1087,
|
|
"step": 2029
|
|
},
|
|
{
|
|
"epoch": 0.79,
|
|
"learning_rate": 1.7240169563540313e-05,
|
|
"loss": 1.0461,
|
|
"step": 2030
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7237251949485588e-05,
|
|
"loss": 1.1381,
|
|
"step": 2031
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7234333041197127e-05,
|
|
"loss": 1.1878,
|
|
"step": 2032
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7231412839196925e-05,
|
|
"loss": 1.0256,
|
|
"step": 2033
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7228491344007197e-05,
|
|
"loss": 1.1674,
|
|
"step": 2034
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.722556855615039e-05,
|
|
"loss": 1.1458,
|
|
"step": 2035
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.722264447614919e-05,
|
|
"loss": 1.1261,
|
|
"step": 2036
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7219719104526508e-05,
|
|
"loss": 1.1481,
|
|
"step": 2037
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7216792441805483e-05,
|
|
"loss": 1.1154,
|
|
"step": 2038
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7213864488509494e-05,
|
|
"loss": 1.0717,
|
|
"step": 2039
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7210935245162137e-05,
|
|
"loss": 1.1503,
|
|
"step": 2040
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.720800471228725e-05,
|
|
"loss": 1.1501,
|
|
"step": 2041
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7205072890408904e-05,
|
|
"loss": 1.1837,
|
|
"step": 2042
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.720213978005139e-05,
|
|
"loss": 1.1734,
|
|
"step": 2043
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7199205381739237e-05,
|
|
"loss": 1.1151,
|
|
"step": 2044
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.71962696959972e-05,
|
|
"loss": 1.1412,
|
|
"step": 2045
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7193332723350265e-05,
|
|
"loss": 1.0939,
|
|
"step": 2046
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.719039446432365e-05,
|
|
"loss": 1.172,
|
|
"step": 2047
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7187454919442804e-05,
|
|
"loss": 1.1491,
|
|
"step": 2048
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7184514089233403e-05,
|
|
"loss": 1.1457,
|
|
"step": 2049
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7181571974221354e-05,
|
|
"loss": 1.1236,
|
|
"step": 2050
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7178628574932793e-05,
|
|
"loss": 1.1651,
|
|
"step": 2051
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7175683891894088e-05,
|
|
"loss": 1.1782,
|
|
"step": 2052
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7172737925631835e-05,
|
|
"loss": 1.1932,
|
|
"step": 2053
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.716979067667286e-05,
|
|
"loss": 1.1708,
|
|
"step": 2054
|
|
},
|
|
{
|
|
"epoch": 0.8,
|
|
"learning_rate": 1.7166842145544212e-05,
|
|
"loss": 1.1487,
|
|
"step": 2055
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7163892332773185e-05,
|
|
"loss": 1.1095,
|
|
"step": 2056
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.716094123888728e-05,
|
|
"loss": 1.1358,
|
|
"step": 2057
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7157988864414258e-05,
|
|
"loss": 1.2157,
|
|
"step": 2058
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.715503520988207e-05,
|
|
"loss": 1.111,
|
|
"step": 2059
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.715208027581893e-05,
|
|
"loss": 1.1671,
|
|
"step": 2060
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.714912406275326e-05,
|
|
"loss": 1.1822,
|
|
"step": 2061
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.714616657121372e-05,
|
|
"loss": 1.1703,
|
|
"step": 2062
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.71432078017292e-05,
|
|
"loss": 1.1599,
|
|
"step": 2063
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7140247754828814e-05,
|
|
"loss": 1.0851,
|
|
"step": 2064
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7137286431041898e-05,
|
|
"loss": 1.1667,
|
|
"step": 2065
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7134323830898036e-05,
|
|
"loss": 1.1658,
|
|
"step": 2066
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.713135995492702e-05,
|
|
"loss": 1.0498,
|
|
"step": 2067
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7128394803658877e-05,
|
|
"loss": 1.0716,
|
|
"step": 2068
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.712542837762387e-05,
|
|
"loss": 1.1562,
|
|
"step": 2069
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.712246067735248e-05,
|
|
"loss": 1.0478,
|
|
"step": 2070
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7119491703375417e-05,
|
|
"loss": 1.1059,
|
|
"step": 2071
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7116521456223626e-05,
|
|
"loss": 1.1773,
|
|
"step": 2072
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7113549936428268e-05,
|
|
"loss": 1.2164,
|
|
"step": 2073
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7110577144520743e-05,
|
|
"loss": 1.2007,
|
|
"step": 2074
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7107603081032674e-05,
|
|
"loss": 1.1055,
|
|
"step": 2075
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.710462774649591e-05,
|
|
"loss": 1.17,
|
|
"step": 2076
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7101651141442527e-05,
|
|
"loss": 1.0859,
|
|
"step": 2077
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.709867326640483e-05,
|
|
"loss": 1.1847,
|
|
"step": 2078
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7095694121915353e-05,
|
|
"loss": 1.216,
|
|
"step": 2079
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.709271370850685e-05,
|
|
"loss": 1.1203,
|
|
"step": 2080
|
|
},
|
|
{
|
|
"epoch": 0.81,
|
|
"learning_rate": 1.7089732026712317e-05,
|
|
"loss": 1.1139,
|
|
"step": 2081
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7086749077064952e-05,
|
|
"loss": 1.1523,
|
|
"step": 2082
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7083764860098206e-05,
|
|
"loss": 1.103,
|
|
"step": 2083
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.708077937634574e-05,
|
|
"loss": 1.2682,
|
|
"step": 2084
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7077792626341446e-05,
|
|
"loss": 1.2545,
|
|
"step": 2085
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7074804610619444e-05,
|
|
"loss": 1.1165,
|
|
"step": 2086
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.707181532971408e-05,
|
|
"loss": 1.1906,
|
|
"step": 2087
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.706882478415992e-05,
|
|
"loss": 1.0844,
|
|
"step": 2088
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.706583297449177e-05,
|
|
"loss": 1.1839,
|
|
"step": 2089
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.706283990124465e-05,
|
|
"loss": 1.1331,
|
|
"step": 2090
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7059845564953807e-05,
|
|
"loss": 1.189,
|
|
"step": 2091
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.705684996615472e-05,
|
|
"loss": 1.1224,
|
|
"step": 2092
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.705385310538309e-05,
|
|
"loss": 1.1367,
|
|
"step": 2093
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.705085498317484e-05,
|
|
"loss": 1.1894,
|
|
"step": 2094
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.704785560006613e-05,
|
|
"loss": 1.195,
|
|
"step": 2095
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.704485495659333e-05,
|
|
"loss": 1.2251,
|
|
"step": 2096
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7041853053293045e-05,
|
|
"loss": 1.1479,
|
|
"step": 2097
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7038849890702108e-05,
|
|
"loss": 1.1044,
|
|
"step": 2098
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.703584546935757e-05,
|
|
"loss": 1.1034,
|
|
"step": 2099
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7032839789796712e-05,
|
|
"loss": 1.1402,
|
|
"step": 2100
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7029832852557034e-05,
|
|
"loss": 1.0996,
|
|
"step": 2101
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7026824658176265e-05,
|
|
"loss": 1.0598,
|
|
"step": 2102
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7023815207192366e-05,
|
|
"loss": 1.1901,
|
|
"step": 2103
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7020804500143508e-05,
|
|
"loss": 1.1,
|
|
"step": 2104
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.70177925375681e-05,
|
|
"loss": 1.1757,
|
|
"step": 2105
|
|
},
|
|
{
|
|
"epoch": 0.82,
|
|
"learning_rate": 1.7014779320004763e-05,
|
|
"loss": 1.0493,
|
|
"step": 2106
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.7011764847992355e-05,
|
|
"loss": 1.1688,
|
|
"step": 2107
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.7008749122069942e-05,
|
|
"loss": 1.1317,
|
|
"step": 2108
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.7005732142776842e-05,
|
|
"loss": 1.1276,
|
|
"step": 2109
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.7002713910652563e-05,
|
|
"loss": 1.1671,
|
|
"step": 2110
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6999694426236862e-05,
|
|
"loss": 1.1653,
|
|
"step": 2111
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.699667369006971e-05,
|
|
"loss": 1.1797,
|
|
"step": 2112
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.699365170269131e-05,
|
|
"loss": 1.1129,
|
|
"step": 2113
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.699062846464207e-05,
|
|
"loss": 1.1523,
|
|
"step": 2114
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.698760397646264e-05,
|
|
"loss": 1.1519,
|
|
"step": 2115
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6984578238693888e-05,
|
|
"loss": 1.1576,
|
|
"step": 2116
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6981551251876905e-05,
|
|
"loss": 1.1746,
|
|
"step": 2117
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6978523016553006e-05,
|
|
"loss": 1.1027,
|
|
"step": 2118
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.697549353326373e-05,
|
|
"loss": 1.23,
|
|
"step": 2119
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6972462802550832e-05,
|
|
"loss": 1.1199,
|
|
"step": 2120
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.69694308249563e-05,
|
|
"loss": 1.1286,
|
|
"step": 2121
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.696639760102234e-05,
|
|
"loss": 1.1901,
|
|
"step": 2122
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6963363131291384e-05,
|
|
"loss": 1.0842,
|
|
"step": 2123
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.696032741630608e-05,
|
|
"loss": 1.1887,
|
|
"step": 2124
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.695729045660931e-05,
|
|
"loss": 1.1966,
|
|
"step": 2125
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.695425225274416e-05,
|
|
"loss": 1.1395,
|
|
"step": 2126
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6951212805253963e-05,
|
|
"loss": 1.1809,
|
|
"step": 2127
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6948172114682252e-05,
|
|
"loss": 1.0747,
|
|
"step": 2128
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6945130181572805e-05,
|
|
"loss": 1.2066,
|
|
"step": 2129
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6942087006469593e-05,
|
|
"loss": 1.2002,
|
|
"step": 2130
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6939042589916838e-05,
|
|
"loss": 1.1289,
|
|
"step": 2131
|
|
},
|
|
{
|
|
"epoch": 0.83,
|
|
"learning_rate": 1.6935996932458962e-05,
|
|
"loss": 1.1462,
|
|
"step": 2132
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6932950034640624e-05,
|
|
"loss": 1.0754,
|
|
"step": 2133
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.69299018970067e-05,
|
|
"loss": 1.1507,
|
|
"step": 2134
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6926852520102283e-05,
|
|
"loss": 1.181,
|
|
"step": 2135
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6923801904472692e-05,
|
|
"loss": 1.104,
|
|
"step": 2136
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.692075005066347e-05,
|
|
"loss": 1.2347,
|
|
"step": 2137
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6917696959220372e-05,
|
|
"loss": 1.1146,
|
|
"step": 2138
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.691464263068939e-05,
|
|
"loss": 1.1594,
|
|
"step": 2139
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6911587065616715e-05,
|
|
"loss": 1.1849,
|
|
"step": 2140
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6908530264548787e-05,
|
|
"loss": 1.1667,
|
|
"step": 2141
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.690547222803224e-05,
|
|
"loss": 1.0922,
|
|
"step": 2142
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6902412956613947e-05,
|
|
"loss": 1.1741,
|
|
"step": 2143
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6899352450840996e-05,
|
|
"loss": 1.0779,
|
|
"step": 2144
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6896290711260693e-05,
|
|
"loss": 1.2029,
|
|
"step": 2145
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6893227738420565e-05,
|
|
"loss": 1.1242,
|
|
"step": 2146
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6890163532868368e-05,
|
|
"loss": 1.1387,
|
|
"step": 2147
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6887098095152067e-05,
|
|
"loss": 1.171,
|
|
"step": 2148
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6884031425819852e-05,
|
|
"loss": 1.129,
|
|
"step": 2149
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6880963525420143e-05,
|
|
"loss": 1.1596,
|
|
"step": 2150
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.687789439450156e-05,
|
|
"loss": 1.1892,
|
|
"step": 2151
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6874824033612956e-05,
|
|
"loss": 1.0944,
|
|
"step": 2152
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6871752443303406e-05,
|
|
"loss": 1.0857,
|
|
"step": 2153
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6868679624122196e-05,
|
|
"loss": 1.2067,
|
|
"step": 2154
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6865605576618845e-05,
|
|
"loss": 1.244,
|
|
"step": 2155
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6862530301343073e-05,
|
|
"loss": 1.1551,
|
|
"step": 2156
|
|
},
|
|
{
|
|
"epoch": 0.84,
|
|
"learning_rate": 1.6859453798844838e-05,
|
|
"loss": 1.1632,
|
|
"step": 2157
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6856376069674302e-05,
|
|
"loss": 1.1779,
|
|
"step": 2158
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.685329711438186e-05,
|
|
"loss": 1.1648,
|
|
"step": 2159
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6850216933518115e-05,
|
|
"loss": 1.166,
|
|
"step": 2160
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6847135527633898e-05,
|
|
"loss": 1.1349,
|
|
"step": 2161
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.684405289728025e-05,
|
|
"loss": 1.1331,
|
|
"step": 2162
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.684096904300844e-05,
|
|
"loss": 1.1528,
|
|
"step": 2163
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6837883965369952e-05,
|
|
"loss": 1.1806,
|
|
"step": 2164
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6834797664916488e-05,
|
|
"loss": 1.1331,
|
|
"step": 2165
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6831710142199967e-05,
|
|
"loss": 1.0833,
|
|
"step": 2166
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.682862139777253e-05,
|
|
"loss": 1.1103,
|
|
"step": 2167
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6825531432186545e-05,
|
|
"loss": 1.0989,
|
|
"step": 2168
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6822440245994573e-05,
|
|
"loss": 1.1387,
|
|
"step": 2169
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6819347839749418e-05,
|
|
"loss": 1.1403,
|
|
"step": 2170
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.681625421400409e-05,
|
|
"loss": 1.0806,
|
|
"step": 2171
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6813159369311828e-05,
|
|
"loss": 1.1532,
|
|
"step": 2172
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.681006330622607e-05,
|
|
"loss": 1.1666,
|
|
"step": 2173
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.680696602530049e-05,
|
|
"loss": 1.1464,
|
|
"step": 2174
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6803867527088972e-05,
|
|
"loss": 1.1846,
|
|
"step": 2175
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.680076781214562e-05,
|
|
"loss": 1.2024,
|
|
"step": 2176
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.679766688102475e-05,
|
|
"loss": 1.1414,
|
|
"step": 2177
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6794564734280904e-05,
|
|
"loss": 1.1623,
|
|
"step": 2178
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6791461372468837e-05,
|
|
"loss": 1.085,
|
|
"step": 2179
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6788356796143516e-05,
|
|
"loss": 1.1273,
|
|
"step": 2180
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.678525100586014e-05,
|
|
"loss": 1.1424,
|
|
"step": 2181
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6782144002174107e-05,
|
|
"loss": 1.0862,
|
|
"step": 2182
|
|
},
|
|
{
|
|
"epoch": 0.85,
|
|
"learning_rate": 1.6779035785641045e-05,
|
|
"loss": 1.1945,
|
|
"step": 2183
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6775926356816794e-05,
|
|
"loss": 1.1643,
|
|
"step": 2184
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6772815716257414e-05,
|
|
"loss": 1.0687,
|
|
"step": 2185
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.676970386451917e-05,
|
|
"loss": 1.1177,
|
|
"step": 2186
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6766590802158567e-05,
|
|
"loss": 1.1652,
|
|
"step": 2187
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6763476529732302e-05,
|
|
"loss": 1.2039,
|
|
"step": 2188
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6760361047797296e-05,
|
|
"loss": 1.1267,
|
|
"step": 2189
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6757244356910698e-05,
|
|
"loss": 1.128,
|
|
"step": 2190
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6754126457629862e-05,
|
|
"loss": 1.2336,
|
|
"step": 2191
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6751007350512356e-05,
|
|
"loss": 1.1886,
|
|
"step": 2192
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.674788703611597e-05,
|
|
"loss": 1.0323,
|
|
"step": 2193
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.674476551499871e-05,
|
|
"loss": 1.1608,
|
|
"step": 2194
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.674164278771879e-05,
|
|
"loss": 1.063,
|
|
"step": 2195
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6738518854834654e-05,
|
|
"loss": 1.1085,
|
|
"step": 2196
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.673539371690495e-05,
|
|
"loss": 1.2147,
|
|
"step": 2197
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6732267374488542e-05,
|
|
"loss": 1.2245,
|
|
"step": 2198
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.672913982814451e-05,
|
|
"loss": 1.1623,
|
|
"step": 2199
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.672601107843216e-05,
|
|
"loss": 1.1761,
|
|
"step": 2200
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6722881125911e-05,
|
|
"loss": 1.1583,
|
|
"step": 2201
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6719749971140756e-05,
|
|
"loss": 1.1133,
|
|
"step": 2202
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6716617614681373e-05,
|
|
"loss": 1.1248,
|
|
"step": 2203
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6713484057093007e-05,
|
|
"loss": 1.1731,
|
|
"step": 2204
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.671034929893603e-05,
|
|
"loss": 1.122,
|
|
"step": 2205
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6707213340771028e-05,
|
|
"loss": 1.1089,
|
|
"step": 2206
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.670407618315881e-05,
|
|
"loss": 1.0649,
|
|
"step": 2207
|
|
},
|
|
{
|
|
"epoch": 0.86,
|
|
"learning_rate": 1.6700937826660384e-05,
|
|
"loss": 1.0775,
|
|
"step": 2208
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6697798271836985e-05,
|
|
"loss": 1.0743,
|
|
"step": 2209
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6694657519250052e-05,
|
|
"loss": 1.1271,
|
|
"step": 2210
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6691515569461255e-05,
|
|
"loss": 1.1687,
|
|
"step": 2211
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6688372423032456e-05,
|
|
"loss": 1.1137,
|
|
"step": 2212
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6685228080525746e-05,
|
|
"loss": 1.1062,
|
|
"step": 2213
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.668208254250343e-05,
|
|
"loss": 1.1209,
|
|
"step": 2214
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6678935809528015e-05,
|
|
"loss": 1.1399,
|
|
"step": 2215
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6675787882162238e-05,
|
|
"loss": 1.1725,
|
|
"step": 2216
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6672638760969034e-05,
|
|
"loss": 1.09,
|
|
"step": 2217
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6669488446511562e-05,
|
|
"loss": 1.0895,
|
|
"step": 2218
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.666633693935319e-05,
|
|
"loss": 1.0865,
|
|
"step": 2219
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6663184240057502e-05,
|
|
"loss": 1.127,
|
|
"step": 2220
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.666003034918829e-05,
|
|
"loss": 1.1812,
|
|
"step": 2221
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.665687526730957e-05,
|
|
"loss": 1.0979,
|
|
"step": 2222
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6653718994985553e-05,
|
|
"loss": 1.1072,
|
|
"step": 2223
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.665056153278068e-05,
|
|
"loss": 1.138,
|
|
"step": 2224
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.66474028812596e-05,
|
|
"loss": 1.1347,
|
|
"step": 2225
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6644243040987165e-05,
|
|
"loss": 1.1836,
|
|
"step": 2226
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6641082012528454e-05,
|
|
"loss": 0.9832,
|
|
"step": 2227
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6637919796448754e-05,
|
|
"loss": 1.1516,
|
|
"step": 2228
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6634756393313557e-05,
|
|
"loss": 1.107,
|
|
"step": 2229
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.663159180368857e-05,
|
|
"loss": 1.2034,
|
|
"step": 2230
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.662842602813972e-05,
|
|
"loss": 1.0975,
|
|
"step": 2231
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6625259067233145e-05,
|
|
"loss": 1.1469,
|
|
"step": 2232
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.662209092153518e-05,
|
|
"loss": 1.2399,
|
|
"step": 2233
|
|
},
|
|
{
|
|
"epoch": 0.87,
|
|
"learning_rate": 1.6618921591612393e-05,
|
|
"loss": 1.1553,
|
|
"step": 2234
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6615751078031545e-05,
|
|
"loss": 1.2042,
|
|
"step": 2235
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6612579381359624e-05,
|
|
"loss": 1.1553,
|
|
"step": 2236
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6609406502163822e-05,
|
|
"loss": 1.1392,
|
|
"step": 2237
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6606232441011537e-05,
|
|
"loss": 1.166,
|
|
"step": 2238
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.660305719847039e-05,
|
|
"loss": 1.1421,
|
|
"step": 2239
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6599880775108204e-05,
|
|
"loss": 1.2234,
|
|
"step": 2240
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6596703171493023e-05,
|
|
"loss": 1.1229,
|
|
"step": 2241
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.659352438819309e-05,
|
|
"loss": 1.1783,
|
|
"step": 2242
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6590344425776872e-05,
|
|
"loss": 1.1498,
|
|
"step": 2243
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6587163284813034e-05,
|
|
"loss": 1.1931,
|
|
"step": 2244
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6583980965870458e-05,
|
|
"loss": 1.1124,
|
|
"step": 2245
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6580797469518234e-05,
|
|
"loss": 1.2005,
|
|
"step": 2246
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6577612796325674e-05,
|
|
"loss": 1.1782,
|
|
"step": 2247
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6574426946862285e-05,
|
|
"loss": 1.1366,
|
|
"step": 2248
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.657123992169779e-05,
|
|
"loss": 1.1577,
|
|
"step": 2249
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6568051721402122e-05,
|
|
"loss": 1.1362,
|
|
"step": 2250
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.656486234654543e-05,
|
|
"loss": 1.1324,
|
|
"step": 2251
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6561671797698067e-05,
|
|
"loss": 1.1495,
|
|
"step": 2252
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6558480075430594e-05,
|
|
"loss": 1.1583,
|
|
"step": 2253
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6555287180313787e-05,
|
|
"loss": 1.0587,
|
|
"step": 2254
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.655209311291863e-05,
|
|
"loss": 1.0536,
|
|
"step": 2255
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6548897873816317e-05,
|
|
"loss": 1.1595,
|
|
"step": 2256
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6545701463578248e-05,
|
|
"loss": 1.1794,
|
|
"step": 2257
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6542503882776035e-05,
|
|
"loss": 1.1233,
|
|
"step": 2258
|
|
},
|
|
{
|
|
"epoch": 0.88,
|
|
"learning_rate": 1.6539305131981502e-05,
|
|
"loss": 1.1679,
|
|
"step": 2259
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6536105211766676e-05,
|
|
"loss": 1.0832,
|
|
"step": 2260
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6532904122703805e-05,
|
|
"loss": 1.1697,
|
|
"step": 2261
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6529701865365334e-05,
|
|
"loss": 1.1647,
|
|
"step": 2262
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6526498440323914e-05,
|
|
"loss": 1.0741,
|
|
"step": 2263
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6523293848152422e-05,
|
|
"loss": 1.2071,
|
|
"step": 2264
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6520088089423926e-05,
|
|
"loss": 1.1382,
|
|
"step": 2265
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6516881164711715e-05,
|
|
"loss": 1.1569,
|
|
"step": 2266
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.651367307458928e-05,
|
|
"loss": 1.186,
|
|
"step": 2267
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.651046381963032e-05,
|
|
"loss": 1.0931,
|
|
"step": 2268
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.650725340040875e-05,
|
|
"loss": 1.153,
|
|
"step": 2269
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6504041817498676e-05,
|
|
"loss": 1.1097,
|
|
"step": 2270
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6500829071474438e-05,
|
|
"loss": 1.1565,
|
|
"step": 2271
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.649761516291056e-05,
|
|
"loss": 1.1462,
|
|
"step": 2272
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6494400092381783e-05,
|
|
"loss": 1.1529,
|
|
"step": 2273
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6491183860463064e-05,
|
|
"loss": 1.0867,
|
|
"step": 2274
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6487966467729553e-05,
|
|
"loss": 1.1298,
|
|
"step": 2275
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6484747914756616e-05,
|
|
"loss": 1.2132,
|
|
"step": 2276
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6481528202119826e-05,
|
|
"loss": 1.1725,
|
|
"step": 2277
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6478307330394963e-05,
|
|
"loss": 1.0832,
|
|
"step": 2278
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.647508530015801e-05,
|
|
"loss": 1.114,
|
|
"step": 2279
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.647186211198517e-05,
|
|
"loss": 1.1538,
|
|
"step": 2280
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.646863776645283e-05,
|
|
"loss": 1.2422,
|
|
"step": 2281
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6465412264137612e-05,
|
|
"loss": 1.1409,
|
|
"step": 2282
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.646218560561632e-05,
|
|
"loss": 1.0904,
|
|
"step": 2283
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6458957791465985e-05,
|
|
"loss": 1.1183,
|
|
"step": 2284
|
|
},
|
|
{
|
|
"epoch": 0.89,
|
|
"learning_rate": 1.6455728822263822e-05,
|
|
"loss": 1.1655,
|
|
"step": 2285
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.645249869858728e-05,
|
|
"loss": 1.1699,
|
|
"step": 2286
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6449267421013994e-05,
|
|
"loss": 1.2029,
|
|
"step": 2287
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.644603499012181e-05,
|
|
"loss": 1.2281,
|
|
"step": 2288
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6442801406488788e-05,
|
|
"loss": 1.1581,
|
|
"step": 2289
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6439566670693178e-05,
|
|
"loss": 1.0889,
|
|
"step": 2290
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6436330783313453e-05,
|
|
"loss": 1.1355,
|
|
"step": 2291
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6433093744928283e-05,
|
|
"loss": 1.1215,
|
|
"step": 2292
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6429855556116548e-05,
|
|
"loss": 1.127,
|
|
"step": 2293
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.642661621745733e-05,
|
|
"loss": 1.1083,
|
|
"step": 2294
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6423375729529912e-05,
|
|
"loss": 1.1249,
|
|
"step": 2295
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.64201340929138e-05,
|
|
"loss": 1.1741,
|
|
"step": 2296
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6416891308188686e-05,
|
|
"loss": 1.1567,
|
|
"step": 2297
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.641364737593448e-05,
|
|
"loss": 1.1794,
|
|
"step": 2298
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6410402296731288e-05,
|
|
"loss": 1.2341,
|
|
"step": 2299
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6407156071159432e-05,
|
|
"loss": 1.2991,
|
|
"step": 2300
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6403908699799423e-05,
|
|
"loss": 1.1529,
|
|
"step": 2301
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6400660183232e-05,
|
|
"loss": 1.127,
|
|
"step": 2302
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6397410522038078e-05,
|
|
"loss": 1.1081,
|
|
"step": 2303
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6394159716798807e-05,
|
|
"loss": 1.0674,
|
|
"step": 2304
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6390907768095513e-05,
|
|
"loss": 1.0939,
|
|
"step": 2305
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.6387654676509755e-05,
|
|
"loss": 1.1179,
|
|
"step": 2306
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.638440044262327e-05,
|
|
"loss": 1.1516,
|
|
"step": 2307
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.638114506701801e-05,
|
|
"loss": 1.1567,
|
|
"step": 2308
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.637788855027614e-05,
|
|
"loss": 1.1894,
|
|
"step": 2309
|
|
},
|
|
{
|
|
"epoch": 0.9,
|
|
"learning_rate": 1.637463089298002e-05,
|
|
"loss": 1.1332,
|
|
"step": 2310
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6371372095712206e-05,
|
|
"loss": 1.1798,
|
|
"step": 2311
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6368112159055477e-05,
|
|
"loss": 1.1964,
|
|
"step": 2312
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.63648510835928e-05,
|
|
"loss": 1.1482,
|
|
"step": 2313
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6361588869907353e-05,
|
|
"loss": 1.1109,
|
|
"step": 2314
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.635832551858252e-05,
|
|
"loss": 1.0977,
|
|
"step": 2315
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6355061030201877e-05,
|
|
"loss": 1.0615,
|
|
"step": 2316
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6351795405349214e-05,
|
|
"loss": 1.0988,
|
|
"step": 2317
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6348528644608523e-05,
|
|
"loss": 1.1885,
|
|
"step": 2318
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6345260748563996e-05,
|
|
"loss": 1.0908,
|
|
"step": 2319
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.634199171780002e-05,
|
|
"loss": 1.1291,
|
|
"step": 2320
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.633872155290121e-05,
|
|
"loss": 1.1475,
|
|
"step": 2321
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.633545025445236e-05,
|
|
"loss": 1.0549,
|
|
"step": 2322
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6332177823038472e-05,
|
|
"loss": 1.1316,
|
|
"step": 2323
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.632890425924475e-05,
|
|
"loss": 1.1214,
|
|
"step": 2324
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.632562956365662e-05,
|
|
"loss": 1.1919,
|
|
"step": 2325
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6322353736859673e-05,
|
|
"loss": 1.0321,
|
|
"step": 2326
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6319076779439734e-05,
|
|
"loss": 1.1791,
|
|
"step": 2327
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6315798691982823e-05,
|
|
"loss": 1.0746,
|
|
"step": 2328
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.631251947507515e-05,
|
|
"loss": 1.1044,
|
|
"step": 2329
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6309239129303143e-05,
|
|
"loss": 1.1108,
|
|
"step": 2330
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.630595765525342e-05,
|
|
"loss": 1.1698,
|
|
"step": 2331
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6302675053512805e-05,
|
|
"loss": 1.2621,
|
|
"step": 2332
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6299391324668326e-05,
|
|
"loss": 1.1063,
|
|
"step": 2333
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6296106469307208e-05,
|
|
"loss": 1.1452,
|
|
"step": 2334
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.629282048801688e-05,
|
|
"loss": 1.1651,
|
|
"step": 2335
|
|
},
|
|
{
|
|
"epoch": 0.91,
|
|
"learning_rate": 1.6289533381384978e-05,
|
|
"loss": 1.1365,
|
|
"step": 2336
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6286245149999323e-05,
|
|
"loss": 1.1389,
|
|
"step": 2337
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.628295579444796e-05,
|
|
"loss": 1.1644,
|
|
"step": 2338
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6279665315319114e-05,
|
|
"loss": 1.1788,
|
|
"step": 2339
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6276373713201218e-05,
|
|
"loss": 1.1614,
|
|
"step": 2340
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6273080988682918e-05,
|
|
"loss": 1.1116,
|
|
"step": 2341
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6269787142353037e-05,
|
|
"loss": 1.1766,
|
|
"step": 2342
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.626649217480062e-05,
|
|
"loss": 1.1314,
|
|
"step": 2343
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6263196086614902e-05,
|
|
"loss": 1.1416,
|
|
"step": 2344
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6259898878385317e-05,
|
|
"loss": 1.0906,
|
|
"step": 2345
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6256600550701513e-05,
|
|
"loss": 1.0738,
|
|
"step": 2346
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.625330110415332e-05,
|
|
"loss": 1.1119,
|
|
"step": 2347
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6250000539330775e-05,
|
|
"loss": 1.0592,
|
|
"step": 2348
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6246698856824122e-05,
|
|
"loss": 1.0847,
|
|
"step": 2349
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6243396057223794e-05,
|
|
"loss": 1.1269,
|
|
"step": 2350
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.624009214112043e-05,
|
|
"loss": 1.1095,
|
|
"step": 2351
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.623678710910487e-05,
|
|
"loss": 1.145,
|
|
"step": 2352
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.623348096176815e-05,
|
|
"loss": 1.0812,
|
|
"step": 2353
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6230173699701506e-05,
|
|
"loss": 1.1902,
|
|
"step": 2354
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6226865323496373e-05,
|
|
"loss": 1.0855,
|
|
"step": 2355
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6223555833744385e-05,
|
|
"loss": 1.0718,
|
|
"step": 2356
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6220245231037378e-05,
|
|
"loss": 1.1574,
|
|
"step": 2357
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.621693351596739e-05,
|
|
"loss": 1.0892,
|
|
"step": 2358
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6213620689126644e-05,
|
|
"loss": 1.0733,
|
|
"step": 2359
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6210306751107577e-05,
|
|
"loss": 1.0769,
|
|
"step": 2360
|
|
},
|
|
{
|
|
"epoch": 0.92,
|
|
"learning_rate": 1.6206991702502815e-05,
|
|
"loss": 1.0683,
|
|
"step": 2361
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6203675543905192e-05,
|
|
"loss": 1.1161,
|
|
"step": 2362
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6200358275907725e-05,
|
|
"loss": 1.1683,
|
|
"step": 2363
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.619703989910365e-05,
|
|
"loss": 1.2341,
|
|
"step": 2364
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6193720414086387e-05,
|
|
"loss": 1.1464,
|
|
"step": 2365
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6190399821449554e-05,
|
|
"loss": 1.1216,
|
|
"step": 2366
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6187078121786976e-05,
|
|
"loss": 1.108,
|
|
"step": 2367
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6183755315692662e-05,
|
|
"loss": 1.1637,
|
|
"step": 2368
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6180431403760836e-05,
|
|
"loss": 1.2141,
|
|
"step": 2369
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.617710638658591e-05,
|
|
"loss": 1.1382,
|
|
"step": 2370
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6173780264762492e-05,
|
|
"loss": 1.1487,
|
|
"step": 2371
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6170453038885394e-05,
|
|
"loss": 1.1997,
|
|
"step": 2372
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.616712470954962e-05,
|
|
"loss": 1.0717,
|
|
"step": 2373
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6163795277350373e-05,
|
|
"loss": 1.1044,
|
|
"step": 2374
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6160464742883054e-05,
|
|
"loss": 1.1958,
|
|
"step": 2375
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.615713310674326e-05,
|
|
"loss": 1.1156,
|
|
"step": 2376
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6153800369526788e-05,
|
|
"loss": 1.1079,
|
|
"step": 2377
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6150466531829623e-05,
|
|
"loss": 1.0362,
|
|
"step": 2378
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6147131594247963e-05,
|
|
"loss": 1.1719,
|
|
"step": 2379
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6143795557378185e-05,
|
|
"loss": 1.0849,
|
|
"step": 2380
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6140458421816876e-05,
|
|
"loss": 1.1318,
|
|
"step": 2381
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6137120188160813e-05,
|
|
"loss": 1.0738,
|
|
"step": 2382
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6133780857006963e-05,
|
|
"loss": 1.163,
|
|
"step": 2383
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.613044042895251e-05,
|
|
"loss": 1.0877,
|
|
"step": 2384
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.612709890459481e-05,
|
|
"loss": 1.0802,
|
|
"step": 2385
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.6123756284531434e-05,
|
|
"loss": 1.1383,
|
|
"step": 2386
|
|
},
|
|
{
|
|
"epoch": 0.93,
|
|
"learning_rate": 1.612041256936013e-05,
|
|
"loss": 1.2022,
|
|
"step": 2387
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6117067759678865e-05,
|
|
"loss": 1.1275,
|
|
"step": 2388
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6113721856085783e-05,
|
|
"loss": 1.1609,
|
|
"step": 2389
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.611037485917923e-05,
|
|
"loss": 1.1449,
|
|
"step": 2390
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.610702676955775e-05,
|
|
"loss": 1.0851,
|
|
"step": 2391
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6103677587820076e-05,
|
|
"loss": 1.1194,
|
|
"step": 2392
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6100327314565143e-05,
|
|
"loss": 1.101,
|
|
"step": 2393
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.609697595039208e-05,
|
|
"loss": 1.1482,
|
|
"step": 2394
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6093623495900207e-05,
|
|
"loss": 1.0751,
|
|
"step": 2395
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.609026995168904e-05,
|
|
"loss": 1.2221,
|
|
"step": 2396
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6086915318358297e-05,
|
|
"loss": 1.1023,
|
|
"step": 2397
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6083559596507884e-05,
|
|
"loss": 1.1781,
|
|
"step": 2398
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6080202786737897e-05,
|
|
"loss": 1.0537,
|
|
"step": 2399
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6076844889648637e-05,
|
|
"loss": 1.2539,
|
|
"step": 2400
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6073485905840595e-05,
|
|
"loss": 1.1309,
|
|
"step": 2401
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6070125835914456e-05,
|
|
"loss": 1.1074,
|
|
"step": 2402
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.60667646804711e-05,
|
|
"loss": 1.0992,
|
|
"step": 2403
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.60634024401116e-05,
|
|
"loss": 1.1456,
|
|
"step": 2404
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6060039115437216e-05,
|
|
"loss": 1.2208,
|
|
"step": 2405
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6056674707049423e-05,
|
|
"loss": 1.1013,
|
|
"step": 2406
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6053309215549867e-05,
|
|
"loss": 1.0986,
|
|
"step": 2407
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.60499426415404e-05,
|
|
"loss": 1.1097,
|
|
"step": 2408
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6046574985623067e-05,
|
|
"loss": 1.1975,
|
|
"step": 2409
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6043206248400102e-05,
|
|
"loss": 1.0767,
|
|
"step": 2410
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6039836430473928e-05,
|
|
"loss": 1.1843,
|
|
"step": 2411
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.603646553244718e-05,
|
|
"loss": 1.1307,
|
|
"step": 2412
|
|
},
|
|
{
|
|
"epoch": 0.94,
|
|
"learning_rate": 1.6033093554922663e-05,
|
|
"loss": 1.2036,
|
|
"step": 2413
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.602972049850339e-05,
|
|
"loss": 1.1001,
|
|
"step": 2414
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.6026346363792565e-05,
|
|
"loss": 1.0988,
|
|
"step": 2415
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.602297115139358e-05,
|
|
"loss": 1.0378,
|
|
"step": 2416
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.6019594861910022e-05,
|
|
"loss": 1.1646,
|
|
"step": 2417
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.6016217495945672e-05,
|
|
"loss": 1.0679,
|
|
"step": 2418
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.6012839054104504e-05,
|
|
"loss": 1.1381,
|
|
"step": 2419
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.6009459536990677e-05,
|
|
"loss": 1.1163,
|
|
"step": 2420
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.6006078945208557e-05,
|
|
"loss": 1.128,
|
|
"step": 2421
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.6002697279362688e-05,
|
|
"loss": 1.1091,
|
|
"step": 2422
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.599931454005781e-05,
|
|
"loss": 1.1561,
|
|
"step": 2423
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.599593072789886e-05,
|
|
"loss": 1.1466,
|
|
"step": 2424
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.599254584349096e-05,
|
|
"loss": 1.1471,
|
|
"step": 2425
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5989159887439428e-05,
|
|
"loss": 1.1331,
|
|
"step": 2426
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5985772860349778e-05,
|
|
"loss": 1.0893,
|
|
"step": 2427
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5982384762827702e-05,
|
|
"loss": 1.072,
|
|
"step": 2428
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5978995595479094e-05,
|
|
"loss": 1.0966,
|
|
"step": 2429
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.597560535891004e-05,
|
|
"loss": 1.1075,
|
|
"step": 2430
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.597221405372681e-05,
|
|
"loss": 1.1326,
|
|
"step": 2431
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5968821680535875e-05,
|
|
"loss": 1.1042,
|
|
"step": 2432
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5965428239943883e-05,
|
|
"loss": 1.148,
|
|
"step": 2433
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5962033732557685e-05,
|
|
"loss": 1.1111,
|
|
"step": 2434
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5958638158984324e-05,
|
|
"loss": 1.0993,
|
|
"step": 2435
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5955241519831023e-05,
|
|
"loss": 1.1221,
|
|
"step": 2436
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.5951843815705203e-05,
|
|
"loss": 1.1241,
|
|
"step": 2437
|
|
},
|
|
{
|
|
"epoch": 0.95,
|
|
"learning_rate": 1.594844504721447e-05,
|
|
"loss": 1.0287,
|
|
"step": 2438
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.594504521496663e-05,
|
|
"loss": 1.1105,
|
|
"step": 2439
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5941644319569665e-05,
|
|
"loss": 1.0778,
|
|
"step": 2440
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5938242361631763e-05,
|
|
"loss": 1.1074,
|
|
"step": 2441
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.593483934176129e-05,
|
|
"loss": 1.0837,
|
|
"step": 2442
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5931435260566807e-05,
|
|
"loss": 1.1408,
|
|
"step": 2443
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5928030118657067e-05,
|
|
"loss": 1.0887,
|
|
"step": 2444
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5924623916641003e-05,
|
|
"loss": 1.0842,
|
|
"step": 2445
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.592121665512775e-05,
|
|
"loss": 1.0741,
|
|
"step": 2446
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5917808334726625e-05,
|
|
"loss": 1.0995,
|
|
"step": 2447
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.591439895604713e-05,
|
|
"loss": 1.1245,
|
|
"step": 2448
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5910988519698977e-05,
|
|
"loss": 1.0885,
|
|
"step": 2449
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.590757702629204e-05,
|
|
"loss": 1.116,
|
|
"step": 2450
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5904164476436396e-05,
|
|
"loss": 1.1151,
|
|
"step": 2451
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.590075087074231e-05,
|
|
"loss": 1.1608,
|
|
"step": 2452
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.589733620982024e-05,
|
|
"loss": 1.2164,
|
|
"step": 2453
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5893920494280823e-05,
|
|
"loss": 1.1558,
|
|
"step": 2454
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5890503724734893e-05,
|
|
"loss": 1.0893,
|
|
"step": 2455
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5887085901793463e-05,
|
|
"loss": 1.0778,
|
|
"step": 2456
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5883667026067745e-05,
|
|
"loss": 1.117,
|
|
"step": 2457
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5880247098169136e-05,
|
|
"loss": 1.1929,
|
|
"step": 2458
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5876826118709216e-05,
|
|
"loss": 1.1543,
|
|
"step": 2459
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.587340408829976e-05,
|
|
"loss": 1.1105,
|
|
"step": 2460
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5869981007552725e-05,
|
|
"loss": 1.1253,
|
|
"step": 2461
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.586655687708026e-05,
|
|
"loss": 1.1166,
|
|
"step": 2462
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.5863131697494704e-05,
|
|
"loss": 1.1484,
|
|
"step": 2463
|
|
},
|
|
{
|
|
"epoch": 0.96,
|
|
"learning_rate": 1.585970546940857e-05,
|
|
"loss": 1.1644,
|
|
"step": 2464
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.585627819343458e-05,
|
|
"loss": 1.102,
|
|
"step": 2465
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5852849870185623e-05,
|
|
"loss": 1.1198,
|
|
"step": 2466
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.584942050027479e-05,
|
|
"loss": 1.0892,
|
|
"step": 2467
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.584599008431535e-05,
|
|
"loss": 1.0761,
|
|
"step": 2468
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5842558622920764e-05,
|
|
"loss": 1.069,
|
|
"step": 2469
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5839126116704674e-05,
|
|
"loss": 1.0773,
|
|
"step": 2470
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.583569256628092e-05,
|
|
"loss": 1.1856,
|
|
"step": 2471
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5832257972263523e-05,
|
|
"loss": 1.1189,
|
|
"step": 2472
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.582882233526668e-05,
|
|
"loss": 1.148,
|
|
"step": 2473
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.582538565590479e-05,
|
|
"loss": 1.053,
|
|
"step": 2474
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5821947934792432e-05,
|
|
"loss": 1.0881,
|
|
"step": 2475
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.581850917254437e-05,
|
|
"loss": 1.1143,
|
|
"step": 2476
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5815069369775563e-05,
|
|
"loss": 1.1083,
|
|
"step": 2477
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.581162852710114e-05,
|
|
"loss": 1.0672,
|
|
"step": 2478
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5808186645136428e-05,
|
|
"loss": 1.1093,
|
|
"step": 2479
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5804743724496935e-05,
|
|
"loss": 1.1464,
|
|
"step": 2480
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5801299765798362e-05,
|
|
"loss": 1.1383,
|
|
"step": 2481
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.579785476965659e-05,
|
|
"loss": 1.0591,
|
|
"step": 2482
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5794408736687676e-05,
|
|
"loss": 1.1525,
|
|
"step": 2483
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5790961667507884e-05,
|
|
"loss": 1.0436,
|
|
"step": 2484
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5787513562733646e-05,
|
|
"loss": 1.1036,
|
|
"step": 2485
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5784064422981584e-05,
|
|
"loss": 1.0983,
|
|
"step": 2486
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5780614248868506e-05,
|
|
"loss": 1.1451,
|
|
"step": 2487
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5777163041011405e-05,
|
|
"loss": 1.2057,
|
|
"step": 2488
|
|
},
|
|
{
|
|
"epoch": 0.97,
|
|
"learning_rate": 1.5773710800027462e-05,
|
|
"loss": 1.143,
|
|
"step": 2489
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5770257526534036e-05,
|
|
"loss": 1.1849,
|
|
"step": 2490
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5766803221148676e-05,
|
|
"loss": 1.1709,
|
|
"step": 2491
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5763347884489107e-05,
|
|
"loss": 1.0892,
|
|
"step": 2492
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5759891517173252e-05,
|
|
"loss": 1.174,
|
|
"step": 2493
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.575643411981921e-05,
|
|
"loss": 1.1239,
|
|
"step": 2494
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5752975693045262e-05,
|
|
"loss": 1.1203,
|
|
"step": 2495
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5749516237469884e-05,
|
|
"loss": 1.0999,
|
|
"step": 2496
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.574605575371172e-05,
|
|
"loss": 1.1606,
|
|
"step": 2497
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5742594242389612e-05,
|
|
"loss": 1.0981,
|
|
"step": 2498
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5739131704122573e-05,
|
|
"loss": 1.1069,
|
|
"step": 2499
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.573566813952982e-05,
|
|
"loss": 1.0656,
|
|
"step": 2500
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.573220354923073e-05,
|
|
"loss": 1.1617,
|
|
"step": 2501
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.572873793384487e-05,
|
|
"loss": 1.1804,
|
|
"step": 2502
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5725271293992004e-05,
|
|
"loss": 1.1766,
|
|
"step": 2503
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5721803630292067e-05,
|
|
"loss": 1.1743,
|
|
"step": 2504
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5718334943365176e-05,
|
|
"loss": 1.187,
|
|
"step": 2505
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5714865233831636e-05,
|
|
"loss": 1.0753,
|
|
"step": 2506
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5711394502311934e-05,
|
|
"loss": 1.1728,
|
|
"step": 2507
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5707922749426735e-05,
|
|
"loss": 1.1456,
|
|
"step": 2508
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.57044499757969e-05,
|
|
"loss": 1.1124,
|
|
"step": 2509
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5700976182043452e-05,
|
|
"loss": 1.1296,
|
|
"step": 2510
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5697501368787618e-05,
|
|
"loss": 1.0837,
|
|
"step": 2511
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5694025536650786e-05,
|
|
"loss": 1.1706,
|
|
"step": 2512
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5690548686254547e-05,
|
|
"loss": 1.1418,
|
|
"step": 2513
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.5687070818220656e-05,
|
|
"loss": 1.1672,
|
|
"step": 2514
|
|
},
|
|
{
|
|
"epoch": 0.98,
|
|
"learning_rate": 1.568359193317106e-05,
|
|
"loss": 1.1824,
|
|
"step": 2515
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5680112031727895e-05,
|
|
"loss": 1.1805,
|
|
"step": 2516
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.567663111451346e-05,
|
|
"loss": 1.1146,
|
|
"step": 2517
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5673149182150246e-05,
|
|
"loss": 1.155,
|
|
"step": 2518
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5669666235260928e-05,
|
|
"loss": 1.07,
|
|
"step": 2519
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5666182274468365e-05,
|
|
"loss": 1.1258,
|
|
"step": 2520
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.566269730039558e-05,
|
|
"loss": 1.1016,
|
|
"step": 2521
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5659211313665798e-05,
|
|
"loss": 1.1406,
|
|
"step": 2522
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.565572431490241e-05,
|
|
"loss": 1.1336,
|
|
"step": 2523
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5652236304728997e-05,
|
|
"loss": 1.1304,
|
|
"step": 2524
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.564874728376932e-05,
|
|
"loss": 1.089,
|
|
"step": 2525
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5645257252647316e-05,
|
|
"loss": 1.1582,
|
|
"step": 2526
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5641766211987102e-05,
|
|
"loss": 1.1433,
|
|
"step": 2527
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5638274162412985e-05,
|
|
"loss": 1.152,
|
|
"step": 2528
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5634781104549442e-05,
|
|
"loss": 1.1281,
|
|
"step": 2529
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5631287039021143e-05,
|
|
"loss": 1.1016,
|
|
"step": 2530
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5627791966452916e-05,
|
|
"loss": 1.1253,
|
|
"step": 2531
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.562429588746979e-05,
|
|
"loss": 1.1254,
|
|
"step": 2532
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5620798802696975e-05,
|
|
"loss": 1.146,
|
|
"step": 2533
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5617300712759838e-05,
|
|
"loss": 1.0991,
|
|
"step": 2534
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.561380161828395e-05,
|
|
"loss": 1.2429,
|
|
"step": 2535
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5610301519895045e-05,
|
|
"loss": 1.154,
|
|
"step": 2536
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5606800418219053e-05,
|
|
"loss": 1.1492,
|
|
"step": 2537
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5603298313882067e-05,
|
|
"loss": 1.0748,
|
|
"step": 2538
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.559979520751037e-05,
|
|
"loss": 1.153,
|
|
"step": 2539
|
|
},
|
|
{
|
|
"epoch": 0.99,
|
|
"learning_rate": 1.5596291099730415e-05,
|
|
"loss": 1.0926,
|
|
"step": 2540
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5592785991168844e-05,
|
|
"loss": 1.1273,
|
|
"step": 2541
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5589279882452476e-05,
|
|
"loss": 1.1907,
|
|
"step": 2542
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.55857727742083e-05,
|
|
"loss": 1.1435,
|
|
"step": 2543
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.55822646670635e-05,
|
|
"loss": 1.1222,
|
|
"step": 2544
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5578755561645417e-05,
|
|
"loss": 1.0486,
|
|
"step": 2545
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.557524545858159e-05,
|
|
"loss": 1.1527,
|
|
"step": 2546
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5571734358499727e-05,
|
|
"loss": 1.1173,
|
|
"step": 2547
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5568222262027716e-05,
|
|
"loss": 1.1444,
|
|
"step": 2548
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5564709169793624e-05,
|
|
"loss": 1.144,
|
|
"step": 2549
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5561195082425693e-05,
|
|
"loss": 1.1257,
|
|
"step": 2550
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5557680000552347e-05,
|
|
"loss": 1.1715,
|
|
"step": 2551
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5554163924802183e-05,
|
|
"loss": 1.1069,
|
|
"step": 2552
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5550646855803986e-05,
|
|
"loss": 1.1165,
|
|
"step": 2553
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5547128794186706e-05,
|
|
"loss": 0.8494,
|
|
"step": 2554
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.554360974057948e-05,
|
|
"loss": 0.8414,
|
|
"step": 2555
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.554008969561161e-05,
|
|
"loss": 0.8299,
|
|
"step": 2556
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.553656865991259e-05,
|
|
"loss": 0.8263,
|
|
"step": 2557
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5533046634112084e-05,
|
|
"loss": 0.8794,
|
|
"step": 2558
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5529523618839937e-05,
|
|
"loss": 0.9325,
|
|
"step": 2559
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5525999614726166e-05,
|
|
"loss": 0.9027,
|
|
"step": 2560
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5522474622400966e-05,
|
|
"loss": 0.9166,
|
|
"step": 2561
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5518948642494703e-05,
|
|
"loss": 0.8366,
|
|
"step": 2562
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5515421675637936e-05,
|
|
"loss": 0.7578,
|
|
"step": 2563
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5511893722461388e-05,
|
|
"loss": 0.8289,
|
|
"step": 2564
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5508364783595957e-05,
|
|
"loss": 0.8803,
|
|
"step": 2565
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"learning_rate": 1.5504834859672723e-05,
|
|
"loss": 0.7657,
|
|
"step": 2566
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5501303951322942e-05,
|
|
"loss": 0.8506,
|
|
"step": 2567
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5497772059178044e-05,
|
|
"loss": 0.7919,
|
|
"step": 2568
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5494239183869635e-05,
|
|
"loss": 0.8418,
|
|
"step": 2569
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5490705326029494e-05,
|
|
"loss": 0.8351,
|
|
"step": 2570
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5487170486289585e-05,
|
|
"loss": 0.8389,
|
|
"step": 2571
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5483634665282037e-05,
|
|
"loss": 0.8658,
|
|
"step": 2572
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.548009786363916e-05,
|
|
"loss": 0.8722,
|
|
"step": 2573
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5476560081993437e-05,
|
|
"loss": 0.8141,
|
|
"step": 2574
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5473021320977528e-05,
|
|
"loss": 0.8639,
|
|
"step": 2575
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5469481581224274e-05,
|
|
"loss": 0.777,
|
|
"step": 2576
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.546594086336667e-05,
|
|
"loss": 0.7939,
|
|
"step": 2577
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5462399168037916e-05,
|
|
"loss": 0.8678,
|
|
"step": 2578
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.545885649587136e-05,
|
|
"loss": 0.7958,
|
|
"step": 2579
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5455312847500547e-05,
|
|
"loss": 0.8522,
|
|
"step": 2580
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5451768223559177e-05,
|
|
"loss": 0.7934,
|
|
"step": 2581
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5448222624681136e-05,
|
|
"loss": 0.857,
|
|
"step": 2582
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.544467605150048e-05,
|
|
"loss": 0.8635,
|
|
"step": 2583
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5441128504651442e-05,
|
|
"loss": 0.7778,
|
|
"step": 2584
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.543757998476843e-05,
|
|
"loss": 0.741,
|
|
"step": 2585
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5434030492486023e-05,
|
|
"loss": 0.8403,
|
|
"step": 2586
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.543048002843897e-05,
|
|
"loss": 0.8231,
|
|
"step": 2587
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.54269285932622e-05,
|
|
"loss": 0.8482,
|
|
"step": 2588
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.542337618759082e-05,
|
|
"loss": 0.8588,
|
|
"step": 2589
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5419822812060103e-05,
|
|
"loss": 0.7903,
|
|
"step": 2590
|
|
},
|
|
{
|
|
"epoch": 1.01,
|
|
"learning_rate": 1.5416268467305494e-05,
|
|
"loss": 0.7457,
|
|
"step": 2591
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5412713153962612e-05,
|
|
"loss": 0.8581,
|
|
"step": 2592
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.540915687266726e-05,
|
|
"loss": 0.7993,
|
|
"step": 2593
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5405599624055394e-05,
|
|
"loss": 0.7856,
|
|
"step": 2594
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5402041408763167e-05,
|
|
"loss": 0.8117,
|
|
"step": 2595
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5398482227426885e-05,
|
|
"loss": 0.8045,
|
|
"step": 2596
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5394922080683037e-05,
|
|
"loss": 0.8097,
|
|
"step": 2597
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5391360969168285e-05,
|
|
"loss": 0.8593,
|
|
"step": 2598
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5387798893519454e-05,
|
|
"loss": 0.7688,
|
|
"step": 2599
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5384235854373555e-05,
|
|
"loss": 0.8466,
|
|
"step": 2600
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5380671852367758e-05,
|
|
"loss": 0.891,
|
|
"step": 2601
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5377106888139414e-05,
|
|
"loss": 0.7881,
|
|
"step": 2602
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5373540962326045e-05,
|
|
"loss": 0.8596,
|
|
"step": 2603
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5369974075565343e-05,
|
|
"loss": 0.8088,
|
|
"step": 2604
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5366406228495173e-05,
|
|
"loss": 0.8143,
|
|
"step": 2605
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5362837421753567e-05,
|
|
"loss": 0.8487,
|
|
"step": 2606
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5359267655978737e-05,
|
|
"loss": 0.7565,
|
|
"step": 2607
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5355696931809058e-05,
|
|
"loss": 0.8534,
|
|
"step": 2608
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5352125249883088e-05,
|
|
"loss": 0.8806,
|
|
"step": 2609
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.534855261083954e-05,
|
|
"loss": 0.7981,
|
|
"step": 2610
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5344979015317316e-05,
|
|
"loss": 0.8546,
|
|
"step": 2611
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5341404463955473e-05,
|
|
"loss": 0.9064,
|
|
"step": 2612
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5337828957393247e-05,
|
|
"loss": 0.8462,
|
|
"step": 2613
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5334252496270046e-05,
|
|
"loss": 0.8629,
|
|
"step": 2614
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.533067508122545e-05,
|
|
"loss": 0.8001,
|
|
"step": 2615
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.5327096712899195e-05,
|
|
"loss": 0.8958,
|
|
"step": 2616
|
|
},
|
|
{
|
|
"epoch": 1.02,
|
|
"learning_rate": 1.532351739193121e-05,
|
|
"loss": 0.8272,
|
|
"step": 2617
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5319937118961576e-05,
|
|
"loss": 0.8674,
|
|
"step": 2618
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.531635589463056e-05,
|
|
"loss": 0.85,
|
|
"step": 2619
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5312773719578578e-05,
|
|
"loss": 0.7697,
|
|
"step": 2620
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5309190594446233e-05,
|
|
"loss": 0.8515,
|
|
"step": 2621
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.53056065198743e-05,
|
|
"loss": 0.8447,
|
|
"step": 2622
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5302021496503708e-05,
|
|
"loss": 0.7697,
|
|
"step": 2623
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5298435524975572e-05,
|
|
"loss": 0.8345,
|
|
"step": 2624
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5294848605931164e-05,
|
|
"loss": 0.8192,
|
|
"step": 2625
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5291260740011935e-05,
|
|
"loss": 0.8234,
|
|
"step": 2626
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5287671927859494e-05,
|
|
"loss": 0.9262,
|
|
"step": 2627
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5284082170115636e-05,
|
|
"loss": 0.8228,
|
|
"step": 2628
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5280491467422307e-05,
|
|
"loss": 0.846,
|
|
"step": 2629
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5276899820421632e-05,
|
|
"loss": 0.8157,
|
|
"step": 2630
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.527330722975591e-05,
|
|
"loss": 0.8622,
|
|
"step": 2631
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5269713696067593e-05,
|
|
"loss": 0.8141,
|
|
"step": 2632
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5266119219999317e-05,
|
|
"loss": 0.8603,
|
|
"step": 2633
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5262523802193877e-05,
|
|
"loss": 0.8466,
|
|
"step": 2634
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5258927443294237e-05,
|
|
"loss": 0.8054,
|
|
"step": 2635
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5255330143943537e-05,
|
|
"loss": 0.8215,
|
|
"step": 2636
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5251731904785078e-05,
|
|
"loss": 0.8966,
|
|
"step": 2637
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5248132726462332e-05,
|
|
"loss": 0.8269,
|
|
"step": 2638
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5244532609618938e-05,
|
|
"loss": 0.8229,
|
|
"step": 2639
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5240931554898702e-05,
|
|
"loss": 0.8442,
|
|
"step": 2640
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.52373295629456e-05,
|
|
"loss": 0.7986,
|
|
"step": 2641
|
|
},
|
|
{
|
|
"epoch": 1.03,
|
|
"learning_rate": 1.5233726634403771e-05,
|
|
"loss": 0.8583,
|
|
"step": 2642
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5230122769917528e-05,
|
|
"loss": 0.8886,
|
|
"step": 2643
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5226517970131345e-05,
|
|
"loss": 0.7667,
|
|
"step": 2644
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5222912235689868e-05,
|
|
"loss": 0.7874,
|
|
"step": 2645
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.521930556723791e-05,
|
|
"loss": 0.8514,
|
|
"step": 2646
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5215697965420448e-05,
|
|
"loss": 0.8333,
|
|
"step": 2647
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5212089430882626e-05,
|
|
"loss": 0.8078,
|
|
"step": 2648
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5208479964269755e-05,
|
|
"loss": 0.7745,
|
|
"step": 2649
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5204869566227323e-05,
|
|
"loss": 0.8535,
|
|
"step": 2650
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5201258237400965e-05,
|
|
"loss": 0.7781,
|
|
"step": 2651
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5197645978436502e-05,
|
|
"loss": 0.8476,
|
|
"step": 2652
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5194032789979903e-05,
|
|
"loss": 0.7708,
|
|
"step": 2653
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5190418672677316e-05,
|
|
"loss": 0.8288,
|
|
"step": 2654
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5186803627175053e-05,
|
|
"loss": 0.8463,
|
|
"step": 2655
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5183187654119591e-05,
|
|
"loss": 0.8514,
|
|
"step": 2656
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5179570754157571e-05,
|
|
"loss": 0.8847,
|
|
"step": 2657
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5175952927935802e-05,
|
|
"loss": 0.8312,
|
|
"step": 2658
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5172334176101258e-05,
|
|
"loss": 0.8248,
|
|
"step": 2659
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5168714499301076e-05,
|
|
"loss": 0.7993,
|
|
"step": 2660
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.516509389818256e-05,
|
|
"loss": 0.8206,
|
|
"step": 2661
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5161472373393186e-05,
|
|
"loss": 0.7609,
|
|
"step": 2662
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5157849925580587e-05,
|
|
"loss": 0.8761,
|
|
"step": 2663
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5154226555392562e-05,
|
|
"loss": 0.8708,
|
|
"step": 2664
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5150602263477075e-05,
|
|
"loss": 0.8672,
|
|
"step": 2665
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5146977050482257e-05,
|
|
"loss": 0.8282,
|
|
"step": 2666
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5143350917056405e-05,
|
|
"loss": 0.8822,
|
|
"step": 2667
|
|
},
|
|
{
|
|
"epoch": 1.04,
|
|
"learning_rate": 1.5139723863847977e-05,
|
|
"loss": 0.8322,
|
|
"step": 2668
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5136095891505596e-05,
|
|
"loss": 0.8678,
|
|
"step": 2669
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5132467000678049e-05,
|
|
"loss": 0.7847,
|
|
"step": 2670
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5128837192014295e-05,
|
|
"loss": 0.8095,
|
|
"step": 2671
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5125206466163441e-05,
|
|
"loss": 0.8863,
|
|
"step": 2672
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5121574823774778e-05,
|
|
"loss": 0.7988,
|
|
"step": 2673
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.511794226549774e-05,
|
|
"loss": 0.8668,
|
|
"step": 2674
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5114308791981944e-05,
|
|
"loss": 0.8365,
|
|
"step": 2675
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5110674403877159e-05,
|
|
"loss": 0.8035,
|
|
"step": 2676
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.510703910183332e-05,
|
|
"loss": 0.7807,
|
|
"step": 2677
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5103402886500526e-05,
|
|
"loss": 0.8625,
|
|
"step": 2678
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5099765758529041e-05,
|
|
"loss": 0.8446,
|
|
"step": 2679
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.509612771856929e-05,
|
|
"loss": 0.8836,
|
|
"step": 2680
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5092488767271858e-05,
|
|
"loss": 0.8116,
|
|
"step": 2681
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5088848905287501e-05,
|
|
"loss": 0.7812,
|
|
"step": 2682
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5085208133267134e-05,
|
|
"loss": 0.8933,
|
|
"step": 2683
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5081566451861828e-05,
|
|
"loss": 0.8055,
|
|
"step": 2684
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5077923861722833e-05,
|
|
"loss": 0.7627,
|
|
"step": 2685
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.507428036350154e-05,
|
|
"loss": 0.794,
|
|
"step": 2686
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5070635957849524e-05,
|
|
"loss": 0.795,
|
|
"step": 2687
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5066990645418505e-05,
|
|
"loss": 0.8656,
|
|
"step": 2688
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5063344426860376e-05,
|
|
"loss": 0.7583,
|
|
"step": 2689
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5059697302827183e-05,
|
|
"loss": 0.8491,
|
|
"step": 2690
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.505604927397115e-05,
|
|
"loss": 0.7707,
|
|
"step": 2691
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5052400340944639e-05,
|
|
"loss": 0.7959,
|
|
"step": 2692
|
|
},
|
|
{
|
|
"epoch": 1.05,
|
|
"learning_rate": 1.5048750504400196e-05,
|
|
"loss": 0.8457,
|
|
"step": 2693
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5045099764990512e-05,
|
|
"loss": 0.844,
|
|
"step": 2694
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5041448123368454e-05,
|
|
"loss": 0.8105,
|
|
"step": 2695
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.503779558018704e-05,
|
|
"loss": 0.7997,
|
|
"step": 2696
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.503414213609945e-05,
|
|
"loss": 0.8271,
|
|
"step": 2697
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.503048779175903e-05,
|
|
"loss": 0.819,
|
|
"step": 2698
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5026832547819283e-05,
|
|
"loss": 0.8104,
|
|
"step": 2699
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5023176404933875e-05,
|
|
"loss": 0.8572,
|
|
"step": 2700
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5019519363756632e-05,
|
|
"loss": 0.8042,
|
|
"step": 2701
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.501586142494154e-05,
|
|
"loss": 0.8753,
|
|
"step": 2702
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5012202589142745e-05,
|
|
"loss": 0.788,
|
|
"step": 2703
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5008542857014562e-05,
|
|
"loss": 0.855,
|
|
"step": 2704
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5004882229211446e-05,
|
|
"loss": 0.7966,
|
|
"step": 2705
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.5001220706388035e-05,
|
|
"loss": 0.8139,
|
|
"step": 2706
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4997558289199114e-05,
|
|
"loss": 0.8101,
|
|
"step": 2707
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4993894978299634e-05,
|
|
"loss": 0.8166,
|
|
"step": 2708
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4990230774344694e-05,
|
|
"loss": 0.8517,
|
|
"step": 2709
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4986565677989573e-05,
|
|
"loss": 0.8264,
|
|
"step": 2710
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.498289968988969e-05,
|
|
"loss": 0.7919,
|
|
"step": 2711
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4979232810700638e-05,
|
|
"loss": 0.8223,
|
|
"step": 2712
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4975565041078152e-05,
|
|
"loss": 0.7857,
|
|
"step": 2713
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4971896381678152e-05,
|
|
"loss": 0.813,
|
|
"step": 2714
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4968226833156689e-05,
|
|
"loss": 0.8401,
|
|
"step": 2715
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4964556396169997e-05,
|
|
"loss": 0.8081,
|
|
"step": 2716
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4960885071374449e-05,
|
|
"loss": 0.8815,
|
|
"step": 2717
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.495721285942659e-05,
|
|
"loss": 0.7895,
|
|
"step": 2718
|
|
},
|
|
{
|
|
"epoch": 1.06,
|
|
"learning_rate": 1.4953539760983123e-05,
|
|
"loss": 0.8404,
|
|
"step": 2719
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4949865776700903e-05,
|
|
"loss": 0.8393,
|
|
"step": 2720
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4946190907236944e-05,
|
|
"loss": 0.8439,
|
|
"step": 2721
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4942515153248425e-05,
|
|
"loss": 0.8602,
|
|
"step": 2722
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4938838515392678e-05,
|
|
"loss": 0.8653,
|
|
"step": 2723
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.493516099432719e-05,
|
|
"loss": 0.7904,
|
|
"step": 2724
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4931482590709619e-05,
|
|
"loss": 0.8335,
|
|
"step": 2725
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4927803305197763e-05,
|
|
"loss": 0.8417,
|
|
"step": 2726
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4924123138449591e-05,
|
|
"loss": 0.7808,
|
|
"step": 2727
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4920442091123222e-05,
|
|
"loss": 0.7857,
|
|
"step": 2728
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.491676016387694e-05,
|
|
"loss": 0.7508,
|
|
"step": 2729
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.491307735736918e-05,
|
|
"loss": 0.8759,
|
|
"step": 2730
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4909393672258536e-05,
|
|
"loss": 0.7609,
|
|
"step": 2731
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4905709109203762e-05,
|
|
"loss": 0.8438,
|
|
"step": 2732
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4902023668863758e-05,
|
|
"loss": 0.7821,
|
|
"step": 2733
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4898337351897599e-05,
|
|
"loss": 0.9021,
|
|
"step": 2734
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4894650158964503e-05,
|
|
"loss": 0.8747,
|
|
"step": 2735
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4890962090723849e-05,
|
|
"loss": 0.9007,
|
|
"step": 2736
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.488727314783517e-05,
|
|
"loss": 0.808,
|
|
"step": 2737
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.488358333095816e-05,
|
|
"loss": 0.8842,
|
|
"step": 2738
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4879892640752665e-05,
|
|
"loss": 0.8436,
|
|
"step": 2739
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4876201077878694e-05,
|
|
"loss": 0.8121,
|
|
"step": 2740
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4872508642996399e-05,
|
|
"loss": 0.8823,
|
|
"step": 2741
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4868815336766105e-05,
|
|
"loss": 0.818,
|
|
"step": 2742
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4865121159848277e-05,
|
|
"loss": 0.7685,
|
|
"step": 2743
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4861426112903545e-05,
|
|
"loss": 0.8328,
|
|
"step": 2744
|
|
},
|
|
{
|
|
"epoch": 1.07,
|
|
"learning_rate": 1.4857730196592695e-05,
|
|
"loss": 0.8767,
|
|
"step": 2745
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4854033411576659e-05,
|
|
"loss": 0.764,
|
|
"step": 2746
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4850335758516539e-05,
|
|
"loss": 0.7664,
|
|
"step": 2747
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4846637238073576e-05,
|
|
"loss": 0.9204,
|
|
"step": 2748
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4842937850909183e-05,
|
|
"loss": 0.8279,
|
|
"step": 2749
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4839237597684913e-05,
|
|
"loss": 0.8286,
|
|
"step": 2750
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4835536479062483e-05,
|
|
"loss": 0.8159,
|
|
"step": 2751
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4831834495703756e-05,
|
|
"loss": 0.8209,
|
|
"step": 2752
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4828131648270767e-05,
|
|
"loss": 0.9131,
|
|
"step": 2753
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4824427937425685e-05,
|
|
"loss": 0.8294,
|
|
"step": 2754
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4820723363830847e-05,
|
|
"loss": 0.8237,
|
|
"step": 2755
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4817017928148736e-05,
|
|
"loss": 0.7825,
|
|
"step": 2756
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4813311631041996e-05,
|
|
"loss": 0.8472,
|
|
"step": 2757
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4809604473173422e-05,
|
|
"loss": 0.8235,
|
|
"step": 2758
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4805896455205964e-05,
|
|
"loss": 0.8564,
|
|
"step": 2759
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.480218757780272e-05,
|
|
"loss": 0.8007,
|
|
"step": 2760
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.479847784162695e-05,
|
|
"loss": 0.9545,
|
|
"step": 2761
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4794767247342063e-05,
|
|
"loss": 0.8469,
|
|
"step": 2762
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4791055795611623e-05,
|
|
"loss": 0.8712,
|
|
"step": 2763
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.478734348709935e-05,
|
|
"loss": 0.7931,
|
|
"step": 2764
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4783630322469106e-05,
|
|
"loss": 0.7884,
|
|
"step": 2765
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4779916302384924e-05,
|
|
"loss": 0.8547,
|
|
"step": 2766
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4776201427510973e-05,
|
|
"loss": 0.8926,
|
|
"step": 2767
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4772485698511587e-05,
|
|
"loss": 0.8905,
|
|
"step": 2768
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4768769116051242e-05,
|
|
"loss": 0.8496,
|
|
"step": 2769
|
|
},
|
|
{
|
|
"epoch": 1.08,
|
|
"learning_rate": 1.4765051680794576e-05,
|
|
"loss": 0.8182,
|
|
"step": 2770
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4761333393406374e-05,
|
|
"loss": 0.8559,
|
|
"step": 2771
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4757614254551577e-05,
|
|
"loss": 0.7741,
|
|
"step": 2772
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4753894264895278e-05,
|
|
"loss": 0.828,
|
|
"step": 2773
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4750173425102718e-05,
|
|
"loss": 0.7814,
|
|
"step": 2774
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4746451735839293e-05,
|
|
"loss": 0.7901,
|
|
"step": 2775
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4742729197770551e-05,
|
|
"loss": 0.826,
|
|
"step": 2776
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4739005811562193e-05,
|
|
"loss": 0.8412,
|
|
"step": 2777
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4735281577880067e-05,
|
|
"loss": 0.8725,
|
|
"step": 2778
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4731556497390179e-05,
|
|
"loss": 0.7834,
|
|
"step": 2779
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.472783057075868e-05,
|
|
"loss": 0.8017,
|
|
"step": 2780
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4724103798651874e-05,
|
|
"loss": 0.8422,
|
|
"step": 2781
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4720376181736225e-05,
|
|
"loss": 0.8519,
|
|
"step": 2782
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.471664772067833e-05,
|
|
"loss": 0.8081,
|
|
"step": 2783
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.471291841614496e-05,
|
|
"loss": 0.8448,
|
|
"step": 2784
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4709188268803013e-05,
|
|
"loss": 0.8665,
|
|
"step": 2785
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4705457279319557e-05,
|
|
"loss": 0.7683,
|
|
"step": 2786
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4701725448361797e-05,
|
|
"loss": 0.8243,
|
|
"step": 2787
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4697992776597096e-05,
|
|
"loss": 0.8018,
|
|
"step": 2788
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4694259264692972e-05,
|
|
"loss": 0.8598,
|
|
"step": 2789
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4690524913317079e-05,
|
|
"loss": 0.7836,
|
|
"step": 2790
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4686789723137231e-05,
|
|
"loss": 0.8199,
|
|
"step": 2791
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4683053694821389e-05,
|
|
"loss": 0.8895,
|
|
"step": 2792
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.467931682903767e-05,
|
|
"loss": 0.8351,
|
|
"step": 2793
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.467557912645433e-05,
|
|
"loss": 0.8768,
|
|
"step": 2794
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4671840587739782e-05,
|
|
"loss": 0.8424,
|
|
"step": 2795
|
|
},
|
|
{
|
|
"epoch": 1.09,
|
|
"learning_rate": 1.4668101213562587e-05,
|
|
"loss": 0.8386,
|
|
"step": 2796
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4664361004591459e-05,
|
|
"loss": 0.8597,
|
|
"step": 2797
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4660619961495249e-05,
|
|
"loss": 0.8144,
|
|
"step": 2798
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4656878084942973e-05,
|
|
"loss": 0.783,
|
|
"step": 2799
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4653135375603782e-05,
|
|
"loss": 0.7943,
|
|
"step": 2800
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4649391834146994e-05,
|
|
"loss": 0.8487,
|
|
"step": 2801
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.464564746124205e-05,
|
|
"loss": 0.7919,
|
|
"step": 2802
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4641902257558567e-05,
|
|
"loss": 0.8342,
|
|
"step": 2803
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4638156223766287e-05,
|
|
"loss": 0.8572,
|
|
"step": 2804
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.463440936053512e-05,
|
|
"loss": 0.7768,
|
|
"step": 2805
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4630661668535106e-05,
|
|
"loss": 0.796,
|
|
"step": 2806
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4626913148436455e-05,
|
|
"loss": 0.7559,
|
|
"step": 2807
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.46231638009095e-05,
|
|
"loss": 0.7932,
|
|
"step": 2808
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4619413626624747e-05,
|
|
"loss": 0.7858,
|
|
"step": 2809
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4615662626252826e-05,
|
|
"loss": 0.843,
|
|
"step": 2810
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4611910800464535e-05,
|
|
"loss": 0.7533,
|
|
"step": 2811
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4608158149930805e-05,
|
|
"loss": 0.8174,
|
|
"step": 2812
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4604404675322723e-05,
|
|
"loss": 0.8289,
|
|
"step": 2813
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4600650377311523e-05,
|
|
"loss": 0.8635,
|
|
"step": 2814
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4596895256568582e-05,
|
|
"loss": 0.871,
|
|
"step": 2815
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4593139313765423e-05,
|
|
"loss": 0.8211,
|
|
"step": 2816
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4589382549573723e-05,
|
|
"loss": 0.8519,
|
|
"step": 2817
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4585624964665302e-05,
|
|
"loss": 0.7748,
|
|
"step": 2818
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4581866559712123e-05,
|
|
"loss": 0.8267,
|
|
"step": 2819
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4578107335386307e-05,
|
|
"loss": 0.8788,
|
|
"step": 2820
|
|
},
|
|
{
|
|
"epoch": 1.1,
|
|
"learning_rate": 1.4574347292360102e-05,
|
|
"loss": 0.8164,
|
|
"step": 2821
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4570586431305928e-05,
|
|
"loss": 0.7452,
|
|
"step": 2822
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.456682475289633e-05,
|
|
"loss": 0.8314,
|
|
"step": 2823
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4563062257804009e-05,
|
|
"loss": 0.8158,
|
|
"step": 2824
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4559298946701806e-05,
|
|
"loss": 0.8434,
|
|
"step": 2825
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4555534820262718e-05,
|
|
"loss": 0.7876,
|
|
"step": 2826
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4551769879159876e-05,
|
|
"loss": 0.8227,
|
|
"step": 2827
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4548004124066569e-05,
|
|
"loss": 0.7672,
|
|
"step": 2828
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4544237555656217e-05,
|
|
"loss": 0.8509,
|
|
"step": 2829
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4540470174602398e-05,
|
|
"loss": 0.8613,
|
|
"step": 2830
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.453670198157883e-05,
|
|
"loss": 0.8064,
|
|
"step": 2831
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4532932977259377e-05,
|
|
"loss": 0.8547,
|
|
"step": 2832
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4529163162318049e-05,
|
|
"loss": 0.8287,
|
|
"step": 2833
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4525392537428998e-05,
|
|
"loss": 0.8351,
|
|
"step": 2834
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.452162110326652e-05,
|
|
"loss": 0.792,
|
|
"step": 2835
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4517848860505064e-05,
|
|
"loss": 0.7903,
|
|
"step": 2836
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4514075809819214e-05,
|
|
"loss": 0.8392,
|
|
"step": 2837
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4510301951883707e-05,
|
|
"loss": 0.8773,
|
|
"step": 2838
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4506527287373415e-05,
|
|
"loss": 0.8599,
|
|
"step": 2839
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4502751816963364e-05,
|
|
"loss": 0.8561,
|
|
"step": 2840
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4498975541328715e-05,
|
|
"loss": 0.9251,
|
|
"step": 2841
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4495198461144782e-05,
|
|
"loss": 0.8065,
|
|
"step": 2842
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4491420577087008e-05,
|
|
"loss": 0.7601,
|
|
"step": 2843
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4487641889831003e-05,
|
|
"loss": 0.8665,
|
|
"step": 2844
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4483862400052497e-05,
|
|
"loss": 0.8442,
|
|
"step": 2845
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4480082108427383e-05,
|
|
"loss": 0.9466,
|
|
"step": 2846
|
|
},
|
|
{
|
|
"epoch": 1.11,
|
|
"learning_rate": 1.4476301015631681e-05,
|
|
"loss": 0.8114,
|
|
"step": 2847
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4472519122341566e-05,
|
|
"loss": 0.8127,
|
|
"step": 2848
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4468736429233352e-05,
|
|
"loss": 0.8118,
|
|
"step": 2849
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4464952936983492e-05,
|
|
"loss": 0.8295,
|
|
"step": 2850
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4461168646268589e-05,
|
|
"loss": 0.856,
|
|
"step": 2851
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4457383557765385e-05,
|
|
"loss": 0.8267,
|
|
"step": 2852
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4453597672150763e-05,
|
|
"loss": 0.8564,
|
|
"step": 2853
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4449810990101756e-05,
|
|
"loss": 0.8325,
|
|
"step": 2854
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.444602351229553e-05,
|
|
"loss": 0.8151,
|
|
"step": 2855
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4442235239409396e-05,
|
|
"loss": 0.8189,
|
|
"step": 2856
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.443844617212082e-05,
|
|
"loss": 0.8031,
|
|
"step": 2857
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.443465631110738e-05,
|
|
"loss": 0.8737,
|
|
"step": 2858
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4430865657046834e-05,
|
|
"loss": 0.8906,
|
|
"step": 2859
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4427074210617045e-05,
|
|
"loss": 0.8265,
|
|
"step": 2860
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4423281972496051e-05,
|
|
"loss": 0.819,
|
|
"step": 2861
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4419488943362003e-05,
|
|
"loss": 0.8216,
|
|
"step": 2862
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4415695123893217e-05,
|
|
"loss": 0.7654,
|
|
"step": 2863
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.441190051476813e-05,
|
|
"loss": 0.7826,
|
|
"step": 2864
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4408105116665336e-05,
|
|
"loss": 0.8288,
|
|
"step": 2865
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4404308930263562e-05,
|
|
"loss": 0.8295,
|
|
"step": 2866
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4400511956241678e-05,
|
|
"loss": 0.8382,
|
|
"step": 2867
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4396714195278696e-05,
|
|
"loss": 0.7956,
|
|
"step": 2868
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4392915648053766e-05,
|
|
"loss": 0.8251,
|
|
"step": 2869
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.438911631524618e-05,
|
|
"loss": 0.8005,
|
|
"step": 2870
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4385316197535373e-05,
|
|
"loss": 0.824,
|
|
"step": 2871
|
|
},
|
|
{
|
|
"epoch": 1.12,
|
|
"learning_rate": 1.4381515295600914e-05,
|
|
"loss": 0.7654,
|
|
"step": 2872
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.437771361012252e-05,
|
|
"loss": 0.8073,
|
|
"step": 2873
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4373911141780039e-05,
|
|
"loss": 0.7996,
|
|
"step": 2874
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4370107891253468e-05,
|
|
"loss": 0.8261,
|
|
"step": 2875
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4366303859222944e-05,
|
|
"loss": 0.7833,
|
|
"step": 2876
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4362499046368732e-05,
|
|
"loss": 0.7794,
|
|
"step": 2877
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4358693453371249e-05,
|
|
"loss": 0.8554,
|
|
"step": 2878
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4354887080911048e-05,
|
|
"loss": 0.8907,
|
|
"step": 2879
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4351079929668816e-05,
|
|
"loss": 0.841,
|
|
"step": 2880
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.434727200032539e-05,
|
|
"loss": 0.8449,
|
|
"step": 2881
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4343463293561734e-05,
|
|
"loss": 0.8394,
|
|
"step": 2882
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4339653810058962e-05,
|
|
"loss": 0.8387,
|
|
"step": 2883
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4335843550498319e-05,
|
|
"loss": 0.8366,
|
|
"step": 2884
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.433203251556119e-05,
|
|
"loss": 0.8905,
|
|
"step": 2885
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4328220705929104e-05,
|
|
"loss": 0.8279,
|
|
"step": 2886
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4324408122283722e-05,
|
|
"loss": 0.8835,
|
|
"step": 2887
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4320594765306846e-05,
|
|
"loss": 0.8207,
|
|
"step": 2888
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4316780635680419e-05,
|
|
"loss": 0.824,
|
|
"step": 2889
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4312965734086519e-05,
|
|
"loss": 0.9271,
|
|
"step": 2890
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4309150061207358e-05,
|
|
"loss": 0.856,
|
|
"step": 2891
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.43053336177253e-05,
|
|
"loss": 0.8688,
|
|
"step": 2892
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4301516404322828e-05,
|
|
"loss": 0.8718,
|
|
"step": 2893
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.429769842168258e-05,
|
|
"loss": 0.8386,
|
|
"step": 2894
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4293879670487316e-05,
|
|
"loss": 0.9506,
|
|
"step": 2895
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.429006015141995e-05,
|
|
"loss": 0.8665,
|
|
"step": 2896
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4286239865163514e-05,
|
|
"loss": 0.825,
|
|
"step": 2897
|
|
},
|
|
{
|
|
"epoch": 1.13,
|
|
"learning_rate": 1.4282418812401197e-05,
|
|
"loss": 0.8873,
|
|
"step": 2898
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.427859699381631e-05,
|
|
"loss": 0.839,
|
|
"step": 2899
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.427477441009231e-05,
|
|
"loss": 0.8365,
|
|
"step": 2900
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4270951061912783e-05,
|
|
"loss": 0.8491,
|
|
"step": 2901
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4267126949961466e-05,
|
|
"loss": 0.8195,
|
|
"step": 2902
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4263302074922209e-05,
|
|
"loss": 0.7996,
|
|
"step": 2903
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4259476437479023e-05,
|
|
"loss": 0.8516,
|
|
"step": 2904
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4255650038316042e-05,
|
|
"loss": 0.8895,
|
|
"step": 2905
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4251822878117537e-05,
|
|
"loss": 0.7367,
|
|
"step": 2906
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4247994957567916e-05,
|
|
"loss": 0.7684,
|
|
"step": 2907
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4244166277351728e-05,
|
|
"loss": 0.8537,
|
|
"step": 2908
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.424033683815365e-05,
|
|
"loss": 0.857,
|
|
"step": 2909
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.42365066406585e-05,
|
|
"loss": 0.7653,
|
|
"step": 2910
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4232675685551232e-05,
|
|
"loss": 0.8206,
|
|
"step": 2911
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4228843973516931e-05,
|
|
"loss": 0.8724,
|
|
"step": 2912
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.422501150524082e-05,
|
|
"loss": 0.8843,
|
|
"step": 2913
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4221178281408254e-05,
|
|
"loss": 0.7712,
|
|
"step": 2914
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4217344302704736e-05,
|
|
"loss": 0.7978,
|
|
"step": 2915
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4213509569815884e-05,
|
|
"loss": 0.8407,
|
|
"step": 2916
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4209674083427468e-05,
|
|
"loss": 0.791,
|
|
"step": 2917
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4205837844225384e-05,
|
|
"loss": 0.873,
|
|
"step": 2918
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4202000852895661e-05,
|
|
"loss": 0.7902,
|
|
"step": 2919
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.419816311012447e-05,
|
|
"loss": 0.8143,
|
|
"step": 2920
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.419432461659811e-05,
|
|
"loss": 0.7886,
|
|
"step": 2921
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.419048537300302e-05,
|
|
"loss": 0.8347,
|
|
"step": 2922
|
|
},
|
|
{
|
|
"epoch": 1.14,
|
|
"learning_rate": 1.4186645380025767e-05,
|
|
"loss": 0.8864,
|
|
"step": 2923
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4182804638353055e-05,
|
|
"loss": 0.7866,
|
|
"step": 2924
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4178963148671722e-05,
|
|
"loss": 0.8144,
|
|
"step": 2925
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4175120911668741e-05,
|
|
"loss": 0.9139,
|
|
"step": 2926
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4171277928031215e-05,
|
|
"loss": 0.8563,
|
|
"step": 2927
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4167434198446385e-05,
|
|
"loss": 0.8154,
|
|
"step": 2928
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4163589723601618e-05,
|
|
"loss": 0.8136,
|
|
"step": 2929
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4159744504184425e-05,
|
|
"loss": 0.8086,
|
|
"step": 2930
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.415589854088244e-05,
|
|
"loss": 0.8337,
|
|
"step": 2931
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4152051834383436e-05,
|
|
"loss": 0.8888,
|
|
"step": 2932
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.414820438537532e-05,
|
|
"loss": 0.785,
|
|
"step": 2933
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4144356194546127e-05,
|
|
"loss": 0.881,
|
|
"step": 2934
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4140507262584029e-05,
|
|
"loss": 0.8731,
|
|
"step": 2935
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4136657590177323e-05,
|
|
"loss": 0.8253,
|
|
"step": 2936
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4132807178014449e-05,
|
|
"loss": 0.829,
|
|
"step": 2937
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4128956026783969e-05,
|
|
"loss": 0.7633,
|
|
"step": 2938
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4125104137174588e-05,
|
|
"loss": 0.8653,
|
|
"step": 2939
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4121251509875134e-05,
|
|
"loss": 0.8239,
|
|
"step": 2940
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.411739814557457e-05,
|
|
"loss": 0.7239,
|
|
"step": 2941
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4113544044961995e-05,
|
|
"loss": 0.8593,
|
|
"step": 2942
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.410968920872663e-05,
|
|
"loss": 0.8038,
|
|
"step": 2943
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4105833637557835e-05,
|
|
"loss": 0.8466,
|
|
"step": 2944
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4101977332145102e-05,
|
|
"loss": 0.7784,
|
|
"step": 2945
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4098120293178048e-05,
|
|
"loss": 0.8345,
|
|
"step": 2946
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4094262521346429e-05,
|
|
"loss": 0.8922,
|
|
"step": 2947
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4090404017340129e-05,
|
|
"loss": 0.8074,
|
|
"step": 2948
|
|
},
|
|
{
|
|
"epoch": 1.15,
|
|
"learning_rate": 1.4086544781849154e-05,
|
|
"loss": 0.7841,
|
|
"step": 2949
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.408268481556366e-05,
|
|
"loss": 0.8729,
|
|
"step": 2950
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4078824119173913e-05,
|
|
"loss": 0.8188,
|
|
"step": 2951
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4074962693370326e-05,
|
|
"loss": 0.8219,
|
|
"step": 2952
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4071100538843431e-05,
|
|
"loss": 0.8525,
|
|
"step": 2953
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.40672376562839e-05,
|
|
"loss": 0.8927,
|
|
"step": 2954
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4063374046382525e-05,
|
|
"loss": 0.8189,
|
|
"step": 2955
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4059509709830236e-05,
|
|
"loss": 0.8369,
|
|
"step": 2956
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.405564464731809e-05,
|
|
"loss": 0.8238,
|
|
"step": 2957
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4051778859537273e-05,
|
|
"loss": 0.7712,
|
|
"step": 2958
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.40479123471791e-05,
|
|
"loss": 0.8067,
|
|
"step": 2959
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4044045110935024e-05,
|
|
"loss": 0.9312,
|
|
"step": 2960
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4040177151496612e-05,
|
|
"loss": 0.8406,
|
|
"step": 2961
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4036308469555576e-05,
|
|
"loss": 0.8017,
|
|
"step": 2962
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4032439065803745e-05,
|
|
"loss": 0.7999,
|
|
"step": 2963
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4028568940933086e-05,
|
|
"loss": 0.7668,
|
|
"step": 2964
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4024698095635687e-05,
|
|
"loss": 0.8491,
|
|
"step": 2965
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4020826530603775e-05,
|
|
"loss": 0.7974,
|
|
"step": 2966
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4016954246529697e-05,
|
|
"loss": 0.8223,
|
|
"step": 2967
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.401308124410593e-05,
|
|
"loss": 0.8679,
|
|
"step": 2968
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4009207524025083e-05,
|
|
"loss": 0.7044,
|
|
"step": 2969
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4005333086979889e-05,
|
|
"loss": 0.8051,
|
|
"step": 2970
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.4001457933663212e-05,
|
|
"loss": 0.8586,
|
|
"step": 2971
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.399758206476805e-05,
|
|
"loss": 0.7951,
|
|
"step": 2972
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.399370548098751e-05,
|
|
"loss": 0.7667,
|
|
"step": 2973
|
|
},
|
|
{
|
|
"epoch": 1.16,
|
|
"learning_rate": 1.3989828183014846e-05,
|
|
"loss": 0.8924,
|
|
"step": 2974
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3985950171543436e-05,
|
|
"loss": 0.8044,
|
|
"step": 2975
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3982071447266781e-05,
|
|
"loss": 0.8076,
|
|
"step": 2976
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3978192010878506e-05,
|
|
"loss": 0.8007,
|
|
"step": 2977
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3974311863072376e-05,
|
|
"loss": 0.8034,
|
|
"step": 2978
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3970431004542269e-05,
|
|
"loss": 0.8663,
|
|
"step": 2979
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3966549435982202e-05,
|
|
"loss": 0.7781,
|
|
"step": 2980
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.396266715808631e-05,
|
|
"loss": 0.8637,
|
|
"step": 2981
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.395878417154886e-05,
|
|
"loss": 0.8748,
|
|
"step": 2982
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3954900477064244e-05,
|
|
"loss": 0.8484,
|
|
"step": 2983
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.395101607532698e-05,
|
|
"loss": 0.9417,
|
|
"step": 2984
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.394713096703172e-05,
|
|
"loss": 0.886,
|
|
"step": 2985
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3943245152873223e-05,
|
|
"loss": 0.8787,
|
|
"step": 2986
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3939358633546403e-05,
|
|
"loss": 0.8756,
|
|
"step": 2987
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3935471409746265e-05,
|
|
"loss": 0.8232,
|
|
"step": 2988
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3931583482167978e-05,
|
|
"loss": 0.7813,
|
|
"step": 2989
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3927694851506805e-05,
|
|
"loss": 0.8043,
|
|
"step": 2990
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3923805518458156e-05,
|
|
"loss": 0.8652,
|
|
"step": 2991
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.391991548371755e-05,
|
|
"loss": 0.8557,
|
|
"step": 2992
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3916024747980648e-05,
|
|
"loss": 0.8088,
|
|
"step": 2993
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.391213331194322e-05,
|
|
"loss": 0.7806,
|
|
"step": 2994
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3908241176301179e-05,
|
|
"loss": 0.7552,
|
|
"step": 2995
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3904348341750541e-05,
|
|
"loss": 0.7234,
|
|
"step": 2996
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3900454808987474e-05,
|
|
"loss": 0.8355,
|
|
"step": 2997
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3896560578708243e-05,
|
|
"loss": 0.8311,
|
|
"step": 2998
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3892665651609259e-05,
|
|
"loss": 0.8956,
|
|
"step": 2999
|
|
},
|
|
{
|
|
"epoch": 1.17,
|
|
"learning_rate": 1.3888770028387045e-05,
|
|
"loss": 0.823,
|
|
"step": 3000
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3884873709738259e-05,
|
|
"loss": 0.8533,
|
|
"step": 3001
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3880976696359671e-05,
|
|
"loss": 0.7962,
|
|
"step": 3002
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3877078988948186e-05,
|
|
"loss": 0.8661,
|
|
"step": 3003
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3873180588200827e-05,
|
|
"loss": 0.7868,
|
|
"step": 3004
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3869281494814746e-05,
|
|
"loss": 0.863,
|
|
"step": 3005
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3865381709487208e-05,
|
|
"loss": 0.7734,
|
|
"step": 3006
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3861481232915616e-05,
|
|
"loss": 0.8006,
|
|
"step": 3007
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3857580065797493e-05,
|
|
"loss": 0.7934,
|
|
"step": 3008
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3853678208830475e-05,
|
|
"loss": 0.8323,
|
|
"step": 3009
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3849775662712333e-05,
|
|
"loss": 0.8278,
|
|
"step": 3010
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3845872428140954e-05,
|
|
"loss": 0.8106,
|
|
"step": 3011
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3841968505814354e-05,
|
|
"loss": 0.7899,
|
|
"step": 3012
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.383806389643067e-05,
|
|
"loss": 0.882,
|
|
"step": 3013
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3834158600688159e-05,
|
|
"loss": 0.8379,
|
|
"step": 3014
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3830252619285202e-05,
|
|
"loss": 0.8553,
|
|
"step": 3015
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3826345952920309e-05,
|
|
"loss": 0.7711,
|
|
"step": 3016
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.38224386022921e-05,
|
|
"loss": 0.8553,
|
|
"step": 3017
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3818530568099328e-05,
|
|
"loss": 0.7961,
|
|
"step": 3018
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3814621851040866e-05,
|
|
"loss": 0.8372,
|
|
"step": 3019
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3810712451815705e-05,
|
|
"loss": 0.8704,
|
|
"step": 3020
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3806802371122961e-05,
|
|
"loss": 0.7559,
|
|
"step": 3021
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.380289160966187e-05,
|
|
"loss": 0.8238,
|
|
"step": 3022
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3798980168131795e-05,
|
|
"loss": 0.8365,
|
|
"step": 3023
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3795068047232216e-05,
|
|
"loss": 0.7812,
|
|
"step": 3024
|
|
},
|
|
{
|
|
"epoch": 1.18,
|
|
"learning_rate": 1.3791155247662731e-05,
|
|
"loss": 0.8053,
|
|
"step": 3025
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3787241770123072e-05,
|
|
"loss": 0.7858,
|
|
"step": 3026
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3783327615313074e-05,
|
|
"loss": 0.8,
|
|
"step": 3027
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3779412783932713e-05,
|
|
"loss": 0.7469,
|
|
"step": 3028
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3775497276682068e-05,
|
|
"loss": 0.8029,
|
|
"step": 3029
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3771581094261355e-05,
|
|
"loss": 0.8511,
|
|
"step": 3030
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3767664237370894e-05,
|
|
"loss": 0.8956,
|
|
"step": 3031
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.376374670671114e-05,
|
|
"loss": 0.8138,
|
|
"step": 3032
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3759828502982662e-05,
|
|
"loss": 0.8414,
|
|
"step": 3033
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3755909626886148e-05,
|
|
"loss": 0.8198,
|
|
"step": 3034
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3751990079122412e-05,
|
|
"loss": 0.8143,
|
|
"step": 3035
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3748069860392381e-05,
|
|
"loss": 0.8738,
|
|
"step": 3036
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3744148971397109e-05,
|
|
"loss": 0.8177,
|
|
"step": 3037
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3740227412837763e-05,
|
|
"loss": 0.7482,
|
|
"step": 3038
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3736305185415635e-05,
|
|
"loss": 0.821,
|
|
"step": 3039
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3732382289832133e-05,
|
|
"loss": 0.8746,
|
|
"step": 3040
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.372845872678879e-05,
|
|
"loss": 0.8697,
|
|
"step": 3041
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3724534496987248e-05,
|
|
"loss": 0.8493,
|
|
"step": 3042
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3720609601129284e-05,
|
|
"loss": 0.7528,
|
|
"step": 3043
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3716684039916776e-05,
|
|
"loss": 0.8616,
|
|
"step": 3044
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3712757814051738e-05,
|
|
"loss": 0.8583,
|
|
"step": 3045
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3708830924236286e-05,
|
|
"loss": 0.7928,
|
|
"step": 3046
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3704903371172673e-05,
|
|
"loss": 0.892,
|
|
"step": 3047
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.370097515556325e-05,
|
|
"loss": 0.8511,
|
|
"step": 3048
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3697046278110509e-05,
|
|
"loss": 0.8385,
|
|
"step": 3049
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.3693116739517041e-05,
|
|
"loss": 0.8249,
|
|
"step": 3050
|
|
},
|
|
{
|
|
"epoch": 1.19,
|
|
"learning_rate": 1.368918654048557e-05,
|
|
"loss": 0.7414,
|
|
"step": 3051
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3685255681718922e-05,
|
|
"loss": 0.791,
|
|
"step": 3052
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3681324163920058e-05,
|
|
"loss": 0.8735,
|
|
"step": 3053
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3677391987792048e-05,
|
|
"loss": 0.8552,
|
|
"step": 3054
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3673459154038078e-05,
|
|
"loss": 0.9009,
|
|
"step": 3055
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3669525663361456e-05,
|
|
"loss": 0.8271,
|
|
"step": 3056
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3665591516465609e-05,
|
|
"loss": 0.7925,
|
|
"step": 3057
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.366165671405407e-05,
|
|
"loss": 0.8431,
|
|
"step": 3058
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3657721256830507e-05,
|
|
"loss": 0.7403,
|
|
"step": 3059
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.365378514549869e-05,
|
|
"loss": 0.7496,
|
|
"step": 3060
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3649848380762513e-05,
|
|
"loss": 0.8133,
|
|
"step": 3061
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3645910963325988e-05,
|
|
"loss": 0.8138,
|
|
"step": 3062
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3641972893893237e-05,
|
|
"loss": 0.7918,
|
|
"step": 3063
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3638034173168502e-05,
|
|
"loss": 0.8202,
|
|
"step": 3064
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3634094801856146e-05,
|
|
"loss": 0.7836,
|
|
"step": 3065
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3630154780660643e-05,
|
|
"loss": 0.746,
|
|
"step": 3066
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3626214110286587e-05,
|
|
"loss": 0.8518,
|
|
"step": 3067
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.362227279143868e-05,
|
|
"loss": 0.8078,
|
|
"step": 3068
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3618330824821753e-05,
|
|
"loss": 0.8392,
|
|
"step": 3069
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3614388211140743e-05,
|
|
"loss": 0.8287,
|
|
"step": 3070
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3610444951100702e-05,
|
|
"loss": 0.8096,
|
|
"step": 3071
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3606501045406806e-05,
|
|
"loss": 0.8907,
|
|
"step": 3072
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.360255649476434e-05,
|
|
"loss": 0.7931,
|
|
"step": 3073
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3598611299878704e-05,
|
|
"loss": 0.8626,
|
|
"step": 3074
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3594665461455416e-05,
|
|
"loss": 0.7589,
|
|
"step": 3075
|
|
},
|
|
{
|
|
"epoch": 1.2,
|
|
"learning_rate": 1.3590718980200109e-05,
|
|
"loss": 0.8699,
|
|
"step": 3076
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3586771856818528e-05,
|
|
"loss": 0.8833,
|
|
"step": 3077
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3582824092016535e-05,
|
|
"loss": 0.8849,
|
|
"step": 3078
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3578875686500108e-05,
|
|
"loss": 0.7767,
|
|
"step": 3079
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3574926640975341e-05,
|
|
"loss": 0.8505,
|
|
"step": 3080
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3570976956148432e-05,
|
|
"loss": 0.801,
|
|
"step": 3081
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.356702663272571e-05,
|
|
"loss": 0.8639,
|
|
"step": 3082
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3563075671413602e-05,
|
|
"loss": 0.8287,
|
|
"step": 3083
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3559124072918662e-05,
|
|
"loss": 0.8526,
|
|
"step": 3084
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.355517183794754e-05,
|
|
"loss": 0.8639,
|
|
"step": 3085
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3551218967207032e-05,
|
|
"loss": 0.8766,
|
|
"step": 3086
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3547265461404009e-05,
|
|
"loss": 0.8761,
|
|
"step": 3087
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3543311321245485e-05,
|
|
"loss": 0.8186,
|
|
"step": 3088
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3539356547438572e-05,
|
|
"loss": 0.865,
|
|
"step": 3089
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3535401140690502e-05,
|
|
"loss": 0.9045,
|
|
"step": 3090
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3531445101708617e-05,
|
|
"loss": 0.7636,
|
|
"step": 3091
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3527488431200373e-05,
|
|
"loss": 0.8464,
|
|
"step": 3092
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3523531129873343e-05,
|
|
"loss": 0.8263,
|
|
"step": 3093
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3519573198435203e-05,
|
|
"loss": 0.8575,
|
|
"step": 3094
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3515614637593753e-05,
|
|
"loss": 0.8831,
|
|
"step": 3095
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3511655448056897e-05,
|
|
"loss": 0.7942,
|
|
"step": 3096
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3507695630532655e-05,
|
|
"loss": 0.801,
|
|
"step": 3097
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3503735185729162e-05,
|
|
"loss": 0.8404,
|
|
"step": 3098
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3499774114354655e-05,
|
|
"loss": 0.8923,
|
|
"step": 3099
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3495812417117497e-05,
|
|
"loss": 0.8118,
|
|
"step": 3100
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3491850094726156e-05,
|
|
"loss": 0.8043,
|
|
"step": 3101
|
|
},
|
|
{
|
|
"epoch": 1.21,
|
|
"learning_rate": 1.3487887147889208e-05,
|
|
"loss": 0.802,
|
|
"step": 3102
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3483923577315347e-05,
|
|
"loss": 0.8029,
|
|
"step": 3103
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3479959383713375e-05,
|
|
"loss": 0.8827,
|
|
"step": 3104
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3475994567792208e-05,
|
|
"loss": 0.8684,
|
|
"step": 3105
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.347202913026087e-05,
|
|
"loss": 0.8038,
|
|
"step": 3106
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3468063071828498e-05,
|
|
"loss": 0.759,
|
|
"step": 3107
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3464096393204344e-05,
|
|
"loss": 0.8494,
|
|
"step": 3108
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3460129095097762e-05,
|
|
"loss": 0.8603,
|
|
"step": 3109
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3456161178218224e-05,
|
|
"loss": 0.8171,
|
|
"step": 3110
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.345219264327531e-05,
|
|
"loss": 0.8293,
|
|
"step": 3111
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.344822349097871e-05,
|
|
"loss": 0.8291,
|
|
"step": 3112
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3444253722038227e-05,
|
|
"loss": 0.8238,
|
|
"step": 3113
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3440283337163771e-05,
|
|
"loss": 0.8091,
|
|
"step": 3114
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3436312337065365e-05,
|
|
"loss": 0.7894,
|
|
"step": 3115
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3432340722453138e-05,
|
|
"loss": 0.8303,
|
|
"step": 3116
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3428368494037336e-05,
|
|
"loss": 0.7898,
|
|
"step": 3117
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3424395652528308e-05,
|
|
"loss": 0.8495,
|
|
"step": 3118
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3420422198636515e-05,
|
|
"loss": 0.829,
|
|
"step": 3119
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3416448133072525e-05,
|
|
"loss": 0.8262,
|
|
"step": 3120
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3412473456547027e-05,
|
|
"loss": 0.906,
|
|
"step": 3121
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3408498169770797e-05,
|
|
"loss": 0.8734,
|
|
"step": 3122
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3404522273454745e-05,
|
|
"loss": 0.8103,
|
|
"step": 3123
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3400545768309869e-05,
|
|
"loss": 0.8755,
|
|
"step": 3124
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.339656865504729e-05,
|
|
"loss": 0.8325,
|
|
"step": 3125
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3392590934378236e-05,
|
|
"loss": 0.7711,
|
|
"step": 3126
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3388612607014036e-05,
|
|
"loss": 0.8188,
|
|
"step": 3127
|
|
},
|
|
{
|
|
"epoch": 1.22,
|
|
"learning_rate": 1.3384633673666132e-05,
|
|
"loss": 0.8805,
|
|
"step": 3128
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3380654135046078e-05,
|
|
"loss": 0.9437,
|
|
"step": 3129
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.337667399186553e-05,
|
|
"loss": 0.7772,
|
|
"step": 3130
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3372693244836255e-05,
|
|
"loss": 0.8637,
|
|
"step": 3131
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3368711894670126e-05,
|
|
"loss": 0.8363,
|
|
"step": 3132
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3364729942079126e-05,
|
|
"loss": 0.7865,
|
|
"step": 3133
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3360747387775352e-05,
|
|
"loss": 0.8187,
|
|
"step": 3134
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3356764232470991e-05,
|
|
"loss": 0.7868,
|
|
"step": 3135
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3352780476878359e-05,
|
|
"loss": 0.7371,
|
|
"step": 3136
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3348796121709862e-05,
|
|
"loss": 0.8305,
|
|
"step": 3137
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3344811167678024e-05,
|
|
"loss": 0.7538,
|
|
"step": 3138
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3340825615495467e-05,
|
|
"loss": 0.8099,
|
|
"step": 3139
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3336839465874934e-05,
|
|
"loss": 0.8492,
|
|
"step": 3140
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3332852719529252e-05,
|
|
"loss": 0.8077,
|
|
"step": 3141
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3328865377171386e-05,
|
|
"loss": 0.7613,
|
|
"step": 3142
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3324877439514372e-05,
|
|
"loss": 0.8081,
|
|
"step": 3143
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3320888907271385e-05,
|
|
"loss": 0.8647,
|
|
"step": 3144
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3316899781155685e-05,
|
|
"loss": 0.7908,
|
|
"step": 3145
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3312910061880646e-05,
|
|
"loss": 0.76,
|
|
"step": 3146
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.330891975015975e-05,
|
|
"loss": 0.7864,
|
|
"step": 3147
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3304928846706582e-05,
|
|
"loss": 0.8376,
|
|
"step": 3148
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3300937352234829e-05,
|
|
"loss": 0.8017,
|
|
"step": 3149
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.329694526745829e-05,
|
|
"loss": 0.8273,
|
|
"step": 3150
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.3292952593090869e-05,
|
|
"loss": 0.8779,
|
|
"step": 3151
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.328895932984657e-05,
|
|
"loss": 0.7838,
|
|
"step": 3152
|
|
},
|
|
{
|
|
"epoch": 1.23,
|
|
"learning_rate": 1.328496547843951e-05,
|
|
"loss": 0.8571,
|
|
"step": 3153
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3280971039583906e-05,
|
|
"loss": 0.7904,
|
|
"step": 3154
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3276976013994078e-05,
|
|
"loss": 0.8632,
|
|
"step": 3155
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3272980402384459e-05,
|
|
"loss": 0.8743,
|
|
"step": 3156
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3268984205469579e-05,
|
|
"loss": 0.8376,
|
|
"step": 3157
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3264987423964075e-05,
|
|
"loss": 0.8799,
|
|
"step": 3158
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3260990058582692e-05,
|
|
"loss": 0.8479,
|
|
"step": 3159
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3256992110040276e-05,
|
|
"loss": 0.8523,
|
|
"step": 3160
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3252993579051778e-05,
|
|
"loss": 0.8654,
|
|
"step": 3161
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.324899446633225e-05,
|
|
"loss": 0.7896,
|
|
"step": 3162
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3244994772596855e-05,
|
|
"loss": 0.842,
|
|
"step": 3163
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3240994498560854e-05,
|
|
"loss": 0.8392,
|
|
"step": 3164
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3236993644939617e-05,
|
|
"loss": 0.8407,
|
|
"step": 3165
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.323299221244861e-05,
|
|
"loss": 0.8674,
|
|
"step": 3166
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3228990201803407e-05,
|
|
"loss": 0.8362,
|
|
"step": 3167
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.322498761371969e-05,
|
|
"loss": 0.8297,
|
|
"step": 3168
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3220984448913234e-05,
|
|
"loss": 0.8002,
|
|
"step": 3169
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3216980708099924e-05,
|
|
"loss": 0.818,
|
|
"step": 3170
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3212976391995752e-05,
|
|
"loss": 0.8291,
|
|
"step": 3171
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3208971501316801e-05,
|
|
"loss": 0.9174,
|
|
"step": 3172
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3204966036779265e-05,
|
|
"loss": 0.8655,
|
|
"step": 3173
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3200959999099441e-05,
|
|
"loss": 0.7892,
|
|
"step": 3174
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3196953388993727e-05,
|
|
"loss": 0.9775,
|
|
"step": 3175
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3192946207178619e-05,
|
|
"loss": 0.8056,
|
|
"step": 3176
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.3188938454370723e-05,
|
|
"loss": 0.8316,
|
|
"step": 3177
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.318493013128674e-05,
|
|
"loss": 0.8346,
|
|
"step": 3178
|
|
},
|
|
{
|
|
"epoch": 1.24,
|
|
"learning_rate": 1.318092123864348e-05,
|
|
"loss": 0.8137,
|
|
"step": 3179
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3176911777157846e-05,
|
|
"loss": 0.8301,
|
|
"step": 3180
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3172901747546854e-05,
|
|
"loss": 0.7873,
|
|
"step": 3181
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.316889115052761e-05,
|
|
"loss": 0.8206,
|
|
"step": 3182
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.316487998681733e-05,
|
|
"loss": 0.839,
|
|
"step": 3183
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3160868257133323e-05,
|
|
"loss": 0.8257,
|
|
"step": 3184
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3156855962193012e-05,
|
|
"loss": 0.947,
|
|
"step": 3185
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.315284310271391e-05,
|
|
"loss": 0.7956,
|
|
"step": 3186
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3148829679413632e-05,
|
|
"loss": 0.9625,
|
|
"step": 3187
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.31448156930099e-05,
|
|
"loss": 0.8046,
|
|
"step": 3188
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.314080114422053e-05,
|
|
"loss": 0.7857,
|
|
"step": 3189
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3136786033763444e-05,
|
|
"loss": 0.9184,
|
|
"step": 3190
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3132770362356662e-05,
|
|
"loss": 0.848,
|
|
"step": 3191
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3128754130718302e-05,
|
|
"loss": 0.7953,
|
|
"step": 3192
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3124737339566584e-05,
|
|
"loss": 0.853,
|
|
"step": 3193
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3120719989619832e-05,
|
|
"loss": 0.8224,
|
|
"step": 3194
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3116702081596464e-05,
|
|
"loss": 0.9168,
|
|
"step": 3195
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3112683616215003e-05,
|
|
"loss": 0.8397,
|
|
"step": 3196
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3108664594194063e-05,
|
|
"loss": 0.7916,
|
|
"step": 3197
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3104645016252371e-05,
|
|
"loss": 0.8693,
|
|
"step": 3198
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.310062488310874e-05,
|
|
"loss": 0.8123,
|
|
"step": 3199
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3096604195482087e-05,
|
|
"loss": 0.8165,
|
|
"step": 3200
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3092582954091437e-05,
|
|
"loss": 0.7747,
|
|
"step": 3201
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3088561159655902e-05,
|
|
"loss": 0.8486,
|
|
"step": 3202
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3084538812894696e-05,
|
|
"loss": 0.8104,
|
|
"step": 3203
|
|
},
|
|
{
|
|
"epoch": 1.25,
|
|
"learning_rate": 1.3080515914527135e-05,
|
|
"loss": 0.797,
|
|
"step": 3204
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3076492465272632e-05,
|
|
"loss": 0.773,
|
|
"step": 3205
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3072468465850694e-05,
|
|
"loss": 0.843,
|
|
"step": 3206
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3068443916980936e-05,
|
|
"loss": 0.7588,
|
|
"step": 3207
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3064418819383065e-05,
|
|
"loss": 0.8049,
|
|
"step": 3208
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3060393173776882e-05,
|
|
"loss": 0.8326,
|
|
"step": 3209
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3056366980882296e-05,
|
|
"loss": 0.8172,
|
|
"step": 3210
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3052340241419308e-05,
|
|
"loss": 0.8469,
|
|
"step": 3211
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3048312956108015e-05,
|
|
"loss": 0.7597,
|
|
"step": 3212
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3044285125668614e-05,
|
|
"loss": 0.885,
|
|
"step": 3213
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3040256750821408e-05,
|
|
"loss": 0.8163,
|
|
"step": 3214
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3036227832286776e-05,
|
|
"loss": 0.8756,
|
|
"step": 3215
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.303219837078522e-05,
|
|
"loss": 0.8295,
|
|
"step": 3216
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3028168367037317e-05,
|
|
"loss": 0.8558,
|
|
"step": 3217
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3024137821763753e-05,
|
|
"loss": 0.7814,
|
|
"step": 3218
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.302010673568531e-05,
|
|
"loss": 0.8131,
|
|
"step": 3219
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3016075109522865e-05,
|
|
"loss": 0.8944,
|
|
"step": 3220
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3012042943997388e-05,
|
|
"loss": 0.8356,
|
|
"step": 3221
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3008010239829951e-05,
|
|
"loss": 0.8019,
|
|
"step": 3222
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.3003976997741724e-05,
|
|
"loss": 0.7668,
|
|
"step": 3223
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.2999943218453961e-05,
|
|
"loss": 0.7813,
|
|
"step": 3224
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.2995908902688029e-05,
|
|
"loss": 0.827,
|
|
"step": 3225
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.2991874051165376e-05,
|
|
"loss": 0.8008,
|
|
"step": 3226
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.2987838664607556e-05,
|
|
"loss": 0.8175,
|
|
"step": 3227
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.2983802743736212e-05,
|
|
"loss": 0.8796,
|
|
"step": 3228
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.2979766289273093e-05,
|
|
"loss": 0.8032,
|
|
"step": 3229
|
|
},
|
|
{
|
|
"epoch": 1.26,
|
|
"learning_rate": 1.2975729301940023e-05,
|
|
"loss": 0.8533,
|
|
"step": 3230
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2971691782458947e-05,
|
|
"loss": 0.8053,
|
|
"step": 3231
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2967653731551881e-05,
|
|
"loss": 0.8287,
|
|
"step": 3232
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2963615149940961e-05,
|
|
"loss": 0.7987,
|
|
"step": 3233
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.295957603834839e-05,
|
|
"loss": 0.8249,
|
|
"step": 3234
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2955536397496489e-05,
|
|
"loss": 0.8189,
|
|
"step": 3235
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.295149622810766e-05,
|
|
"loss": 0.8088,
|
|
"step": 3236
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2947455530904411e-05,
|
|
"loss": 0.8349,
|
|
"step": 3237
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.294341430660933e-05,
|
|
"loss": 0.8304,
|
|
"step": 3238
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2939372555945112e-05,
|
|
"loss": 0.7939,
|
|
"step": 3239
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2935330279634534e-05,
|
|
"loss": 0.8077,
|
|
"step": 3240
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2931287478400482e-05,
|
|
"loss": 0.7803,
|
|
"step": 3241
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2927244152965928e-05,
|
|
"loss": 0.7882,
|
|
"step": 3242
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2923200304053932e-05,
|
|
"loss": 0.7969,
|
|
"step": 3243
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2919155932387655e-05,
|
|
"loss": 0.8048,
|
|
"step": 3244
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.291511103869035e-05,
|
|
"loss": 0.8079,
|
|
"step": 3245
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2911065623685369e-05,
|
|
"loss": 0.8007,
|
|
"step": 3246
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2907019688096141e-05,
|
|
"loss": 0.8528,
|
|
"step": 3247
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2902973232646203e-05,
|
|
"loss": 0.8338,
|
|
"step": 3248
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2898926258059183e-05,
|
|
"loss": 0.7262,
|
|
"step": 3249
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2894878765058797e-05,
|
|
"loss": 0.8464,
|
|
"step": 3250
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2890830754368855e-05,
|
|
"loss": 0.8602,
|
|
"step": 3251
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.288678222671326e-05,
|
|
"loss": 0.7246,
|
|
"step": 3252
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.288273318281601e-05,
|
|
"loss": 0.7951,
|
|
"step": 3253
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2878683623401192e-05,
|
|
"loss": 0.8388,
|
|
"step": 3254
|
|
},
|
|
{
|
|
"epoch": 1.27,
|
|
"learning_rate": 1.2874633549192987e-05,
|
|
"loss": 0.9271,
|
|
"step": 3255
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2870582960915669e-05,
|
|
"loss": 0.8744,
|
|
"step": 3256
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2866531859293601e-05,
|
|
"loss": 0.7856,
|
|
"step": 3257
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2862480245051239e-05,
|
|
"loss": 0.7722,
|
|
"step": 3258
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2858428118913131e-05,
|
|
"loss": 0.7767,
|
|
"step": 3259
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2854375481603916e-05,
|
|
"loss": 0.831,
|
|
"step": 3260
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2850322333848326e-05,
|
|
"loss": 0.8604,
|
|
"step": 3261
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2846268676371184e-05,
|
|
"loss": 0.8219,
|
|
"step": 3262
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2842214509897399e-05,
|
|
"loss": 0.8371,
|
|
"step": 3263
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.283815983515198e-05,
|
|
"loss": 0.7876,
|
|
"step": 3264
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.283410465286002e-05,
|
|
"loss": 0.8448,
|
|
"step": 3265
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2830048963746703e-05,
|
|
"loss": 0.8423,
|
|
"step": 3266
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.282599276853731e-05,
|
|
"loss": 0.7869,
|
|
"step": 3267
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2821936067957208e-05,
|
|
"loss": 0.811,
|
|
"step": 3268
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.281787886273185e-05,
|
|
"loss": 0.8158,
|
|
"step": 3269
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2813821153586789e-05,
|
|
"loss": 0.7747,
|
|
"step": 3270
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2809762941247658e-05,
|
|
"loss": 0.7714,
|
|
"step": 3271
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.280570422644019e-05,
|
|
"loss": 0.8408,
|
|
"step": 3272
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2801645009890195e-05,
|
|
"loss": 0.9267,
|
|
"step": 3273
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.279758529232359e-05,
|
|
"loss": 0.8344,
|
|
"step": 3274
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2793525074466367e-05,
|
|
"loss": 0.8299,
|
|
"step": 3275
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2789464357044613e-05,
|
|
"loss": 0.8883,
|
|
"step": 3276
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.27854031407845e-05,
|
|
"loss": 0.8115,
|
|
"step": 3277
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2781341426412298e-05,
|
|
"loss": 0.7158,
|
|
"step": 3278
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.277727921465436e-05,
|
|
"loss": 0.8502,
|
|
"step": 3279
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2773216506237129e-05,
|
|
"loss": 0.8513,
|
|
"step": 3280
|
|
},
|
|
{
|
|
"epoch": 1.28,
|
|
"learning_rate": 1.2769153301887138e-05,
|
|
"loss": 0.8477,
|
|
"step": 3281
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2765089602331006e-05,
|
|
"loss": 0.7555,
|
|
"step": 3282
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2761025408295443e-05,
|
|
"loss": 0.721,
|
|
"step": 3283
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2756960720507244e-05,
|
|
"loss": 0.8706,
|
|
"step": 3284
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2752895539693298e-05,
|
|
"loss": 0.8227,
|
|
"step": 3285
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2748829866580574e-05,
|
|
"loss": 0.7942,
|
|
"step": 3286
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2744763701896146e-05,
|
|
"loss": 0.8772,
|
|
"step": 3287
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2740697046367148e-05,
|
|
"loss": 0.8235,
|
|
"step": 3288
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2736629900720832e-05,
|
|
"loss": 0.8991,
|
|
"step": 3289
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2732562265684512e-05,
|
|
"loss": 0.8303,
|
|
"step": 3290
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2728494141985608e-05,
|
|
"loss": 0.8647,
|
|
"step": 3291
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2724425530351616e-05,
|
|
"loss": 0.8317,
|
|
"step": 3292
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2720356431510128e-05,
|
|
"loss": 0.8114,
|
|
"step": 3293
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2716286846188817e-05,
|
|
"loss": 0.7986,
|
|
"step": 3294
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2712216775115444e-05,
|
|
"loss": 0.7844,
|
|
"step": 3295
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2708146219017857e-05,
|
|
"loss": 0.7918,
|
|
"step": 3296
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2704075178623991e-05,
|
|
"loss": 0.8342,
|
|
"step": 3297
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.270000365466187e-05,
|
|
"loss": 0.8106,
|
|
"step": 3298
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.26959316478596e-05,
|
|
"loss": 0.8108,
|
|
"step": 3299
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.269185915894538e-05,
|
|
"loss": 0.8428,
|
|
"step": 3300
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2687786188647484e-05,
|
|
"loss": 0.8764,
|
|
"step": 3301
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2683712737694285e-05,
|
|
"loss": 0.7486,
|
|
"step": 3302
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2679638806814234e-05,
|
|
"loss": 0.7714,
|
|
"step": 3303
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.267556439673587e-05,
|
|
"loss": 0.8122,
|
|
"step": 3304
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.2671489508187814e-05,
|
|
"loss": 0.7389,
|
|
"step": 3305
|
|
},
|
|
{
|
|
"epoch": 1.29,
|
|
"learning_rate": 1.266741414189878e-05,
|
|
"loss": 0.8342,
|
|
"step": 3306
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2663338298597562e-05,
|
|
"loss": 0.7589,
|
|
"step": 3307
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2659261979013043e-05,
|
|
"loss": 0.7938,
|
|
"step": 3308
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2655185183874188e-05,
|
|
"loss": 0.8033,
|
|
"step": 3309
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2651107913910043e-05,
|
|
"loss": 0.7569,
|
|
"step": 3310
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2647030169849752e-05,
|
|
"loss": 0.7634,
|
|
"step": 3311
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2642951952422533e-05,
|
|
"loss": 0.8317,
|
|
"step": 3312
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2638873262357688e-05,
|
|
"loss": 0.8088,
|
|
"step": 3313
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2634794100384609e-05,
|
|
"loss": 0.8122,
|
|
"step": 3314
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2630714467232772e-05,
|
|
"loss": 0.8419,
|
|
"step": 3315
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.262663436363173e-05,
|
|
"loss": 0.7813,
|
|
"step": 3316
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2622553790311133e-05,
|
|
"loss": 0.8249,
|
|
"step": 3317
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2618472748000706e-05,
|
|
"loss": 0.8032,
|
|
"step": 3318
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2614391237430255e-05,
|
|
"loss": 0.8452,
|
|
"step": 3319
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2610309259329675e-05,
|
|
"loss": 0.8158,
|
|
"step": 3320
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2606226814428947e-05,
|
|
"loss": 0.8536,
|
|
"step": 3321
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2602143903458135e-05,
|
|
"loss": 0.744,
|
|
"step": 3322
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2598060527147376e-05,
|
|
"loss": 0.7812,
|
|
"step": 3323
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2593976686226906e-05,
|
|
"loss": 0.8332,
|
|
"step": 3324
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2589892381427026e-05,
|
|
"loss": 0.7906,
|
|
"step": 3325
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.258580761347814e-05,
|
|
"loss": 0.8136,
|
|
"step": 3326
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2581722383110719e-05,
|
|
"loss": 0.8507,
|
|
"step": 3327
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2577636691055326e-05,
|
|
"loss": 0.7858,
|
|
"step": 3328
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2573550538042598e-05,
|
|
"loss": 0.9472,
|
|
"step": 3329
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2569463924803265e-05,
|
|
"loss": 0.7649,
|
|
"step": 3330
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2565376852068127e-05,
|
|
"loss": 0.8261,
|
|
"step": 3331
|
|
},
|
|
{
|
|
"epoch": 1.3,
|
|
"learning_rate": 1.2561289320568082e-05,
|
|
"loss": 0.8276,
|
|
"step": 3332
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2557201331034093e-05,
|
|
"loss": 0.8003,
|
|
"step": 3333
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2553112884197219e-05,
|
|
"loss": 0.8815,
|
|
"step": 3334
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2549023980788588e-05,
|
|
"loss": 0.808,
|
|
"step": 3335
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2544934621539425e-05,
|
|
"loss": 0.8268,
|
|
"step": 3336
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2540844807181018e-05,
|
|
"loss": 0.8584,
|
|
"step": 3337
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2536754538444752e-05,
|
|
"loss": 0.7564,
|
|
"step": 3338
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2532663816062089e-05,
|
|
"loss": 0.8959,
|
|
"step": 3339
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2528572640764567e-05,
|
|
"loss": 0.8085,
|
|
"step": 3340
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2524481013283809e-05,
|
|
"loss": 0.8231,
|
|
"step": 3341
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2520388934351519e-05,
|
|
"loss": 0.807,
|
|
"step": 3342
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2516296404699484e-05,
|
|
"loss": 0.7768,
|
|
"step": 3343
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2512203425059564e-05,
|
|
"loss": 0.8379,
|
|
"step": 3344
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2508109996163707e-05,
|
|
"loss": 0.8367,
|
|
"step": 3345
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2504016118743936e-05,
|
|
"loss": 0.8597,
|
|
"step": 3346
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2499921793532362e-05,
|
|
"loss": 0.8339,
|
|
"step": 3347
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2495827021261166e-05,
|
|
"loss": 0.8001,
|
|
"step": 3348
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2491731802662616e-05,
|
|
"loss": 0.847,
|
|
"step": 3349
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2487636138469057e-05,
|
|
"loss": 0.816,
|
|
"step": 3350
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2483540029412917e-05,
|
|
"loss": 0.8043,
|
|
"step": 3351
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2479443476226696e-05,
|
|
"loss": 0.7703,
|
|
"step": 3352
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2475346479642982e-05,
|
|
"loss": 0.8276,
|
|
"step": 3353
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2471249040394438e-05,
|
|
"loss": 0.8577,
|
|
"step": 3354
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2467151159213804e-05,
|
|
"loss": 0.8345,
|
|
"step": 3355
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2463052836833906e-05,
|
|
"loss": 0.8257,
|
|
"step": 3356
|
|
},
|
|
{
|
|
"epoch": 1.31,
|
|
"learning_rate": 1.2458954073987643e-05,
|
|
"loss": 0.7489,
|
|
"step": 3357
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2454854871407993e-05,
|
|
"loss": 0.8099,
|
|
"step": 3358
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2450755229828016e-05,
|
|
"loss": 0.9014,
|
|
"step": 3359
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2446655149980849e-05,
|
|
"loss": 0.8476,
|
|
"step": 3360
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2442554632599703e-05,
|
|
"loss": 0.823,
|
|
"step": 3361
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2438453678417874e-05,
|
|
"loss": 0.831,
|
|
"step": 3362
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2434352288168738e-05,
|
|
"loss": 0.8331,
|
|
"step": 3363
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2430250462585736e-05,
|
|
"loss": 0.7804,
|
|
"step": 3364
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2426148202402405e-05,
|
|
"loss": 0.866,
|
|
"step": 3365
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2422045508352336e-05,
|
|
"loss": 0.8671,
|
|
"step": 3366
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2417942381169224e-05,
|
|
"loss": 0.8256,
|
|
"step": 3367
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.241383882158682e-05,
|
|
"loss": 0.742,
|
|
"step": 3368
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2409734830338974e-05,
|
|
"loss": 0.7864,
|
|
"step": 3369
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2405630408159584e-05,
|
|
"loss": 0.776,
|
|
"step": 3370
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2401525555782654e-05,
|
|
"loss": 0.8702,
|
|
"step": 3371
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2397420273942246e-05,
|
|
"loss": 0.8713,
|
|
"step": 3372
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.239331456337251e-05,
|
|
"loss": 0.7897,
|
|
"step": 3373
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2389208424807662e-05,
|
|
"loss": 0.7442,
|
|
"step": 3374
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2385101858982004e-05,
|
|
"loss": 0.8214,
|
|
"step": 3375
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2380994866629914e-05,
|
|
"loss": 0.8655,
|
|
"step": 3376
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2376887448485837e-05,
|
|
"loss": 0.7784,
|
|
"step": 3377
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2372779605284305e-05,
|
|
"loss": 0.8065,
|
|
"step": 3378
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2368671337759917e-05,
|
|
"loss": 0.8675,
|
|
"step": 3379
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2364562646647357e-05,
|
|
"loss": 0.8225,
|
|
"step": 3380
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2360453532681376e-05,
|
|
"loss": 0.8769,
|
|
"step": 3381
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2356343996596809e-05,
|
|
"loss": 0.8692,
|
|
"step": 3382
|
|
},
|
|
{
|
|
"epoch": 1.32,
|
|
"learning_rate": 1.2352234039128555e-05,
|
|
"loss": 0.8642,
|
|
"step": 3383
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2348123661011602e-05,
|
|
"loss": 0.8132,
|
|
"step": 3384
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2344012862981002e-05,
|
|
"loss": 0.7934,
|
|
"step": 3385
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2339901645771893e-05,
|
|
"loss": 0.7945,
|
|
"step": 3386
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2335790010119473e-05,
|
|
"loss": 0.8328,
|
|
"step": 3387
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.233167795675903e-05,
|
|
"loss": 0.947,
|
|
"step": 3388
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2327565486425919e-05,
|
|
"loss": 0.7569,
|
|
"step": 3389
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.232345259985557e-05,
|
|
"loss": 0.8339,
|
|
"step": 3390
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2319339297783488e-05,
|
|
"loss": 0.843,
|
|
"step": 3391
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2315225580945252e-05,
|
|
"loss": 0.8191,
|
|
"step": 3392
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2311111450076518e-05,
|
|
"loss": 0.803,
|
|
"step": 3393
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2306996905913011e-05,
|
|
"loss": 0.8511,
|
|
"step": 3394
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2302881949190535e-05,
|
|
"loss": 0.8359,
|
|
"step": 3395
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2298766580644963e-05,
|
|
"loss": 0.761,
|
|
"step": 3396
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2294650801012247e-05,
|
|
"loss": 0.8499,
|
|
"step": 3397
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2290534611028406e-05,
|
|
"loss": 0.8249,
|
|
"step": 3398
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.228641801142954e-05,
|
|
"loss": 0.791,
|
|
"step": 3399
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2282301002951815e-05,
|
|
"loss": 0.659,
|
|
"step": 3400
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2278183586331472e-05,
|
|
"loss": 0.8046,
|
|
"step": 3401
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2274065762304832e-05,
|
|
"loss": 0.8564,
|
|
"step": 3402
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2269947531608277e-05,
|
|
"loss": 0.8466,
|
|
"step": 3403
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2265828894978273e-05,
|
|
"loss": 0.8644,
|
|
"step": 3404
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2261709853151352e-05,
|
|
"loss": 0.8284,
|
|
"step": 3405
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2257590406864118e-05,
|
|
"loss": 0.7805,
|
|
"step": 3406
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.2253470556853251e-05,
|
|
"loss": 0.7558,
|
|
"step": 3407
|
|
},
|
|
{
|
|
"epoch": 1.33,
|
|
"learning_rate": 1.22493503038555e-05,
|
|
"loss": 0.8153,
|
|
"step": 3408
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.224522964860769e-05,
|
|
"loss": 0.7902,
|
|
"step": 3409
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2241108591846714e-05,
|
|
"loss": 0.7699,
|
|
"step": 3410
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2236987134309536e-05,
|
|
"loss": 0.7886,
|
|
"step": 3411
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.22328652767332e-05,
|
|
"loss": 0.7841,
|
|
"step": 3412
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.222874301985481e-05,
|
|
"loss": 0.8279,
|
|
"step": 3413
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2224620364411546e-05,
|
|
"loss": 0.7758,
|
|
"step": 3414
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2220497311140666e-05,
|
|
"loss": 0.7489,
|
|
"step": 3415
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2216373860779489e-05,
|
|
"loss": 0.8327,
|
|
"step": 3416
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2212250014065412e-05,
|
|
"loss": 0.8356,
|
|
"step": 3417
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2208125771735897e-05,
|
|
"loss": 0.8498,
|
|
"step": 3418
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2204001134528485e-05,
|
|
"loss": 0.8051,
|
|
"step": 3419
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2199876103180774e-05,
|
|
"loss": 0.7688,
|
|
"step": 3420
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2195750678430452e-05,
|
|
"loss": 0.8089,
|
|
"step": 3421
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2191624861015255e-05,
|
|
"loss": 0.8243,
|
|
"step": 3422
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.218749865167301e-05,
|
|
"loss": 0.824,
|
|
"step": 3423
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2183372051141597e-05,
|
|
"loss": 0.8057,
|
|
"step": 3424
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2179245060158984e-05,
|
|
"loss": 0.8641,
|
|
"step": 3425
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2175117679463187e-05,
|
|
"loss": 0.7793,
|
|
"step": 3426
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2170989909792313e-05,
|
|
"loss": 0.8463,
|
|
"step": 3427
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2166861751884519e-05,
|
|
"loss": 0.8258,
|
|
"step": 3428
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2162733206478048e-05,
|
|
"loss": 0.8701,
|
|
"step": 3429
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2158604274311205e-05,
|
|
"loss": 0.8173,
|
|
"step": 3430
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2154474956122362e-05,
|
|
"loss": 0.8422,
|
|
"step": 3431
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2150345252649966e-05,
|
|
"loss": 0.8449,
|
|
"step": 3432
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2146215164632527e-05,
|
|
"loss": 0.8247,
|
|
"step": 3433
|
|
},
|
|
{
|
|
"epoch": 1.34,
|
|
"learning_rate": 1.2142084692808627e-05,
|
|
"loss": 0.828,
|
|
"step": 3434
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2137953837916914e-05,
|
|
"loss": 0.7913,
|
|
"step": 3435
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.213382260069611e-05,
|
|
"loss": 0.8106,
|
|
"step": 3436
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2129690981885e-05,
|
|
"loss": 0.8168,
|
|
"step": 3437
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2125558982222441e-05,
|
|
"loss": 0.8035,
|
|
"step": 3438
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.212142660244735e-05,
|
|
"loss": 0.8243,
|
|
"step": 3439
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.211729384329873e-05,
|
|
"loss": 0.7868,
|
|
"step": 3440
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2113160705515626e-05,
|
|
"loss": 0.8532,
|
|
"step": 3441
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2109027189837174e-05,
|
|
"loss": 0.8264,
|
|
"step": 3442
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2104893297002566e-05,
|
|
"loss": 0.8665,
|
|
"step": 3443
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2100759027751063e-05,
|
|
"loss": 0.7994,
|
|
"step": 3444
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2096624382821995e-05,
|
|
"loss": 0.8401,
|
|
"step": 3445
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.209248936295476e-05,
|
|
"loss": 0.7513,
|
|
"step": 3446
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2088353968888815e-05,
|
|
"loss": 0.8138,
|
|
"step": 3447
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2084218201363697e-05,
|
|
"loss": 0.8487,
|
|
"step": 3448
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2080082061119001e-05,
|
|
"loss": 0.8398,
|
|
"step": 3449
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2075945548894388e-05,
|
|
"loss": 0.8063,
|
|
"step": 3450
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2071808665429592e-05,
|
|
"loss": 0.9326,
|
|
"step": 3451
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2067671411464408e-05,
|
|
"loss": 0.7502,
|
|
"step": 3452
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2063533787738697e-05,
|
|
"loss": 0.8493,
|
|
"step": 3453
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.205939579499239e-05,
|
|
"loss": 0.7971,
|
|
"step": 3454
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2055257433965479e-05,
|
|
"loss": 0.8667,
|
|
"step": 3455
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2051118705398034e-05,
|
|
"loss": 0.7248,
|
|
"step": 3456
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2046979610030167e-05,
|
|
"loss": 0.806,
|
|
"step": 3457
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2042840148602082e-05,
|
|
"loss": 0.8599,
|
|
"step": 3458
|
|
},
|
|
{
|
|
"epoch": 1.35,
|
|
"learning_rate": 1.2038700321854031e-05,
|
|
"loss": 0.7482,
|
|
"step": 3459
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2034560130526341e-05,
|
|
"loss": 0.7889,
|
|
"step": 3460
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2030419575359394e-05,
|
|
"loss": 0.7803,
|
|
"step": 3461
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2026278657093651e-05,
|
|
"loss": 0.7593,
|
|
"step": 3462
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.202213737646962e-05,
|
|
"loss": 0.8096,
|
|
"step": 3463
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2017995734227892e-05,
|
|
"loss": 0.8333,
|
|
"step": 3464
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2013853731109113e-05,
|
|
"loss": 0.8221,
|
|
"step": 3465
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2009711367853996e-05,
|
|
"loss": 0.825,
|
|
"step": 3466
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2005568645203314e-05,
|
|
"loss": 0.8657,
|
|
"step": 3467
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.2001425563897909e-05,
|
|
"loss": 0.8155,
|
|
"step": 3468
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1997282124678686e-05,
|
|
"loss": 0.7775,
|
|
"step": 3469
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1993138328286617e-05,
|
|
"loss": 0.805,
|
|
"step": 3470
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1988994175462731e-05,
|
|
"loss": 0.7467,
|
|
"step": 3471
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1984849666948123e-05,
|
|
"loss": 0.8694,
|
|
"step": 3472
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1980704803483963e-05,
|
|
"loss": 0.8248,
|
|
"step": 3473
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.197655958581146e-05,
|
|
"loss": 0.7722,
|
|
"step": 3474
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1972414014671914e-05,
|
|
"loss": 0.8867,
|
|
"step": 3475
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1968268090806665e-05,
|
|
"loss": 0.734,
|
|
"step": 3476
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1964121814957136e-05,
|
|
"loss": 0.8663,
|
|
"step": 3477
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1959975187864793e-05,
|
|
"loss": 0.8444,
|
|
"step": 3478
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1955828210271187e-05,
|
|
"loss": 0.7924,
|
|
"step": 3479
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1951680882917906e-05,
|
|
"loss": 0.8569,
|
|
"step": 3480
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1947533206546627e-05,
|
|
"loss": 0.7952,
|
|
"step": 3481
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1943385181899066e-05,
|
|
"loss": 0.8754,
|
|
"step": 3482
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1939236809717017e-05,
|
|
"loss": 0.831,
|
|
"step": 3483
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1935088090742332e-05,
|
|
"loss": 0.8997,
|
|
"step": 3484
|
|
},
|
|
{
|
|
"epoch": 1.36,
|
|
"learning_rate": 1.1930939025716923e-05,
|
|
"loss": 0.8184,
|
|
"step": 3485
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1926789615382766e-05,
|
|
"loss": 0.8606,
|
|
"step": 3486
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1922639860481894e-05,
|
|
"loss": 0.8263,
|
|
"step": 3487
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.191848976175641e-05,
|
|
"loss": 0.8218,
|
|
"step": 3488
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1914339319948472e-05,
|
|
"loss": 0.7513,
|
|
"step": 3489
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.19101885358003e-05,
|
|
"loss": 0.7841,
|
|
"step": 3490
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1906037410054178e-05,
|
|
"loss": 0.7977,
|
|
"step": 3491
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1901885943452448e-05,
|
|
"loss": 0.8492,
|
|
"step": 3492
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1897734136737517e-05,
|
|
"loss": 0.8334,
|
|
"step": 3493
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1893581990651848e-05,
|
|
"loss": 0.8407,
|
|
"step": 3494
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1889429505937966e-05,
|
|
"loss": 0.7352,
|
|
"step": 3495
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.188527668333846e-05,
|
|
"loss": 0.8477,
|
|
"step": 3496
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1881123523595976e-05,
|
|
"loss": 0.8268,
|
|
"step": 3497
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1876970027453223e-05,
|
|
"loss": 0.855,
|
|
"step": 3498
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1872816195652965e-05,
|
|
"loss": 0.8702,
|
|
"step": 3499
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1868662028938032e-05,
|
|
"loss": 0.8521,
|
|
"step": 3500
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.186450752805131e-05,
|
|
"loss": 0.8109,
|
|
"step": 3501
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1860352693735747e-05,
|
|
"loss": 0.7808,
|
|
"step": 3502
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1856197526734347e-05,
|
|
"loss": 0.788,
|
|
"step": 3503
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1852042027790181e-05,
|
|
"loss": 0.8452,
|
|
"step": 3504
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1847886197646372e-05,
|
|
"loss": 0.9082,
|
|
"step": 3505
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1843730037046104e-05,
|
|
"loss": 0.8485,
|
|
"step": 3506
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.183957354673262e-05,
|
|
"loss": 0.8029,
|
|
"step": 3507
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.183541672744923e-05,
|
|
"loss": 0.7685,
|
|
"step": 3508
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1831259579939282e-05,
|
|
"loss": 0.867,
|
|
"step": 3509
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1827102104946211e-05,
|
|
"loss": 0.7754,
|
|
"step": 3510
|
|
},
|
|
{
|
|
"epoch": 1.37,
|
|
"learning_rate": 1.1822944303213486e-05,
|
|
"loss": 0.8507,
|
|
"step": 3511
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1818786175484652e-05,
|
|
"loss": 0.828,
|
|
"step": 3512
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1814627722503296e-05,
|
|
"loss": 0.7792,
|
|
"step": 3513
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1810468945013081e-05,
|
|
"loss": 0.811,
|
|
"step": 3514
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1806309843757708e-05,
|
|
"loss": 0.8613,
|
|
"step": 3515
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.180215041948096e-05,
|
|
"loss": 0.9092,
|
|
"step": 3516
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1797990672926652e-05,
|
|
"loss": 0.8072,
|
|
"step": 3517
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1793830604838678e-05,
|
|
"loss": 0.7974,
|
|
"step": 3518
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1789670215960975e-05,
|
|
"loss": 0.8373,
|
|
"step": 3519
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.178550950703755e-05,
|
|
"loss": 0.7611,
|
|
"step": 3520
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.178134847881245e-05,
|
|
"loss": 0.8677,
|
|
"step": 3521
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1777187132029794e-05,
|
|
"loss": 0.8131,
|
|
"step": 3522
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1773025467433759e-05,
|
|
"loss": 0.8362,
|
|
"step": 3523
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1768863485768564e-05,
|
|
"loss": 0.7758,
|
|
"step": 3524
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1764701187778502e-05,
|
|
"loss": 0.7626,
|
|
"step": 3525
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.176053857420791e-05,
|
|
"loss": 0.7844,
|
|
"step": 3526
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1756375645801185e-05,
|
|
"loss": 0.7554,
|
|
"step": 3527
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1752212403302785e-05,
|
|
"loss": 0.7766,
|
|
"step": 3528
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1748048847457216e-05,
|
|
"loss": 0.7805,
|
|
"step": 3529
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1743884979009048e-05,
|
|
"loss": 0.7945,
|
|
"step": 3530
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.17397207987029e-05,
|
|
"loss": 0.8294,
|
|
"step": 3531
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1735556307283453e-05,
|
|
"loss": 0.8759,
|
|
"step": 3532
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1731391505495442e-05,
|
|
"loss": 0.8622,
|
|
"step": 3533
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1727226394083649e-05,
|
|
"loss": 0.7996,
|
|
"step": 3534
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1723060973792924e-05,
|
|
"loss": 0.8086,
|
|
"step": 3535
|
|
},
|
|
{
|
|
"epoch": 1.38,
|
|
"learning_rate": 1.1718895245368167e-05,
|
|
"loss": 0.8523,
|
|
"step": 3536
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1714729209554332e-05,
|
|
"loss": 0.8088,
|
|
"step": 3537
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1710562867096429e-05,
|
|
"loss": 0.868,
|
|
"step": 3538
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1706396218739522e-05,
|
|
"loss": 0.8063,
|
|
"step": 3539
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.170222926522873e-05,
|
|
"loss": 0.8093,
|
|
"step": 3540
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.169806200730923e-05,
|
|
"loss": 0.7443,
|
|
"step": 3541
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1693894445726245e-05,
|
|
"loss": 0.7938,
|
|
"step": 3542
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1689726581225064e-05,
|
|
"loss": 0.7948,
|
|
"step": 3543
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.168555841455102e-05,
|
|
"loss": 0.8392,
|
|
"step": 3544
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1681389946449504e-05,
|
|
"loss": 0.9113,
|
|
"step": 3545
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1677221177665963e-05,
|
|
"loss": 0.8012,
|
|
"step": 3546
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1673052108945893e-05,
|
|
"loss": 0.8975,
|
|
"step": 3547
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1668882741034846e-05,
|
|
"loss": 0.7858,
|
|
"step": 3548
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.166471307467843e-05,
|
|
"loss": 0.7809,
|
|
"step": 3549
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1660543110622302e-05,
|
|
"loss": 0.8665,
|
|
"step": 3550
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.165637284961218e-05,
|
|
"loss": 0.7537,
|
|
"step": 3551
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1652202292393822e-05,
|
|
"loss": 0.7654,
|
|
"step": 3552
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.164803143971305e-05,
|
|
"loss": 0.8024,
|
|
"step": 3553
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1643860292315732e-05,
|
|
"loss": 0.7487,
|
|
"step": 3554
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1639688850947798e-05,
|
|
"loss": 0.8185,
|
|
"step": 3555
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1635517116355219e-05,
|
|
"loss": 0.8414,
|
|
"step": 3556
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.163134508928403e-05,
|
|
"loss": 0.8316,
|
|
"step": 3557
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1627172770480304e-05,
|
|
"loss": 0.8258,
|
|
"step": 3558
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.162300016069018e-05,
|
|
"loss": 0.8755,
|
|
"step": 3559
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1618827260659844e-05,
|
|
"loss": 0.8722,
|
|
"step": 3560
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.1614654071135531e-05,
|
|
"loss": 0.8574,
|
|
"step": 3561
|
|
},
|
|
{
|
|
"epoch": 1.39,
|
|
"learning_rate": 1.161048059286353e-05,
|
|
"loss": 0.7901,
|
|
"step": 3562
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1606306826590186e-05,
|
|
"loss": 0.7841,
|
|
"step": 3563
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1602132773061885e-05,
|
|
"loss": 0.7707,
|
|
"step": 3564
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1597958433025072e-05,
|
|
"loss": 0.7716,
|
|
"step": 3565
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1593783807226249e-05,
|
|
"loss": 0.7846,
|
|
"step": 3566
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.158960889641195e-05,
|
|
"loss": 0.8406,
|
|
"step": 3567
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1585433701328784e-05,
|
|
"loss": 0.8468,
|
|
"step": 3568
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1581258222723389e-05,
|
|
"loss": 0.8511,
|
|
"step": 3569
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1577082461342472e-05,
|
|
"loss": 0.7766,
|
|
"step": 3570
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1572906417932775e-05,
|
|
"loss": 0.7747,
|
|
"step": 3571
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1568730093241103e-05,
|
|
"loss": 0.9052,
|
|
"step": 3572
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1564553488014298e-05,
|
|
"loss": 0.8215,
|
|
"step": 3573
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1560376602999272e-05,
|
|
"loss": 0.7676,
|
|
"step": 3574
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1556199438942963e-05,
|
|
"loss": 0.8692,
|
|
"step": 3575
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.155202199659238e-05,
|
|
"loss": 0.7592,
|
|
"step": 3576
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1547844276694566e-05,
|
|
"loss": 0.8029,
|
|
"step": 3577
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1543666279996625e-05,
|
|
"loss": 0.7675,
|
|
"step": 3578
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1539488007245704e-05,
|
|
"loss": 0.8128,
|
|
"step": 3579
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1535309459189e-05,
|
|
"loss": 0.786,
|
|
"step": 3580
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1531130636573762e-05,
|
|
"loss": 0.747,
|
|
"step": 3581
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1526951540147287e-05,
|
|
"loss": 0.7986,
|
|
"step": 3582
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.152277217065692e-05,
|
|
"loss": 0.8269,
|
|
"step": 3583
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1518592528850052e-05,
|
|
"loss": 0.8524,
|
|
"step": 3584
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1514412615474132e-05,
|
|
"loss": 0.8436,
|
|
"step": 3585
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.1510232431276647e-05,
|
|
"loss": 0.8162,
|
|
"step": 3586
|
|
},
|
|
{
|
|
"epoch": 1.4,
|
|
"learning_rate": 1.150605197700514e-05,
|
|
"loss": 0.8673,
|
|
"step": 3587
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1501871253407196e-05,
|
|
"loss": 0.829,
|
|
"step": 3588
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1497690261230452e-05,
|
|
"loss": 0.8207,
|
|
"step": 3589
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1493509001222596e-05,
|
|
"loss": 0.7467,
|
|
"step": 3590
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1489327474131357e-05,
|
|
"loss": 0.8566,
|
|
"step": 3591
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1485145680704519e-05,
|
|
"loss": 0.8631,
|
|
"step": 3592
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1480963621689904e-05,
|
|
"loss": 0.836,
|
|
"step": 3593
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1476781297835393e-05,
|
|
"loss": 0.8102,
|
|
"step": 3594
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1472598709888905e-05,
|
|
"loss": 0.7384,
|
|
"step": 3595
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1468415858598413e-05,
|
|
"loss": 0.8071,
|
|
"step": 3596
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.146423274471193e-05,
|
|
"loss": 0.8258,
|
|
"step": 3597
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1460049368977519e-05,
|
|
"loss": 0.8595,
|
|
"step": 3598
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1455865732143296e-05,
|
|
"loss": 0.8337,
|
|
"step": 3599
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1451681834957417e-05,
|
|
"loss": 0.8285,
|
|
"step": 3600
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1447497678168085e-05,
|
|
"loss": 0.694,
|
|
"step": 3601
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1443313262523547e-05,
|
|
"loss": 0.8076,
|
|
"step": 3602
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.143912858877211e-05,
|
|
"loss": 0.7888,
|
|
"step": 3603
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1434943657662103e-05,
|
|
"loss": 0.7729,
|
|
"step": 3604
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1430758469941927e-05,
|
|
"loss": 0.8432,
|
|
"step": 3605
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1426573026360009e-05,
|
|
"loss": 0.815,
|
|
"step": 3606
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1422387327664835e-05,
|
|
"loss": 0.7924,
|
|
"step": 3607
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1418201374604923e-05,
|
|
"loss": 0.8195,
|
|
"step": 3608
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1414015167928856e-05,
|
|
"loss": 0.8883,
|
|
"step": 3609
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.140982870838524e-05,
|
|
"loss": 0.7706,
|
|
"step": 3610
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1405641996722748e-05,
|
|
"loss": 0.8548,
|
|
"step": 3611
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1401455033690076e-05,
|
|
"loss": 0.8115,
|
|
"step": 3612
|
|
},
|
|
{
|
|
"epoch": 1.41,
|
|
"learning_rate": 1.1397267820035986e-05,
|
|
"loss": 0.7368,
|
|
"step": 3613
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1393080356509267e-05,
|
|
"loss": 0.7966,
|
|
"step": 3614
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1388892643858766e-05,
|
|
"loss": 0.8524,
|
|
"step": 3615
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.138470468283337e-05,
|
|
"loss": 0.8692,
|
|
"step": 3616
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1380516474182006e-05,
|
|
"loss": 0.8688,
|
|
"step": 3617
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1376328018653652e-05,
|
|
"loss": 0.8245,
|
|
"step": 3618
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1372139316997326e-05,
|
|
"loss": 0.8104,
|
|
"step": 3619
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.136795036996209e-05,
|
|
"loss": 0.7416,
|
|
"step": 3620
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1363761178297055e-05,
|
|
"loss": 0.767,
|
|
"step": 3621
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1359571742751365e-05,
|
|
"loss": 0.7972,
|
|
"step": 3622
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1355382064074221e-05,
|
|
"loss": 0.7942,
|
|
"step": 3623
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1351192143014858e-05,
|
|
"loss": 0.8375,
|
|
"step": 3624
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1347001980322556e-05,
|
|
"loss": 0.7071,
|
|
"step": 3625
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1342811576746644e-05,
|
|
"loss": 0.8745,
|
|
"step": 3626
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1338620933036483e-05,
|
|
"loss": 0.7258,
|
|
"step": 3627
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1334430049941491e-05,
|
|
"loss": 0.7995,
|
|
"step": 3628
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1330238928211112e-05,
|
|
"loss": 0.8618,
|
|
"step": 3629
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1326047568594852e-05,
|
|
"loss": 0.8236,
|
|
"step": 3630
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1321855971842243e-05,
|
|
"loss": 0.7445,
|
|
"step": 3631
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1317664138702868e-05,
|
|
"loss": 0.7587,
|
|
"step": 3632
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.131347206992635e-05,
|
|
"loss": 0.7954,
|
|
"step": 3633
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1309279766262355e-05,
|
|
"loss": 0.8867,
|
|
"step": 3634
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1305087228460589e-05,
|
|
"loss": 0.8207,
|
|
"step": 3635
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1300894457270802e-05,
|
|
"loss": 0.8664,
|
|
"step": 3636
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.1296701453442784e-05,
|
|
"loss": 0.7509,
|
|
"step": 3637
|
|
},
|
|
{
|
|
"epoch": 1.42,
|
|
"learning_rate": 1.129250821772637e-05,
|
|
"loss": 0.8004,
|
|
"step": 3638
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.128831475087143e-05,
|
|
"loss": 0.8091,
|
|
"step": 3639
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1284121053627882e-05,
|
|
"loss": 0.8469,
|
|
"step": 3640
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1279927126745683e-05,
|
|
"loss": 0.8223,
|
|
"step": 3641
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1275732970974828e-05,
|
|
"loss": 0.8149,
|
|
"step": 3642
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1271538587065355e-05,
|
|
"loss": 0.757,
|
|
"step": 3643
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1267343975767353e-05,
|
|
"loss": 0.8504,
|
|
"step": 3644
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1263149137830929e-05,
|
|
"loss": 0.7826,
|
|
"step": 3645
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1258954074006252e-05,
|
|
"loss": 0.7208,
|
|
"step": 3646
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1254758785043516e-05,
|
|
"loss": 0.7831,
|
|
"step": 3647
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1250563271692971e-05,
|
|
"loss": 0.8438,
|
|
"step": 3648
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1246367534704889e-05,
|
|
"loss": 0.8115,
|
|
"step": 3649
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1242171574829599e-05,
|
|
"loss": 0.8229,
|
|
"step": 3650
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.123797539281746e-05,
|
|
"loss": 0.8503,
|
|
"step": 3651
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.123377898941887e-05,
|
|
"loss": 0.8725,
|
|
"step": 3652
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1229582365384271e-05,
|
|
"loss": 0.7756,
|
|
"step": 3653
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1225385521464143e-05,
|
|
"loss": 0.8002,
|
|
"step": 3654
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1221188458409007e-05,
|
|
"loss": 0.8505,
|
|
"step": 3655
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1216991176969418e-05,
|
|
"loss": 0.8378,
|
|
"step": 3656
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1212793677895976e-05,
|
|
"loss": 0.8075,
|
|
"step": 3657
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1208595961939317e-05,
|
|
"loss": 0.7965,
|
|
"step": 3658
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1204398029850119e-05,
|
|
"loss": 0.7797,
|
|
"step": 3659
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.120019988237909e-05,
|
|
"loss": 0.8104,
|
|
"step": 3660
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1196001520276986e-05,
|
|
"loss": 0.8337,
|
|
"step": 3661
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1191802944294597e-05,
|
|
"loss": 0.8637,
|
|
"step": 3662
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.1187604155182753e-05,
|
|
"loss": 0.892,
|
|
"step": 3663
|
|
},
|
|
{
|
|
"epoch": 1.43,
|
|
"learning_rate": 1.118340515369232e-05,
|
|
"loss": 0.8446,
|
|
"step": 3664
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1179205940574204e-05,
|
|
"loss": 0.7697,
|
|
"step": 3665
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1175006516579345e-05,
|
|
"loss": 0.7541,
|
|
"step": 3666
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.117080688245873e-05,
|
|
"loss": 0.854,
|
|
"step": 3667
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1166607038963366e-05,
|
|
"loss": 0.8271,
|
|
"step": 3668
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1162406986844322e-05,
|
|
"loss": 0.8204,
|
|
"step": 3669
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1158206726852683e-05,
|
|
"loss": 0.8597,
|
|
"step": 3670
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.115400625973958e-05,
|
|
"loss": 0.8297,
|
|
"step": 3671
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1149805586256182e-05,
|
|
"loss": 0.8236,
|
|
"step": 3672
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1145604707153692e-05,
|
|
"loss": 0.8255,
|
|
"step": 3673
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.114140362318335e-05,
|
|
"loss": 0.7972,
|
|
"step": 3674
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1137202335096437e-05,
|
|
"loss": 0.8512,
|
|
"step": 3675
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1133000843644263e-05,
|
|
"loss": 0.8083,
|
|
"step": 3676
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1128799149578182e-05,
|
|
"loss": 0.8206,
|
|
"step": 3677
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1124597253649576e-05,
|
|
"loss": 0.8256,
|
|
"step": 3678
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.112039515660987e-05,
|
|
"loss": 0.8316,
|
|
"step": 3679
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1116192859210527e-05,
|
|
"loss": 0.7883,
|
|
"step": 3680
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1111990362203034e-05,
|
|
"loss": 0.7744,
|
|
"step": 3681
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1107787666338926e-05,
|
|
"loss": 0.78,
|
|
"step": 3682
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1103584772369764e-05,
|
|
"loss": 0.8161,
|
|
"step": 3683
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1099381681047157e-05,
|
|
"loss": 0.7485,
|
|
"step": 3684
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1095178393122734e-05,
|
|
"loss": 0.8178,
|
|
"step": 3685
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1090974909348172e-05,
|
|
"loss": 0.7855,
|
|
"step": 3686
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1086771230475176e-05,
|
|
"loss": 0.7971,
|
|
"step": 3687
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1082567357255484e-05,
|
|
"loss": 0.7504,
|
|
"step": 3688
|
|
},
|
|
{
|
|
"epoch": 1.44,
|
|
"learning_rate": 1.1078363290440877e-05,
|
|
"loss": 0.8809,
|
|
"step": 3689
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1074159030783162e-05,
|
|
"loss": 0.7655,
|
|
"step": 3690
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.106995457903419e-05,
|
|
"loss": 0.7971,
|
|
"step": 3691
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1065749935945835e-05,
|
|
"loss": 0.7729,
|
|
"step": 3692
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1061545102270015e-05,
|
|
"loss": 0.8376,
|
|
"step": 3693
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1057340078758676e-05,
|
|
"loss": 0.7894,
|
|
"step": 3694
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1053134866163799e-05,
|
|
"loss": 0.8096,
|
|
"step": 3695
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1048929465237401e-05,
|
|
"loss": 0.8027,
|
|
"step": 3696
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.104472387673153e-05,
|
|
"loss": 0.7927,
|
|
"step": 3697
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1040518101398277e-05,
|
|
"loss": 0.7919,
|
|
"step": 3698
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1036312139989746e-05,
|
|
"loss": 0.8168,
|
|
"step": 3699
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1032105993258098e-05,
|
|
"loss": 0.8595,
|
|
"step": 3700
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1027899661955505e-05,
|
|
"loss": 0.7597,
|
|
"step": 3701
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1023693146834197e-05,
|
|
"loss": 0.8236,
|
|
"step": 3702
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1019486448646407e-05,
|
|
"loss": 0.7328,
|
|
"step": 3703
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1015279568144427e-05,
|
|
"loss": 0.8006,
|
|
"step": 3704
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1011072506080565e-05,
|
|
"loss": 0.7949,
|
|
"step": 3705
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1006865263207176e-05,
|
|
"loss": 0.8014,
|
|
"step": 3706
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.1002657840276627e-05,
|
|
"loss": 0.7444,
|
|
"step": 3707
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.099845023804134e-05,
|
|
"loss": 0.8612,
|
|
"step": 3708
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.0994242457253752e-05,
|
|
"loss": 0.808,
|
|
"step": 3709
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.0990034498666342e-05,
|
|
"loss": 0.7958,
|
|
"step": 3710
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.0985826363031613e-05,
|
|
"loss": 0.8106,
|
|
"step": 3711
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.0981618051102105e-05,
|
|
"loss": 0.8294,
|
|
"step": 3712
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.0977409563630391e-05,
|
|
"loss": 0.7706,
|
|
"step": 3713
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.0973200901369066e-05,
|
|
"loss": 0.8164,
|
|
"step": 3714
|
|
},
|
|
{
|
|
"epoch": 1.45,
|
|
"learning_rate": 1.096899206507077e-05,
|
|
"loss": 0.8102,
|
|
"step": 3715
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.096478305548816e-05,
|
|
"loss": 0.79,
|
|
"step": 3716
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0960573873373936e-05,
|
|
"loss": 0.819,
|
|
"step": 3717
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.095636451948082e-05,
|
|
"loss": 0.8402,
|
|
"step": 3718
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0952154994561574e-05,
|
|
"loss": 0.7997,
|
|
"step": 3719
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0947945299368978e-05,
|
|
"loss": 0.786,
|
|
"step": 3720
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0943735434655855e-05,
|
|
"loss": 0.9033,
|
|
"step": 3721
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0939525401175049e-05,
|
|
"loss": 0.8118,
|
|
"step": 3722
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.093531519967944e-05,
|
|
"loss": 0.8467,
|
|
"step": 3723
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0931104830921935e-05,
|
|
"loss": 0.87,
|
|
"step": 3724
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0926894295655472e-05,
|
|
"loss": 0.8779,
|
|
"step": 3725
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.092268359463302e-05,
|
|
"loss": 0.7582,
|
|
"step": 3726
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0918472728607576e-05,
|
|
"loss": 0.8463,
|
|
"step": 3727
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0914261698332168e-05,
|
|
"loss": 0.7881,
|
|
"step": 3728
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0910050504559846e-05,
|
|
"loss": 0.8627,
|
|
"step": 3729
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0905839148043702e-05,
|
|
"loss": 0.803,
|
|
"step": 3730
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0901627629536846e-05,
|
|
"loss": 0.8452,
|
|
"step": 3731
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0897415949792427e-05,
|
|
"loss": 0.8322,
|
|
"step": 3732
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0893204109563614e-05,
|
|
"loss": 0.8309,
|
|
"step": 3733
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0888992109603607e-05,
|
|
"loss": 0.8407,
|
|
"step": 3734
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0884779950665636e-05,
|
|
"loss": 0.9029,
|
|
"step": 3735
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0880567633502962e-05,
|
|
"loss": 0.825,
|
|
"step": 3736
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0876355158868866e-05,
|
|
"loss": 0.8287,
|
|
"step": 3737
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0872142527516665e-05,
|
|
"loss": 0.8037,
|
|
"step": 3738
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0867929740199708e-05,
|
|
"loss": 0.8577,
|
|
"step": 3739
|
|
},
|
|
{
|
|
"epoch": 1.46,
|
|
"learning_rate": 1.0863716797671353e-05,
|
|
"loss": 0.8447,
|
|
"step": 3740
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.085950370068501e-05,
|
|
"loss": 0.8787,
|
|
"step": 3741
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0855290449994095e-05,
|
|
"loss": 0.762,
|
|
"step": 3742
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.085107704635207e-05,
|
|
"loss": 0.7906,
|
|
"step": 3743
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0846863490512409e-05,
|
|
"loss": 0.7441,
|
|
"step": 3744
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0842649783228624e-05,
|
|
"loss": 0.8334,
|
|
"step": 3745
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0838435925254246e-05,
|
|
"loss": 0.8671,
|
|
"step": 3746
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0834221917342843e-05,
|
|
"loss": 0.8591,
|
|
"step": 3747
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0830007760247995e-05,
|
|
"loss": 0.8248,
|
|
"step": 3748
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0825793454723325e-05,
|
|
"loss": 0.9317,
|
|
"step": 3749
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0821579001522471e-05,
|
|
"loss": 0.8488,
|
|
"step": 3750
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0817364401399099e-05,
|
|
"loss": 0.7328,
|
|
"step": 3751
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0813149655106913e-05,
|
|
"loss": 0.8421,
|
|
"step": 3752
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0808934763399623e-05,
|
|
"loss": 0.8649,
|
|
"step": 3753
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0804719727030986e-05,
|
|
"loss": 0.8368,
|
|
"step": 3754
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0800504546754762e-05,
|
|
"loss": 0.8274,
|
|
"step": 3755
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0796289223324763e-05,
|
|
"loss": 0.8634,
|
|
"step": 3756
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0792073757494801e-05,
|
|
"loss": 0.8564,
|
|
"step": 3757
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0787858150018734e-05,
|
|
"loss": 0.7885,
|
|
"step": 3758
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0783642401650431e-05,
|
|
"loss": 0.8082,
|
|
"step": 3759
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0779426513143798e-05,
|
|
"loss": 0.8436,
|
|
"step": 3760
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0775210485252752e-05,
|
|
"loss": 0.7646,
|
|
"step": 3761
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.077099431873125e-05,
|
|
"loss": 0.7703,
|
|
"step": 3762
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0766778014333261e-05,
|
|
"loss": 0.8546,
|
|
"step": 3763
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0762561572812789e-05,
|
|
"loss": 0.8016,
|
|
"step": 3764
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0758344994923857e-05,
|
|
"loss": 0.8029,
|
|
"step": 3765
|
|
},
|
|
{
|
|
"epoch": 1.47,
|
|
"learning_rate": 1.0754128281420511e-05,
|
|
"loss": 0.8309,
|
|
"step": 3766
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0749911433056826e-05,
|
|
"loss": 0.812,
|
|
"step": 3767
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.07456944505869e-05,
|
|
"loss": 0.8117,
|
|
"step": 3768
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0741477334764847e-05,
|
|
"loss": 0.846,
|
|
"step": 3769
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0737260086344818e-05,
|
|
"loss": 0.7968,
|
|
"step": 3770
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0733042706080978e-05,
|
|
"loss": 0.7636,
|
|
"step": 3771
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.072882519472752e-05,
|
|
"loss": 0.819,
|
|
"step": 3772
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.072460755303866e-05,
|
|
"loss": 0.8669,
|
|
"step": 3773
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0720389781768635e-05,
|
|
"loss": 0.7994,
|
|
"step": 3774
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0716171881671704e-05,
|
|
"loss": 0.7925,
|
|
"step": 3775
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0711953853502154e-05,
|
|
"loss": 0.7847,
|
|
"step": 3776
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0707735698014298e-05,
|
|
"loss": 0.7606,
|
|
"step": 3777
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0703517415962461e-05,
|
|
"loss": 0.7709,
|
|
"step": 3778
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0699299008100996e-05,
|
|
"loss": 0.7722,
|
|
"step": 3779
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0695080475184281e-05,
|
|
"loss": 0.8415,
|
|
"step": 3780
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.069086181796671e-05,
|
|
"loss": 0.8422,
|
|
"step": 3781
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0686643037202706e-05,
|
|
"loss": 0.7647,
|
|
"step": 3782
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0682424133646712e-05,
|
|
"loss": 0.7443,
|
|
"step": 3783
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0678205108053189e-05,
|
|
"loss": 0.8042,
|
|
"step": 3784
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0673985961176624e-05,
|
|
"loss": 0.8327,
|
|
"step": 3785
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0669766693771529e-05,
|
|
"loss": 0.8666,
|
|
"step": 3786
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0665547306592425e-05,
|
|
"loss": 0.7978,
|
|
"step": 3787
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0661327800393871e-05,
|
|
"loss": 0.8153,
|
|
"step": 3788
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0657108175930432e-05,
|
|
"loss": 0.756,
|
|
"step": 3789
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0652888433956707e-05,
|
|
"loss": 0.7789,
|
|
"step": 3790
|
|
},
|
|
{
|
|
"epoch": 1.48,
|
|
"learning_rate": 1.0648668575227307e-05,
|
|
"loss": 0.8069,
|
|
"step": 3791
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0644448600496866e-05,
|
|
"loss": 0.8585,
|
|
"step": 3792
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0640228510520046e-05,
|
|
"loss": 0.7667,
|
|
"step": 3793
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0636008306051511e-05,
|
|
"loss": 0.887,
|
|
"step": 3794
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0631787987845971e-05,
|
|
"loss": 0.8112,
|
|
"step": 3795
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0627567556658135e-05,
|
|
"loss": 0.8236,
|
|
"step": 3796
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0623347013242746e-05,
|
|
"loss": 0.8687,
|
|
"step": 3797
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0619126358354552e-05,
|
|
"loss": 0.7506,
|
|
"step": 3798
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0614905592748343e-05,
|
|
"loss": 0.8162,
|
|
"step": 3799
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0610684717178905e-05,
|
|
"loss": 0.8957,
|
|
"step": 3800
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0606463732401063e-05,
|
|
"loss": 0.8377,
|
|
"step": 3801
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0602242639169649e-05,
|
|
"loss": 0.8495,
|
|
"step": 3802
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0598021438239519e-05,
|
|
"loss": 0.8248,
|
|
"step": 3803
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0593800130365548e-05,
|
|
"loss": 0.8167,
|
|
"step": 3804
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0589578716302634e-05,
|
|
"loss": 0.7554,
|
|
"step": 3805
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0585357196805684e-05,
|
|
"loss": 0.8297,
|
|
"step": 3806
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0581135572629633e-05,
|
|
"loss": 0.9863,
|
|
"step": 3807
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0576913844529432e-05,
|
|
"loss": 0.8087,
|
|
"step": 3808
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0572692013260051e-05,
|
|
"loss": 0.7847,
|
|
"step": 3809
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0568470079576475e-05,
|
|
"loss": 0.7316,
|
|
"step": 3810
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0564248044233711e-05,
|
|
"loss": 0.7807,
|
|
"step": 3811
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0560025907986789e-05,
|
|
"loss": 0.824,
|
|
"step": 3812
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0555803671590738e-05,
|
|
"loss": 0.8929,
|
|
"step": 3813
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0551581335800634e-05,
|
|
"loss": 0.8188,
|
|
"step": 3814
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0547358901371542e-05,
|
|
"loss": 0.8752,
|
|
"step": 3815
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0543136369058568e-05,
|
|
"loss": 0.8644,
|
|
"step": 3816
|
|
},
|
|
{
|
|
"epoch": 1.49,
|
|
"learning_rate": 1.0538913739616817e-05,
|
|
"loss": 0.8047,
|
|
"step": 3817
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.053469101380142e-05,
|
|
"loss": 0.7765,
|
|
"step": 3818
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0530468192367531e-05,
|
|
"loss": 0.8515,
|
|
"step": 3819
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.052624527607031e-05,
|
|
"loss": 0.8758,
|
|
"step": 3820
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.052202226566494e-05,
|
|
"loss": 0.8311,
|
|
"step": 3821
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.051779916190662e-05,
|
|
"loss": 0.7933,
|
|
"step": 3822
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0513575965550565e-05,
|
|
"loss": 0.8415,
|
|
"step": 3823
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0509352677352002e-05,
|
|
"loss": 0.8342,
|
|
"step": 3824
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0505129298066188e-05,
|
|
"loss": 0.8415,
|
|
"step": 3825
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0500905828448382e-05,
|
|
"loss": 0.7954,
|
|
"step": 3826
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0496682269253869e-05,
|
|
"loss": 0.8429,
|
|
"step": 3827
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0492458621237938e-05,
|
|
"loss": 0.8459,
|
|
"step": 3828
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.048823488515591e-05,
|
|
"loss": 0.8329,
|
|
"step": 3829
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0484011061763108e-05,
|
|
"loss": 0.8627,
|
|
"step": 3830
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0479787151814876e-05,
|
|
"loss": 0.8095,
|
|
"step": 3831
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.047556315606658e-05,
|
|
"loss": 0.8608,
|
|
"step": 3832
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0471339075273583e-05,
|
|
"loss": 0.8471,
|
|
"step": 3833
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.046711491019129e-05,
|
|
"loss": 0.8192,
|
|
"step": 3834
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.046289066157509e-05,
|
|
"loss": 0.7942,
|
|
"step": 3835
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0458666330180418e-05,
|
|
"loss": 0.8126,
|
|
"step": 3836
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0454441916762698e-05,
|
|
"loss": 0.7656,
|
|
"step": 3837
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0450217422077387e-05,
|
|
"loss": 0.7168,
|
|
"step": 3838
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0445992846879942e-05,
|
|
"loss": 0.8401,
|
|
"step": 3839
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0441768191925848e-05,
|
|
"loss": 0.8258,
|
|
"step": 3840
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0437543457970591e-05,
|
|
"loss": 0.8139,
|
|
"step": 3841
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0433318645769684e-05,
|
|
"loss": 0.7783,
|
|
"step": 3842
|
|
},
|
|
{
|
|
"epoch": 1.5,
|
|
"learning_rate": 1.0429093756078645e-05,
|
|
"loss": 0.8323,
|
|
"step": 3843
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0424868789653006e-05,
|
|
"loss": 0.7717,
|
|
"step": 3844
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0420643747248323e-05,
|
|
"loss": 0.8013,
|
|
"step": 3845
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0416418629620149e-05,
|
|
"loss": 0.8005,
|
|
"step": 3846
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0412193437524067e-05,
|
|
"loss": 0.814,
|
|
"step": 3847
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0407968171715659e-05,
|
|
"loss": 0.8165,
|
|
"step": 3848
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0403742832950537e-05,
|
|
"loss": 0.86,
|
|
"step": 3849
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0399517421984302e-05,
|
|
"loss": 0.7968,
|
|
"step": 3850
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0395291939572593e-05,
|
|
"loss": 0.7514,
|
|
"step": 3851
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0391066386471045e-05,
|
|
"loss": 0.8597,
|
|
"step": 3852
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0386840763435315e-05,
|
|
"loss": 0.8036,
|
|
"step": 3853
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0382615071221065e-05,
|
|
"loss": 0.8593,
|
|
"step": 3854
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0378389310583978e-05,
|
|
"loss": 0.8173,
|
|
"step": 3855
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0374163482279737e-05,
|
|
"loss": 0.8373,
|
|
"step": 3856
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0369937587064057e-05,
|
|
"loss": 0.8449,
|
|
"step": 3857
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0365711625692636e-05,
|
|
"loss": 0.7864,
|
|
"step": 3858
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0361485598921213e-05,
|
|
"loss": 0.7978,
|
|
"step": 3859
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.035725950750552e-05,
|
|
"loss": 0.79,
|
|
"step": 3860
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0353033352201311e-05,
|
|
"loss": 0.7911,
|
|
"step": 3861
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0348807133764345e-05,
|
|
"loss": 0.8249,
|
|
"step": 3862
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0344580852950392e-05,
|
|
"loss": 0.8697,
|
|
"step": 3863
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0340354510515235e-05,
|
|
"loss": 0.7695,
|
|
"step": 3864
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0336128107214675e-05,
|
|
"loss": 0.8083,
|
|
"step": 3865
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0331901643804511e-05,
|
|
"loss": 0.8419,
|
|
"step": 3866
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0327675121040562e-05,
|
|
"loss": 0.8139,
|
|
"step": 3867
|
|
},
|
|
{
|
|
"epoch": 1.51,
|
|
"learning_rate": 1.0323448539678653e-05,
|
|
"loss": 0.8864,
|
|
"step": 3868
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0319221900474621e-05,
|
|
"loss": 0.8461,
|
|
"step": 3869
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0314995204184317e-05,
|
|
"loss": 0.7806,
|
|
"step": 3870
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0310768451563593e-05,
|
|
"loss": 0.7788,
|
|
"step": 3871
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0306541643368321e-05,
|
|
"loss": 0.7695,
|
|
"step": 3872
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0302314780354381e-05,
|
|
"loss": 0.8841,
|
|
"step": 3873
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0298087863277655e-05,
|
|
"loss": 0.8498,
|
|
"step": 3874
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0293860892894044e-05,
|
|
"loss": 0.8323,
|
|
"step": 3875
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0289633869959452e-05,
|
|
"loss": 0.8428,
|
|
"step": 3876
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0285406795229798e-05,
|
|
"loss": 0.878,
|
|
"step": 3877
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0281179669461006e-05,
|
|
"loss": 0.8044,
|
|
"step": 3878
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.027695249340901e-05,
|
|
"loss": 0.7788,
|
|
"step": 3879
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0272725267829754e-05,
|
|
"loss": 0.8228,
|
|
"step": 3880
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.026849799347919e-05,
|
|
"loss": 0.7965,
|
|
"step": 3881
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0264270671113278e-05,
|
|
"loss": 0.7295,
|
|
"step": 3882
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0260043301487992e-05,
|
|
"loss": 0.7998,
|
|
"step": 3883
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0255815885359306e-05,
|
|
"loss": 0.8325,
|
|
"step": 3884
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0251588423483205e-05,
|
|
"loss": 0.7919,
|
|
"step": 3885
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0247360916615694e-05,
|
|
"loss": 0.7856,
|
|
"step": 3886
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.024313336551276e-05,
|
|
"loss": 0.7313,
|
|
"step": 3887
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.023890577093043e-05,
|
|
"loss": 0.8179,
|
|
"step": 3888
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0234678133624708e-05,
|
|
"loss": 0.8193,
|
|
"step": 3889
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.023045045435163e-05,
|
|
"loss": 0.8304,
|
|
"step": 3890
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0226222733867222e-05,
|
|
"loss": 0.7862,
|
|
"step": 3891
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0221994972927534e-05,
|
|
"loss": 0.8875,
|
|
"step": 3892
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0217767172288603e-05,
|
|
"loss": 0.8251,
|
|
"step": 3893
|
|
},
|
|
{
|
|
"epoch": 1.52,
|
|
"learning_rate": 1.0213539332706493e-05,
|
|
"loss": 0.7881,
|
|
"step": 3894
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0209311454937262e-05,
|
|
"loss": 0.8572,
|
|
"step": 3895
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.020508353973698e-05,
|
|
"loss": 0.8861,
|
|
"step": 3896
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0200855587861724e-05,
|
|
"loss": 0.8074,
|
|
"step": 3897
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0196627600067576e-05,
|
|
"loss": 0.8571,
|
|
"step": 3898
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0192399577110623e-05,
|
|
"loss": 0.8581,
|
|
"step": 3899
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0188171519746961e-05,
|
|
"loss": 0.861,
|
|
"step": 3900
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0183943428732692e-05,
|
|
"loss": 0.8674,
|
|
"step": 3901
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0179715304823921e-05,
|
|
"loss": 0.8268,
|
|
"step": 3902
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0175487148776762e-05,
|
|
"loss": 0.8074,
|
|
"step": 3903
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0171258961347332e-05,
|
|
"loss": 0.9292,
|
|
"step": 3904
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0167030743291764e-05,
|
|
"loss": 0.7417,
|
|
"step": 3905
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0162802495366175e-05,
|
|
"loss": 0.8159,
|
|
"step": 3906
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0158574218326712e-05,
|
|
"loss": 0.786,
|
|
"step": 3907
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0154345912929503e-05,
|
|
"loss": 0.839,
|
|
"step": 3908
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0150117579930707e-05,
|
|
"loss": 0.9004,
|
|
"step": 3909
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0145889220086465e-05,
|
|
"loss": 0.8285,
|
|
"step": 3910
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0141660834152932e-05,
|
|
"loss": 0.7538,
|
|
"step": 3911
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0137432422886277e-05,
|
|
"loss": 0.824,
|
|
"step": 3912
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0133203987042655e-05,
|
|
"loss": 0.7764,
|
|
"step": 3913
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.012897552737824e-05,
|
|
"loss": 0.7609,
|
|
"step": 3914
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.01247470446492e-05,
|
|
"loss": 0.8264,
|
|
"step": 3915
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.012051853961172e-05,
|
|
"loss": 0.7688,
|
|
"step": 3916
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0116290013021977e-05,
|
|
"loss": 0.8028,
|
|
"step": 3917
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0112061465636152e-05,
|
|
"loss": 0.8631,
|
|
"step": 3918
|
|
},
|
|
{
|
|
"epoch": 1.53,
|
|
"learning_rate": 1.0107832898210438e-05,
|
|
"loss": 0.7739,
|
|
"step": 3919
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.010360431150103e-05,
|
|
"loss": 0.7697,
|
|
"step": 3920
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0099375706264118e-05,
|
|
"loss": 0.7789,
|
|
"step": 3921
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0095147083255905e-05,
|
|
"loss": 0.7571,
|
|
"step": 3922
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0090918443232593e-05,
|
|
"loss": 0.7309,
|
|
"step": 3923
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0086689786950386e-05,
|
|
"loss": 0.8158,
|
|
"step": 3924
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0082461115165492e-05,
|
|
"loss": 0.7826,
|
|
"step": 3925
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.007823242863412e-05,
|
|
"loss": 0.7538,
|
|
"step": 3926
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0074003728112493e-05,
|
|
"loss": 0.7587,
|
|
"step": 3927
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0069775014356815e-05,
|
|
"loss": 0.7654,
|
|
"step": 3928
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0065546288123312e-05,
|
|
"loss": 0.7924,
|
|
"step": 3929
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0061317550168204e-05,
|
|
"loss": 0.7947,
|
|
"step": 3930
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0057088801247709e-05,
|
|
"loss": 0.7775,
|
|
"step": 3931
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.005286004211806e-05,
|
|
"loss": 0.8722,
|
|
"step": 3932
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0048631273535476e-05,
|
|
"loss": 0.8281,
|
|
"step": 3933
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.004440249625619e-05,
|
|
"loss": 0.7563,
|
|
"step": 3934
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0040173711036431e-05,
|
|
"loss": 0.7961,
|
|
"step": 3935
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0035944918632429e-05,
|
|
"loss": 0.8777,
|
|
"step": 3936
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0031716119800416e-05,
|
|
"loss": 0.8032,
|
|
"step": 3937
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0027487315296628e-05,
|
|
"loss": 0.8138,
|
|
"step": 3938
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0023258505877298e-05,
|
|
"loss": 0.8847,
|
|
"step": 3939
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0019029692298665e-05,
|
|
"loss": 0.8228,
|
|
"step": 3940
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0014800875316959e-05,
|
|
"loss": 0.8342,
|
|
"step": 3941
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0010572055688425e-05,
|
|
"loss": 0.7786,
|
|
"step": 3942
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.000634323416929e-05,
|
|
"loss": 0.7944,
|
|
"step": 3943
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 1.0002114411515806e-05,
|
|
"loss": 0.807,
|
|
"step": 3944
|
|
},
|
|
{
|
|
"epoch": 1.54,
|
|
"learning_rate": 9.997885588484199e-06,
|
|
"loss": 0.8422,
|
|
"step": 3945
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.993656765830711e-06,
|
|
"loss": 0.8037,
|
|
"step": 3946
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.989427944311578e-06,
|
|
"loss": 0.7721,
|
|
"step": 3947
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.985199124683041e-06,
|
|
"loss": 0.7753,
|
|
"step": 3948
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.98097030770134e-06,
|
|
"loss": 0.8474,
|
|
"step": 3949
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.976741494122705e-06,
|
|
"loss": 0.8222,
|
|
"step": 3950
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.972512684703373e-06,
|
|
"loss": 0.8474,
|
|
"step": 3951
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.968283880199585e-06,
|
|
"loss": 0.7966,
|
|
"step": 3952
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.96405508136757e-06,
|
|
"loss": 0.7725,
|
|
"step": 3953
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.95982628896357e-06,
|
|
"loss": 0.8256,
|
|
"step": 3954
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.955597503743811e-06,
|
|
"loss": 0.7992,
|
|
"step": 3955
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.951368726464526e-06,
|
|
"loss": 0.7978,
|
|
"step": 3956
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.94713995788194e-06,
|
|
"loss": 0.8319,
|
|
"step": 3957
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.942911198752293e-06,
|
|
"loss": 0.8024,
|
|
"step": 3958
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.9386824498318e-06,
|
|
"loss": 0.7495,
|
|
"step": 3959
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.934453711876692e-06,
|
|
"loss": 0.838,
|
|
"step": 3960
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.930224985643187e-06,
|
|
"loss": 0.8106,
|
|
"step": 3961
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.925996271887512e-06,
|
|
"loss": 0.7554,
|
|
"step": 3962
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.921767571365882e-06,
|
|
"loss": 0.7973,
|
|
"step": 3963
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.917538884834512e-06,
|
|
"loss": 0.8109,
|
|
"step": 3964
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.913310213049618e-06,
|
|
"loss": 0.8392,
|
|
"step": 3965
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.909081556767409e-06,
|
|
"loss": 0.8622,
|
|
"step": 3966
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.904852916744099e-06,
|
|
"loss": 0.803,
|
|
"step": 3967
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.900624293735885e-06,
|
|
"loss": 0.8187,
|
|
"step": 3968
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.896395688498973e-06,
|
|
"loss": 0.7927,
|
|
"step": 3969
|
|
},
|
|
{
|
|
"epoch": 1.55,
|
|
"learning_rate": 9.892167101789563e-06,
|
|
"loss": 0.8305,
|
|
"step": 3970
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.887938534363852e-06,
|
|
"loss": 0.757,
|
|
"step": 3971
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.883709986978028e-06,
|
|
"loss": 0.7527,
|
|
"step": 3972
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.879481460388283e-06,
|
|
"loss": 0.8307,
|
|
"step": 3973
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.875252955350798e-06,
|
|
"loss": 0.8396,
|
|
"step": 3974
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.871024472621765e-06,
|
|
"loss": 0.7579,
|
|
"step": 3975
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.866796012957347e-06,
|
|
"loss": 0.8451,
|
|
"step": 3976
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.862567577113726e-06,
|
|
"loss": 0.8913,
|
|
"step": 3977
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.858339165847067e-06,
|
|
"loss": 0.8244,
|
|
"step": 3978
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.854110779913537e-06,
|
|
"loss": 0.906,
|
|
"step": 3979
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.849882420069298e-06,
|
|
"loss": 0.7984,
|
|
"step": 3980
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.8456540870705e-06,
|
|
"loss": 0.8303,
|
|
"step": 3981
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.841425781673292e-06,
|
|
"loss": 0.8815,
|
|
"step": 3982
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.837197504633827e-06,
|
|
"loss": 0.8577,
|
|
"step": 3983
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.832969256708242e-06,
|
|
"loss": 0.7218,
|
|
"step": 3984
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.82874103865267e-06,
|
|
"loss": 0.9027,
|
|
"step": 3985
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.82451285122324e-06,
|
|
"loss": 0.7973,
|
|
"step": 3986
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.820284695176082e-06,
|
|
"loss": 0.7347,
|
|
"step": 3987
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.816056571267314e-06,
|
|
"loss": 0.7959,
|
|
"step": 3988
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.811828480253042e-06,
|
|
"loss": 0.7825,
|
|
"step": 3989
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.80760042288938e-06,
|
|
"loss": 0.7709,
|
|
"step": 3990
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.803372399932427e-06,
|
|
"loss": 0.8113,
|
|
"step": 3991
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.799144412138276e-06,
|
|
"loss": 0.8351,
|
|
"step": 3992
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.794916460263022e-06,
|
|
"loss": 0.6956,
|
|
"step": 3993
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.79068854506274e-06,
|
|
"loss": 0.8695,
|
|
"step": 3994
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.786460667293512e-06,
|
|
"loss": 0.7542,
|
|
"step": 3995
|
|
},
|
|
{
|
|
"epoch": 1.56,
|
|
"learning_rate": 9.782232827711399e-06,
|
|
"loss": 0.8237,
|
|
"step": 3996
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.778005027072473e-06,
|
|
"loss": 0.7817,
|
|
"step": 3997
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.773777266132782e-06,
|
|
"loss": 0.852,
|
|
"step": 3998
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.769549545648373e-06,
|
|
"loss": 0.7218,
|
|
"step": 3999
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.765321866375294e-06,
|
|
"loss": 0.7191,
|
|
"step": 4000
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.761094229069575e-06,
|
|
"loss": 0.8156,
|
|
"step": 4001
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.756866634487243e-06,
|
|
"loss": 0.8391,
|
|
"step": 4002
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.752639083384309e-06,
|
|
"loss": 0.8767,
|
|
"step": 4003
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.748411576516794e-06,
|
|
"loss": 0.8143,
|
|
"step": 4004
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.744184114640695e-06,
|
|
"loss": 0.8132,
|
|
"step": 4005
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.739956698512013e-06,
|
|
"loss": 0.8329,
|
|
"step": 4006
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.735729328886724e-06,
|
|
"loss": 0.8275,
|
|
"step": 4007
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.731502006520814e-06,
|
|
"loss": 0.7749,
|
|
"step": 4008
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.727274732170248e-06,
|
|
"loss": 0.7184,
|
|
"step": 4009
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.723047506590997e-06,
|
|
"loss": 0.7909,
|
|
"step": 4010
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.718820330538999e-06,
|
|
"loss": 0.763,
|
|
"step": 4011
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.714593204770206e-06,
|
|
"loss": 0.844,
|
|
"step": 4012
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.710366130040548e-06,
|
|
"loss": 0.8796,
|
|
"step": 4013
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.706139107105958e-06,
|
|
"loss": 0.789,
|
|
"step": 4014
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.701912136722347e-06,
|
|
"loss": 0.7855,
|
|
"step": 4015
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.697685219645622e-06,
|
|
"loss": 0.8094,
|
|
"step": 4016
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.693458356631677e-06,
|
|
"loss": 0.8366,
|
|
"step": 4017
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.68923154843641e-06,
|
|
"loss": 0.8068,
|
|
"step": 4018
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.685004795815688e-06,
|
|
"loss": 0.7887,
|
|
"step": 4019
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.680778099525382e-06,
|
|
"loss": 0.7689,
|
|
"step": 4020
|
|
},
|
|
{
|
|
"epoch": 1.57,
|
|
"learning_rate": 9.67655146032135e-06,
|
|
"loss": 0.7734,
|
|
"step": 4021
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.67232487895944e-06,
|
|
"loss": 0.7949,
|
|
"step": 4022
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.668098356195494e-06,
|
|
"loss": 0.7727,
|
|
"step": 4023
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.663871892785328e-06,
|
|
"loss": 0.7589,
|
|
"step": 4024
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.659645489484766e-06,
|
|
"loss": 0.8413,
|
|
"step": 4025
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.655419147049612e-06,
|
|
"loss": 0.731,
|
|
"step": 4026
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.651192866235662e-06,
|
|
"loss": 0.8767,
|
|
"step": 4027
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.646966647798692e-06,
|
|
"loss": 0.8434,
|
|
"step": 4028
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.642740492494483e-06,
|
|
"loss": 0.8494,
|
|
"step": 4029
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.638514401078789e-06,
|
|
"loss": 0.7936,
|
|
"step": 4030
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.634288374307369e-06,
|
|
"loss": 0.8045,
|
|
"step": 4031
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.63006241293595e-06,
|
|
"loss": 0.8024,
|
|
"step": 4032
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.625836517720265e-06,
|
|
"loss": 0.897,
|
|
"step": 4033
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.621610689416025e-06,
|
|
"loss": 0.8157,
|
|
"step": 4034
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.617384928778935e-06,
|
|
"loss": 0.8346,
|
|
"step": 4035
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.61315923656469e-06,
|
|
"loss": 0.8951,
|
|
"step": 4036
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.60893361352896e-06,
|
|
"loss": 0.8152,
|
|
"step": 4037
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.604708060427408e-06,
|
|
"loss": 0.7762,
|
|
"step": 4038
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.6004825780157e-06,
|
|
"loss": 0.8131,
|
|
"step": 4039
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.59625716704947e-06,
|
|
"loss": 0.8669,
|
|
"step": 4040
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.592031828284343e-06,
|
|
"loss": 0.7853,
|
|
"step": 4041
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.587806562475935e-06,
|
|
"loss": 0.8397,
|
|
"step": 4042
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.583581370379851e-06,
|
|
"loss": 0.838,
|
|
"step": 4043
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.579356252751682e-06,
|
|
"loss": 0.7522,
|
|
"step": 4044
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.575131210346998e-06,
|
|
"loss": 0.7819,
|
|
"step": 4045
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.570906243921358e-06,
|
|
"loss": 0.7748,
|
|
"step": 4046
|
|
},
|
|
{
|
|
"epoch": 1.58,
|
|
"learning_rate": 9.566681354230318e-06,
|
|
"loss": 0.809,
|
|
"step": 4047
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.562456542029409e-06,
|
|
"loss": 0.801,
|
|
"step": 4048
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.558231808074157e-06,
|
|
"loss": 0.8154,
|
|
"step": 4049
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.554007153120061e-06,
|
|
"loss": 0.7448,
|
|
"step": 4050
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.549782577922616e-06,
|
|
"loss": 0.8042,
|
|
"step": 4051
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.545558083237302e-06,
|
|
"loss": 0.875,
|
|
"step": 4052
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.541333669819585e-06,
|
|
"loss": 0.813,
|
|
"step": 4053
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.537109338424911e-06,
|
|
"loss": 0.8692,
|
|
"step": 4054
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.532885089808713e-06,
|
|
"loss": 0.8548,
|
|
"step": 4055
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.528660924726416e-06,
|
|
"loss": 0.883,
|
|
"step": 4056
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.524436843933427e-06,
|
|
"loss": 0.7615,
|
|
"step": 4057
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.520212848185127e-06,
|
|
"loss": 0.8252,
|
|
"step": 4058
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.515988938236895e-06,
|
|
"loss": 0.7928,
|
|
"step": 4059
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.511765114844093e-06,
|
|
"loss": 0.9005,
|
|
"step": 4060
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.507541378762064e-06,
|
|
"loss": 0.7935,
|
|
"step": 4061
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.503317730746138e-06,
|
|
"loss": 0.769,
|
|
"step": 4062
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.49909417155162e-06,
|
|
"loss": 0.7602,
|
|
"step": 4063
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.494870701933814e-06,
|
|
"loss": 0.8202,
|
|
"step": 4064
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.490647322647996e-06,
|
|
"loss": 0.8108,
|
|
"step": 4065
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.486424034449442e-06,
|
|
"loss": 0.8105,
|
|
"step": 4066
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.482200838093382e-06,
|
|
"loss": 0.8704,
|
|
"step": 4067
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.477977734335061e-06,
|
|
"loss": 0.7732,
|
|
"step": 4068
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.47375472392969e-06,
|
|
"loss": 0.906,
|
|
"step": 4069
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.469531807632474e-06,
|
|
"loss": 0.7631,
|
|
"step": 4070
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.465308986198581e-06,
|
|
"loss": 0.8342,
|
|
"step": 4071
|
|
},
|
|
{
|
|
"epoch": 1.59,
|
|
"learning_rate": 9.461086260383188e-06,
|
|
"loss": 0.7874,
|
|
"step": 4072
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.456863630941434e-06,
|
|
"loss": 0.7822,
|
|
"step": 4073
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.452641098628458e-06,
|
|
"loss": 0.8258,
|
|
"step": 4074
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.44841866419937e-06,
|
|
"loss": 0.8392,
|
|
"step": 4075
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.444196328409264e-06,
|
|
"loss": 0.7178,
|
|
"step": 4076
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.439974092013216e-06,
|
|
"loss": 0.8059,
|
|
"step": 4077
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.43575195576629e-06,
|
|
"loss": 0.8479,
|
|
"step": 4078
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.43152992042353e-06,
|
|
"loss": 0.8164,
|
|
"step": 4079
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.427307986739952e-06,
|
|
"loss": 0.8509,
|
|
"step": 4080
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.423086155470571e-06,
|
|
"loss": 0.8215,
|
|
"step": 4081
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.41886442737037e-06,
|
|
"loss": 0.7956,
|
|
"step": 4082
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.414642803194323e-06,
|
|
"loss": 0.8334,
|
|
"step": 4083
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.410421283697371e-06,
|
|
"loss": 0.8803,
|
|
"step": 4084
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.406199869634455e-06,
|
|
"loss": 0.8622,
|
|
"step": 4085
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.401978561760486e-06,
|
|
"loss": 0.7541,
|
|
"step": 4086
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.397757360830353e-06,
|
|
"loss": 0.8795,
|
|
"step": 4087
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.39353626759894e-06,
|
|
"loss": 0.8623,
|
|
"step": 4088
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.389315282821097e-06,
|
|
"loss": 0.7987,
|
|
"step": 4089
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.385094407251658e-06,
|
|
"loss": 0.8429,
|
|
"step": 4090
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.380873641645448e-06,
|
|
"loss": 0.8608,
|
|
"step": 4091
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.37665298675726e-06,
|
|
"loss": 0.8186,
|
|
"step": 4092
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.372432443341867e-06,
|
|
"loss": 0.8207,
|
|
"step": 4093
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.36821201215403e-06,
|
|
"loss": 0.8041,
|
|
"step": 4094
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.363991693948489e-06,
|
|
"loss": 0.8767,
|
|
"step": 4095
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.359771489479961e-06,
|
|
"loss": 0.8783,
|
|
"step": 4096
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.355551399503137e-06,
|
|
"loss": 0.8109,
|
|
"step": 4097
|
|
},
|
|
{
|
|
"epoch": 1.6,
|
|
"learning_rate": 9.351331424772694e-06,
|
|
"loss": 0.8939,
|
|
"step": 4098
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.347111566043295e-06,
|
|
"loss": 0.8045,
|
|
"step": 4099
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.342891824069566e-06,
|
|
"loss": 0.8264,
|
|
"step": 4100
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.338672199606134e-06,
|
|
"loss": 0.7823,
|
|
"step": 4101
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.334452693407577e-06,
|
|
"loss": 0.8477,
|
|
"step": 4102
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.330233306228475e-06,
|
|
"loss": 0.8103,
|
|
"step": 4103
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.326014038823376e-06,
|
|
"loss": 0.8074,
|
|
"step": 4104
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.321794891946813e-06,
|
|
"loss": 0.8827,
|
|
"step": 4105
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.317575866353293e-06,
|
|
"loss": 0.8315,
|
|
"step": 4106
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.313356962797297e-06,
|
|
"loss": 0.8043,
|
|
"step": 4107
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.30913818203329e-06,
|
|
"loss": 0.8227,
|
|
"step": 4108
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.304919524815724e-06,
|
|
"loss": 0.8016,
|
|
"step": 4109
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.300700991899007e-06,
|
|
"loss": 0.8636,
|
|
"step": 4110
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.296482584037542e-06,
|
|
"loss": 0.8515,
|
|
"step": 4111
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.292264301985702e-06,
|
|
"loss": 0.7996,
|
|
"step": 4112
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.288046146497844e-06,
|
|
"loss": 0.8593,
|
|
"step": 4113
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.2838281183283e-06,
|
|
"loss": 0.7278,
|
|
"step": 4114
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.27961021823137e-06,
|
|
"loss": 0.8025,
|
|
"step": 4115
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.275392446961344e-06,
|
|
"loss": 0.8722,
|
|
"step": 4116
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.271174805272482e-06,
|
|
"loss": 0.7353,
|
|
"step": 4117
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.266957293919029e-06,
|
|
"loss": 0.8418,
|
|
"step": 4118
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.262739913655186e-06,
|
|
"loss": 0.7918,
|
|
"step": 4119
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.258522665235155e-06,
|
|
"loss": 0.8487,
|
|
"step": 4120
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.254305549413106e-06,
|
|
"loss": 0.8561,
|
|
"step": 4121
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.250088566943179e-06,
|
|
"loss": 0.7365,
|
|
"step": 4122
|
|
},
|
|
{
|
|
"epoch": 1.61,
|
|
"learning_rate": 9.24587171857949e-06,
|
|
"loss": 0.8202,
|
|
"step": 4123
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.241655005076147e-06,
|
|
"loss": 0.8023,
|
|
"step": 4124
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.23743842718721e-06,
|
|
"loss": 0.877,
|
|
"step": 4125
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.233221985666739e-06,
|
|
"loss": 0.7503,
|
|
"step": 4126
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.229005681268755e-06,
|
|
"loss": 0.7595,
|
|
"step": 4127
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.224789514747252e-06,
|
|
"loss": 0.7421,
|
|
"step": 4128
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.220573486856205e-06,
|
|
"loss": 0.7714,
|
|
"step": 4129
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.21635759834957e-06,
|
|
"loss": 0.8079,
|
|
"step": 4130
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.21214184998127e-06,
|
|
"loss": 0.8179,
|
|
"step": 4131
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.207926242505202e-06,
|
|
"loss": 0.8077,
|
|
"step": 4132
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.20371077667524e-06,
|
|
"loss": 0.8103,
|
|
"step": 4133
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.199495453245238e-06,
|
|
"loss": 0.8479,
|
|
"step": 4134
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.195280272969019e-06,
|
|
"loss": 0.8249,
|
|
"step": 4135
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.191065236600378e-06,
|
|
"loss": 0.81,
|
|
"step": 4136
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.186850344893088e-06,
|
|
"loss": 0.7929,
|
|
"step": 4137
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.1826355986009e-06,
|
|
"loss": 0.804,
|
|
"step": 4138
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.178420998477534e-06,
|
|
"loss": 0.8294,
|
|
"step": 4139
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.174206545276678e-06,
|
|
"loss": 0.7993,
|
|
"step": 4140
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.169992239752007e-06,
|
|
"loss": 0.8075,
|
|
"step": 4141
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.165778082657162e-06,
|
|
"loss": 0.798,
|
|
"step": 4142
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.161564074745754e-06,
|
|
"loss": 0.8197,
|
|
"step": 4143
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.157350216771379e-06,
|
|
"loss": 0.8864,
|
|
"step": 4144
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.153136509487593e-06,
|
|
"loss": 0.8233,
|
|
"step": 4145
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.148922953647932e-06,
|
|
"loss": 0.7756,
|
|
"step": 4146
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.144709550005903e-06,
|
|
"loss": 0.7505,
|
|
"step": 4147
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.140496299314993e-06,
|
|
"loss": 0.8542,
|
|
"step": 4148
|
|
},
|
|
{
|
|
"epoch": 1.62,
|
|
"learning_rate": 9.136283202328649e-06,
|
|
"loss": 0.8571,
|
|
"step": 4149
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.132070259800295e-06,
|
|
"loss": 0.8184,
|
|
"step": 4150
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.127857472483335e-06,
|
|
"loss": 0.801,
|
|
"step": 4151
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.123644841131139e-06,
|
|
"loss": 0.8036,
|
|
"step": 4152
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.119432366497043e-06,
|
|
"loss": 0.8013,
|
|
"step": 4153
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.115220049334367e-06,
|
|
"loss": 0.8099,
|
|
"step": 4154
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.111007890396398e-06,
|
|
"loss": 0.8914,
|
|
"step": 4155
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.106795890436391e-06,
|
|
"loss": 0.8087,
|
|
"step": 4156
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.102584050207578e-06,
|
|
"loss": 0.8106,
|
|
"step": 4157
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.098372370463155e-06,
|
|
"loss": 0.8286,
|
|
"step": 4158
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.094160851956302e-06,
|
|
"loss": 0.8337,
|
|
"step": 4159
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.089949495440155e-06,
|
|
"loss": 0.7154,
|
|
"step": 4160
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.085738301667839e-06,
|
|
"loss": 0.8117,
|
|
"step": 4161
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.081527271392426e-06,
|
|
"loss": 0.846,
|
|
"step": 4162
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.07731640536698e-06,
|
|
"loss": 0.8439,
|
|
"step": 4163
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.073105704344526e-06,
|
|
"loss": 0.7501,
|
|
"step": 4164
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.068895169078068e-06,
|
|
"loss": 0.8278,
|
|
"step": 4165
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.064684800320564e-06,
|
|
"loss": 0.7972,
|
|
"step": 4166
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.060474598824954e-06,
|
|
"loss": 0.8404,
|
|
"step": 4167
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.056264565344148e-06,
|
|
"loss": 0.7821,
|
|
"step": 4168
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.052054700631024e-06,
|
|
"loss": 0.8066,
|
|
"step": 4169
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.04784500543843e-06,
|
|
"loss": 0.8263,
|
|
"step": 4170
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.043635480519182e-06,
|
|
"loss": 0.8032,
|
|
"step": 4171
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.039426126626067e-06,
|
|
"loss": 0.7735,
|
|
"step": 4172
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.035216944511843e-06,
|
|
"loss": 0.8107,
|
|
"step": 4173
|
|
},
|
|
{
|
|
"epoch": 1.63,
|
|
"learning_rate": 9.031007934929237e-06,
|
|
"loss": 0.8239,
|
|
"step": 4174
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 9.026799098630936e-06,
|
|
"loss": 0.8647,
|
|
"step": 4175
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 9.022590436369614e-06,
|
|
"loss": 0.7988,
|
|
"step": 4176
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 9.018381948897898e-06,
|
|
"loss": 0.8164,
|
|
"step": 4177
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 9.014173636968392e-06,
|
|
"loss": 0.8397,
|
|
"step": 4178
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 9.009965501333663e-06,
|
|
"loss": 0.7631,
|
|
"step": 4179
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 9.00575754274625e-06,
|
|
"loss": 0.7949,
|
|
"step": 4180
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 9.001549761958663e-06,
|
|
"loss": 0.8295,
|
|
"step": 4181
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.997342159723372e-06,
|
|
"loss": 0.7761,
|
|
"step": 4182
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.993134736792829e-06,
|
|
"loss": 0.8704,
|
|
"step": 4183
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.988927493919437e-06,
|
|
"loss": 0.7581,
|
|
"step": 4184
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.984720431855576e-06,
|
|
"loss": 0.7572,
|
|
"step": 4185
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.980513551353595e-06,
|
|
"loss": 0.8051,
|
|
"step": 4186
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.976306853165809e-06,
|
|
"loss": 0.7807,
|
|
"step": 4187
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.972100338044497e-06,
|
|
"loss": 0.7694,
|
|
"step": 4188
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.967894006741906e-06,
|
|
"loss": 0.7888,
|
|
"step": 4189
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.963687860010254e-06,
|
|
"loss": 0.7938,
|
|
"step": 4190
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.959481898601729e-06,
|
|
"loss": 0.7608,
|
|
"step": 4191
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.95527612326847e-06,
|
|
"loss": 0.837,
|
|
"step": 4192
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.951070534762602e-06,
|
|
"loss": 0.7766,
|
|
"step": 4193
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.946865133836205e-06,
|
|
"loss": 0.802,
|
|
"step": 4194
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.942659921241325e-06,
|
|
"loss": 0.7778,
|
|
"step": 4195
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.93845489772999e-06,
|
|
"loss": 0.7777,
|
|
"step": 4196
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.934250064054168e-06,
|
|
"loss": 0.7707,
|
|
"step": 4197
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.930045420965812e-06,
|
|
"loss": 0.798,
|
|
"step": 4198
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.925840969216836e-06,
|
|
"loss": 0.802,
|
|
"step": 4199
|
|
},
|
|
{
|
|
"epoch": 1.64,
|
|
"learning_rate": 8.921636709559127e-06,
|
|
"loss": 0.7875,
|
|
"step": 4200
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.917432642744519e-06,
|
|
"loss": 0.9035,
|
|
"step": 4201
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.91322876952483e-06,
|
|
"loss": 0.8062,
|
|
"step": 4202
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.90902509065183e-06,
|
|
"loss": 0.8643,
|
|
"step": 4203
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.904821606877269e-06,
|
|
"loss": 0.8062,
|
|
"step": 4204
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.900618318952847e-06,
|
|
"loss": 0.746,
|
|
"step": 4205
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.896415227630238e-06,
|
|
"loss": 0.8787,
|
|
"step": 4206
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.892212333661076e-06,
|
|
"loss": 0.8016,
|
|
"step": 4207
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.888009637796968e-06,
|
|
"loss": 0.8377,
|
|
"step": 4208
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.883807140789478e-06,
|
|
"loss": 0.725,
|
|
"step": 4209
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.879604843390131e-06,
|
|
"loss": 0.8054,
|
|
"step": 4210
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.875402746350425e-06,
|
|
"loss": 0.7805,
|
|
"step": 4211
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.871200850421822e-06,
|
|
"loss": 0.845,
|
|
"step": 4212
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.866999156355742e-06,
|
|
"loss": 0.7829,
|
|
"step": 4213
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.862797664903565e-06,
|
|
"loss": 0.7907,
|
|
"step": 4214
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.858596376816651e-06,
|
|
"loss": 0.7777,
|
|
"step": 4215
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.854395292846308e-06,
|
|
"loss": 0.9026,
|
|
"step": 4216
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.850194413743823e-06,
|
|
"loss": 0.7694,
|
|
"step": 4217
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.845993740260423e-06,
|
|
"loss": 0.8211,
|
|
"step": 4218
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.84179327314732e-06,
|
|
"loss": 0.7192,
|
|
"step": 4219
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.83759301315568e-06,
|
|
"loss": 0.7908,
|
|
"step": 4220
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.833392961036634e-06,
|
|
"loss": 0.7837,
|
|
"step": 4221
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.829193117541276e-06,
|
|
"loss": 0.859,
|
|
"step": 4222
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.824993483420658e-06,
|
|
"loss": 0.8552,
|
|
"step": 4223
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.8207940594258e-06,
|
|
"loss": 0.8738,
|
|
"step": 4224
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.816594846307682e-06,
|
|
"loss": 0.8478,
|
|
"step": 4225
|
|
},
|
|
{
|
|
"epoch": 1.65,
|
|
"learning_rate": 8.812395844817252e-06,
|
|
"loss": 0.8182,
|
|
"step": 4226
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.808197055705406e-06,
|
|
"loss": 0.8083,
|
|
"step": 4227
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.803998479723015e-06,
|
|
"loss": 0.8096,
|
|
"step": 4228
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.799800117620911e-06,
|
|
"loss": 0.8412,
|
|
"step": 4229
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.795601970149884e-06,
|
|
"loss": 0.8434,
|
|
"step": 4230
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.791404038060686e-06,
|
|
"loss": 0.8566,
|
|
"step": 4231
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.787206322104025e-06,
|
|
"loss": 0.8567,
|
|
"step": 4232
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.783008823030584e-06,
|
|
"loss": 0.7719,
|
|
"step": 4233
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.778811541590994e-06,
|
|
"loss": 0.7787,
|
|
"step": 4234
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.77461447853586e-06,
|
|
"loss": 0.7747,
|
|
"step": 4235
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.770417634615734e-06,
|
|
"loss": 0.8894,
|
|
"step": 4236
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.766221010581134e-06,
|
|
"loss": 0.7762,
|
|
"step": 4237
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.762024607182542e-06,
|
|
"loss": 0.799,
|
|
"step": 4238
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.757828425170403e-06,
|
|
"loss": 0.8314,
|
|
"step": 4239
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.753632465295113e-06,
|
|
"loss": 0.8153,
|
|
"step": 4240
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.74943672830703e-06,
|
|
"loss": 0.7625,
|
|
"step": 4241
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.745241214956484e-06,
|
|
"loss": 0.756,
|
|
"step": 4242
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.741045925993753e-06,
|
|
"loss": 0.8341,
|
|
"step": 4243
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.736850862169075e-06,
|
|
"loss": 0.7671,
|
|
"step": 4244
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.73265602423265e-06,
|
|
"loss": 0.8494,
|
|
"step": 4245
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.728461412934643e-06,
|
|
"loss": 0.7796,
|
|
"step": 4246
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.724267029025175e-06,
|
|
"loss": 0.77,
|
|
"step": 4247
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.720072873254322e-06,
|
|
"loss": 0.7522,
|
|
"step": 4248
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.71587894637212e-06,
|
|
"loss": 0.8872,
|
|
"step": 4249
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.711685249128574e-06,
|
|
"loss": 0.8063,
|
|
"step": 4250
|
|
},
|
|
{
|
|
"epoch": 1.66,
|
|
"learning_rate": 8.707491782273632e-06,
|
|
"loss": 0.8462,
|
|
"step": 4251
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.703298546557221e-06,
|
|
"loss": 0.8097,
|
|
"step": 4252
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.6991055427292e-06,
|
|
"loss": 0.7728,
|
|
"step": 4253
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.694912771539413e-06,
|
|
"loss": 0.8536,
|
|
"step": 4254
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.690720233737645e-06,
|
|
"loss": 0.7718,
|
|
"step": 4255
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.686527930073653e-06,
|
|
"loss": 0.771,
|
|
"step": 4256
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.682335861297133e-06,
|
|
"loss": 0.804,
|
|
"step": 4257
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.678144028157758e-06,
|
|
"loss": 0.8093,
|
|
"step": 4258
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.673952431405148e-06,
|
|
"loss": 0.7686,
|
|
"step": 4259
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.66976107178889e-06,
|
|
"loss": 0.8656,
|
|
"step": 4260
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.665569950058514e-06,
|
|
"loss": 0.8794,
|
|
"step": 4261
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.66137906696352e-06,
|
|
"loss": 0.8984,
|
|
"step": 4262
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.65718842325336e-06,
|
|
"loss": 0.7871,
|
|
"step": 4263
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.652998019677446e-06,
|
|
"loss": 0.7959,
|
|
"step": 4264
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.648807856985146e-06,
|
|
"loss": 0.769,
|
|
"step": 4265
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.64461793592578e-06,
|
|
"loss": 0.7574,
|
|
"step": 4266
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.640428257248637e-06,
|
|
"loss": 0.8181,
|
|
"step": 4267
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.63623882170295e-06,
|
|
"loss": 0.8141,
|
|
"step": 4268
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.632049630037914e-06,
|
|
"loss": 0.8969,
|
|
"step": 4269
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.627860683002677e-06,
|
|
"loss": 0.7,
|
|
"step": 4270
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.62367198134635e-06,
|
|
"loss": 0.7931,
|
|
"step": 4271
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.619483525817995e-06,
|
|
"loss": 0.7383,
|
|
"step": 4272
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.615295317166634e-06,
|
|
"loss": 0.8179,
|
|
"step": 4273
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.611107356141236e-06,
|
|
"loss": 0.8355,
|
|
"step": 4274
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.606919643490736e-06,
|
|
"loss": 0.7422,
|
|
"step": 4275
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.602732179964017e-06,
|
|
"loss": 0.7633,
|
|
"step": 4276
|
|
},
|
|
{
|
|
"epoch": 1.67,
|
|
"learning_rate": 8.598544966309926e-06,
|
|
"loss": 0.8341,
|
|
"step": 4277
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.594358003277257e-06,
|
|
"loss": 0.7729,
|
|
"step": 4278
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.590171291614763e-06,
|
|
"loss": 0.8459,
|
|
"step": 4279
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.585984832071147e-06,
|
|
"loss": 0.7934,
|
|
"step": 4280
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.581798625395077e-06,
|
|
"loss": 0.8081,
|
|
"step": 4281
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.577612672335172e-06,
|
|
"loss": 0.7419,
|
|
"step": 4282
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.573426973639996e-06,
|
|
"loss": 0.8078,
|
|
"step": 4283
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.569241530058075e-06,
|
|
"loss": 0.8035,
|
|
"step": 4284
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.565056342337898e-06,
|
|
"loss": 0.8523,
|
|
"step": 4285
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.560871411227896e-06,
|
|
"loss": 0.8181,
|
|
"step": 4286
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.556686737476455e-06,
|
|
"loss": 0.811,
|
|
"step": 4287
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.552502321831917e-06,
|
|
"loss": 0.8032,
|
|
"step": 4288
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.548318165042584e-06,
|
|
"loss": 0.8476,
|
|
"step": 4289
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.544134267856703e-06,
|
|
"loss": 0.7608,
|
|
"step": 4290
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.539950631022483e-06,
|
|
"loss": 0.796,
|
|
"step": 4291
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.535767255288075e-06,
|
|
"loss": 0.8396,
|
|
"step": 4292
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.531584141401592e-06,
|
|
"loss": 0.8331,
|
|
"step": 4293
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.527401290111095e-06,
|
|
"loss": 0.837,
|
|
"step": 4294
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.52321870216461e-06,
|
|
"loss": 0.7782,
|
|
"step": 4295
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.519036378310098e-06,
|
|
"loss": 0.7593,
|
|
"step": 4296
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.514854319295484e-06,
|
|
"loss": 0.7431,
|
|
"step": 4297
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.510672525868643e-06,
|
|
"loss": 0.7564,
|
|
"step": 4298
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.506490998777407e-06,
|
|
"loss": 0.8382,
|
|
"step": 4299
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.502309738769551e-06,
|
|
"loss": 0.8027,
|
|
"step": 4300
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.498128746592806e-06,
|
|
"loss": 0.8287,
|
|
"step": 4301
|
|
},
|
|
{
|
|
"epoch": 1.68,
|
|
"learning_rate": 8.493948022994864e-06,
|
|
"loss": 0.8767,
|
|
"step": 4302
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.489767568723357e-06,
|
|
"loss": 0.8121,
|
|
"step": 4303
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.485587384525873e-06,
|
|
"loss": 0.7926,
|
|
"step": 4304
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.48140747114995e-06,
|
|
"loss": 0.7927,
|
|
"step": 4305
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.477227829343084e-06,
|
|
"loss": 0.8295,
|
|
"step": 4306
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.473048459852716e-06,
|
|
"loss": 0.8415,
|
|
"step": 4307
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.468869363426243e-06,
|
|
"loss": 0.8044,
|
|
"step": 4308
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.464690540811002e-06,
|
|
"loss": 0.7976,
|
|
"step": 4309
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.4605119927543e-06,
|
|
"loss": 0.8071,
|
|
"step": 4310
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.456333720003376e-06,
|
|
"loss": 0.8337,
|
|
"step": 4311
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.452155723305439e-06,
|
|
"loss": 0.788,
|
|
"step": 4312
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.447978003407625e-06,
|
|
"loss": 0.7852,
|
|
"step": 4313
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.44380056105704e-06,
|
|
"loss": 0.7703,
|
|
"step": 4314
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.43962339700073e-06,
|
|
"loss": 0.7823,
|
|
"step": 4315
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.4354465119857e-06,
|
|
"loss": 0.7354,
|
|
"step": 4316
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.431269906758903e-06,
|
|
"loss": 0.8178,
|
|
"step": 4317
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.427093582067228e-06,
|
|
"loss": 0.8224,
|
|
"step": 4318
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.422917538657532e-06,
|
|
"loss": 0.7801,
|
|
"step": 4319
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.418741777276611e-06,
|
|
"loss": 0.7448,
|
|
"step": 4320
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.414566298671221e-06,
|
|
"loss": 0.7576,
|
|
"step": 4321
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.410391103588052e-06,
|
|
"loss": 0.8008,
|
|
"step": 4322
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.406216192773755e-06,
|
|
"loss": 0.7542,
|
|
"step": 4323
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.40204156697493e-06,
|
|
"loss": 0.849,
|
|
"step": 4324
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.39786722693812e-06,
|
|
"loss": 0.7701,
|
|
"step": 4325
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.393693173409818e-06,
|
|
"loss": 0.8408,
|
|
"step": 4326
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.389519407136471e-06,
|
|
"loss": 0.8236,
|
|
"step": 4327
|
|
},
|
|
{
|
|
"epoch": 1.69,
|
|
"learning_rate": 8.38534592886447e-06,
|
|
"loss": 0.787,
|
|
"step": 4328
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.381172739340156e-06,
|
|
"loss": 0.779,
|
|
"step": 4329
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.376999839309821e-06,
|
|
"loss": 0.7274,
|
|
"step": 4330
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.372827229519697e-06,
|
|
"loss": 0.7992,
|
|
"step": 4331
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.368654910715975e-06,
|
|
"loss": 0.8074,
|
|
"step": 4332
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.36448288364478e-06,
|
|
"loss": 0.8577,
|
|
"step": 4333
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.360311149052205e-06,
|
|
"loss": 0.7432,
|
|
"step": 4334
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.35613970768427e-06,
|
|
"loss": 0.8101,
|
|
"step": 4335
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.351968560286951e-06,
|
|
"loss": 0.7658,
|
|
"step": 4336
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.34779770760618e-06,
|
|
"loss": 0.7583,
|
|
"step": 4337
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.343627150387825e-06,
|
|
"loss": 0.8105,
|
|
"step": 4338
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.3394568893777e-06,
|
|
"loss": 0.8043,
|
|
"step": 4339
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.335286925321571e-06,
|
|
"loss": 0.7719,
|
|
"step": 4340
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.331117258965155e-06,
|
|
"loss": 0.8299,
|
|
"step": 4341
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.32694789105411e-06,
|
|
"loss": 0.863,
|
|
"step": 4342
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.322778822334042e-06,
|
|
"loss": 0.8132,
|
|
"step": 4343
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.318610053550498e-06,
|
|
"loss": 0.7884,
|
|
"step": 4344
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.314441585448982e-06,
|
|
"loss": 0.7976,
|
|
"step": 4345
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.310273418774938e-06,
|
|
"loss": 0.7893,
|
|
"step": 4346
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.306105554273758e-06,
|
|
"loss": 0.7694,
|
|
"step": 4347
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.301937992690773e-06,
|
|
"loss": 0.7762,
|
|
"step": 4348
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.297770734771273e-06,
|
|
"loss": 0.6976,
|
|
"step": 4349
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.29360378126048e-06,
|
|
"loss": 0.7675,
|
|
"step": 4350
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.289437132903574e-06,
|
|
"loss": 0.8533,
|
|
"step": 4351
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.285270790445671e-06,
|
|
"loss": 0.8195,
|
|
"step": 4352
|
|
},
|
|
{
|
|
"epoch": 1.7,
|
|
"learning_rate": 8.281104754631836e-06,
|
|
"loss": 0.7351,
|
|
"step": 4353
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.276939026207076e-06,
|
|
"loss": 0.7509,
|
|
"step": 4354
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.272773605916353e-06,
|
|
"loss": 0.7934,
|
|
"step": 4355
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.268608494504564e-06,
|
|
"loss": 0.8445,
|
|
"step": 4356
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.26444369271655e-06,
|
|
"loss": 0.8418,
|
|
"step": 4357
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.2602792012971e-06,
|
|
"loss": 0.7669,
|
|
"step": 4358
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.256115020990955e-06,
|
|
"loss": 0.7356,
|
|
"step": 4359
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.251951152542788e-06,
|
|
"loss": 0.8721,
|
|
"step": 4360
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.247787596697217e-06,
|
|
"loss": 0.7346,
|
|
"step": 4361
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.243624354198817e-06,
|
|
"loss": 0.7607,
|
|
"step": 4362
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.239461425792092e-06,
|
|
"loss": 0.7989,
|
|
"step": 4363
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.235298812221503e-06,
|
|
"loss": 0.7434,
|
|
"step": 4364
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.231136514231437e-06,
|
|
"loss": 0.8869,
|
|
"step": 4365
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.226974532566245e-06,
|
|
"loss": 0.8556,
|
|
"step": 4366
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.222812867970207e-06,
|
|
"loss": 0.7082,
|
|
"step": 4367
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.218651521187552e-06,
|
|
"loss": 0.8525,
|
|
"step": 4368
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.214490492962457e-06,
|
|
"loss": 0.7631,
|
|
"step": 4369
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.210329784039028e-06,
|
|
"loss": 0.7796,
|
|
"step": 4370
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.206169395161324e-06,
|
|
"loss": 0.8406,
|
|
"step": 4371
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.20200932707335e-06,
|
|
"loss": 0.8083,
|
|
"step": 4372
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.197849580519045e-06,
|
|
"loss": 0.8238,
|
|
"step": 4373
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.193690156242293e-06,
|
|
"loss": 0.7899,
|
|
"step": 4374
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.189531054986922e-06,
|
|
"loss": 0.7868,
|
|
"step": 4375
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.185372277496704e-06,
|
|
"loss": 0.7986,
|
|
"step": 4376
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.181213824515353e-06,
|
|
"loss": 0.815,
|
|
"step": 4377
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.177055696786516e-06,
|
|
"loss": 0.7611,
|
|
"step": 4378
|
|
},
|
|
{
|
|
"epoch": 1.71,
|
|
"learning_rate": 8.17289789505379e-06,
|
|
"loss": 0.7756,
|
|
"step": 4379
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.168740420060718e-06,
|
|
"loss": 0.812,
|
|
"step": 4380
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.164583272550777e-06,
|
|
"loss": 0.7329,
|
|
"step": 4381
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.160426453267383e-06,
|
|
"loss": 0.8778,
|
|
"step": 4382
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.156269962953899e-06,
|
|
"loss": 0.7903,
|
|
"step": 4383
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.152113802353631e-06,
|
|
"loss": 0.7883,
|
|
"step": 4384
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.147957972209819e-06,
|
|
"loss": 0.8997,
|
|
"step": 4385
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.143802473265655e-06,
|
|
"loss": 0.6767,
|
|
"step": 4386
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.139647306264257e-06,
|
|
"loss": 0.791,
|
|
"step": 4387
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.135492471948694e-06,
|
|
"loss": 0.8019,
|
|
"step": 4388
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.13133797106197e-06,
|
|
"loss": 0.8927,
|
|
"step": 4389
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.127183804347038e-06,
|
|
"loss": 0.7454,
|
|
"step": 4390
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.123029972546782e-06,
|
|
"loss": 0.7712,
|
|
"step": 4391
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.118876476404027e-06,
|
|
"loss": 0.7535,
|
|
"step": 4392
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.114723316661541e-06,
|
|
"loss": 0.844,
|
|
"step": 4393
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.110570494062037e-06,
|
|
"loss": 0.7738,
|
|
"step": 4394
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.106418009348157e-06,
|
|
"loss": 0.7519,
|
|
"step": 4395
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.102265863262486e-06,
|
|
"loss": 0.7833,
|
|
"step": 4396
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.098114056547553e-06,
|
|
"loss": 0.7815,
|
|
"step": 4397
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.093962589945824e-06,
|
|
"loss": 0.8694,
|
|
"step": 4398
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.089811464199703e-06,
|
|
"loss": 0.8585,
|
|
"step": 4399
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.08566068005153e-06,
|
|
"loss": 0.8169,
|
|
"step": 4400
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.081510238243591e-06,
|
|
"loss": 0.8747,
|
|
"step": 4401
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.077360139518104e-06,
|
|
"loss": 0.7949,
|
|
"step": 4402
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.07321038461724e-06,
|
|
"loss": 0.8065,
|
|
"step": 4403
|
|
},
|
|
{
|
|
"epoch": 1.72,
|
|
"learning_rate": 8.069060974283079e-06,
|
|
"loss": 0.8329,
|
|
"step": 4404
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.06491190925767e-06,
|
|
"loss": 0.7769,
|
|
"step": 4405
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.060763190282983e-06,
|
|
"loss": 0.8007,
|
|
"step": 4406
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.05661481810094e-06,
|
|
"loss": 0.806,
|
|
"step": 4407
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.052466793453378e-06,
|
|
"loss": 0.7477,
|
|
"step": 4408
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.048319117082096e-06,
|
|
"loss": 0.8093,
|
|
"step": 4409
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.044171789728816e-06,
|
|
"loss": 0.8056,
|
|
"step": 4410
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.040024812135207e-06,
|
|
"loss": 0.722,
|
|
"step": 4411
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.035878185042869e-06,
|
|
"loss": 0.7975,
|
|
"step": 4412
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.031731909193337e-06,
|
|
"loss": 0.8435,
|
|
"step": 4413
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.027585985328088e-06,
|
|
"loss": 0.8134,
|
|
"step": 4414
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.02344041418854e-06,
|
|
"loss": 0.8147,
|
|
"step": 4415
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.019295196516044e-06,
|
|
"loss": 0.8164,
|
|
"step": 4416
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.015150333051878e-06,
|
|
"loss": 0.814,
|
|
"step": 4417
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.011005824537272e-06,
|
|
"loss": 0.7592,
|
|
"step": 4418
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.006861671713386e-06,
|
|
"loss": 0.7789,
|
|
"step": 4419
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 8.002717875321317e-06,
|
|
"loss": 0.8337,
|
|
"step": 4420
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.998574436102095e-06,
|
|
"loss": 0.8312,
|
|
"step": 4421
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.99443135479669e-06,
|
|
"loss": 0.8136,
|
|
"step": 4422
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.990288632146009e-06,
|
|
"loss": 0.8482,
|
|
"step": 4423
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.986146268890887e-06,
|
|
"loss": 0.871,
|
|
"step": 4424
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.98200426577211e-06,
|
|
"loss": 0.8034,
|
|
"step": 4425
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.977862623530383e-06,
|
|
"loss": 0.822,
|
|
"step": 4426
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.973721342906352e-06,
|
|
"loss": 0.7804,
|
|
"step": 4427
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.969580424640608e-06,
|
|
"loss": 0.7997,
|
|
"step": 4428
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.965439869473664e-06,
|
|
"loss": 0.8052,
|
|
"step": 4429
|
|
},
|
|
{
|
|
"epoch": 1.73,
|
|
"learning_rate": 7.961299678145974e-06,
|
|
"loss": 0.8113,
|
|
"step": 4430
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.95715985139792e-06,
|
|
"loss": 0.7294,
|
|
"step": 4431
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.953020389969835e-06,
|
|
"loss": 0.7438,
|
|
"step": 4432
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.948881294601973e-06,
|
|
"loss": 0.8512,
|
|
"step": 4433
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.944742566034524e-06,
|
|
"loss": 0.8033,
|
|
"step": 4434
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.940604205007614e-06,
|
|
"loss": 0.7952,
|
|
"step": 4435
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.936466212261306e-06,
|
|
"loss": 0.8303,
|
|
"step": 4436
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.932328588535594e-06,
|
|
"loss": 0.8085,
|
|
"step": 4437
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.928191334570413e-06,
|
|
"loss": 0.7746,
|
|
"step": 4438
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.924054451105614e-06,
|
|
"loss": 0.8011,
|
|
"step": 4439
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.919917938881002e-06,
|
|
"loss": 0.7804,
|
|
"step": 4440
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.915781798636303e-06,
|
|
"loss": 0.862,
|
|
"step": 4441
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.911646031111188e-06,
|
|
"loss": 0.7497,
|
|
"step": 4442
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.907510637045244e-06,
|
|
"loss": 0.8123,
|
|
"step": 4443
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.903375617178007e-06,
|
|
"loss": 0.7875,
|
|
"step": 4444
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.899240972248937e-06,
|
|
"loss": 0.7655,
|
|
"step": 4445
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.895106702997437e-06,
|
|
"loss": 0.8114,
|
|
"step": 4446
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.890972810162829e-06,
|
|
"loss": 0.7881,
|
|
"step": 4447
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.886839294484378e-06,
|
|
"loss": 0.8595,
|
|
"step": 4448
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.882706156701274e-06,
|
|
"loss": 0.8158,
|
|
"step": 4449
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.87857339755265e-06,
|
|
"loss": 0.8173,
|
|
"step": 4450
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.874441017777564e-06,
|
|
"loss": 0.7664,
|
|
"step": 4451
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.870309018115002e-06,
|
|
"loss": 0.818,
|
|
"step": 4452
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.866177399303893e-06,
|
|
"loss": 0.7657,
|
|
"step": 4453
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.862046162083088e-06,
|
|
"loss": 0.7965,
|
|
"step": 4454
|
|
},
|
|
{
|
|
"epoch": 1.74,
|
|
"learning_rate": 7.85791530719138e-06,
|
|
"loss": 0.8254,
|
|
"step": 4455
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.853784835367478e-06,
|
|
"loss": 0.7457,
|
|
"step": 4456
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.849654747350037e-06,
|
|
"loss": 0.7871,
|
|
"step": 4457
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.84552504387764e-06,
|
|
"loss": 0.8517,
|
|
"step": 4458
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.8413957256888e-06,
|
|
"loss": 0.8231,
|
|
"step": 4459
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.837266793521955e-06,
|
|
"loss": 0.8196,
|
|
"step": 4460
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.833138248115483e-06,
|
|
"loss": 0.818,
|
|
"step": 4461
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.82901009020769e-06,
|
|
"loss": 0.7278,
|
|
"step": 4462
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.824882320536814e-06,
|
|
"loss": 0.7977,
|
|
"step": 4463
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.820754939841021e-06,
|
|
"loss": 0.7362,
|
|
"step": 4464
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.816627948858404e-06,
|
|
"loss": 0.8145,
|
|
"step": 4465
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.812501348326993e-06,
|
|
"loss": 0.8339,
|
|
"step": 4466
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.808375138984747e-06,
|
|
"loss": 0.7642,
|
|
"step": 4467
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.804249321569555e-06,
|
|
"loss": 0.759,
|
|
"step": 4468
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.80012389681923e-06,
|
|
"loss": 0.7801,
|
|
"step": 4469
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.795998865471519e-06,
|
|
"loss": 0.7784,
|
|
"step": 4470
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.791874228264105e-06,
|
|
"loss": 0.7548,
|
|
"step": 4471
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.787749985934591e-06,
|
|
"loss": 0.8672,
|
|
"step": 4472
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.783626139220514e-06,
|
|
"loss": 0.8644,
|
|
"step": 4473
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.779502688859335e-06,
|
|
"loss": 0.7342,
|
|
"step": 4474
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.775379635588454e-06,
|
|
"loss": 0.7318,
|
|
"step": 4475
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.77125698014519e-06,
|
|
"loss": 0.8047,
|
|
"step": 4476
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.767134723266804e-06,
|
|
"loss": 0.8356,
|
|
"step": 4477
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.763012865690465e-06,
|
|
"loss": 0.8046,
|
|
"step": 4478
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.75889140815329e-06,
|
|
"loss": 0.7813,
|
|
"step": 4479
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.754770351392311e-06,
|
|
"loss": 0.8499,
|
|
"step": 4480
|
|
},
|
|
{
|
|
"epoch": 1.75,
|
|
"learning_rate": 7.750649696144503e-06,
|
|
"loss": 0.8758,
|
|
"step": 4481
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.746529443146752e-06,
|
|
"loss": 0.788,
|
|
"step": 4482
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.742409593135885e-06,
|
|
"loss": 0.832,
|
|
"step": 4483
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.73829014684865e-06,
|
|
"loss": 0.7709,
|
|
"step": 4484
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.73417110502173e-06,
|
|
"loss": 0.8389,
|
|
"step": 4485
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.730052468391726e-06,
|
|
"loss": 0.8182,
|
|
"step": 4486
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.725934237695171e-06,
|
|
"loss": 0.8575,
|
|
"step": 4487
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.72181641366853e-06,
|
|
"loss": 0.7853,
|
|
"step": 4488
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.717698997048188e-06,
|
|
"loss": 0.7422,
|
|
"step": 4489
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.713581988570465e-06,
|
|
"loss": 0.8337,
|
|
"step": 4490
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.709465388971596e-06,
|
|
"loss": 0.8187,
|
|
"step": 4491
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.705349198987758e-06,
|
|
"loss": 0.7522,
|
|
"step": 4492
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.70123341935504e-06,
|
|
"loss": 0.7659,
|
|
"step": 4493
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.69711805080947e-06,
|
|
"loss": 0.8696,
|
|
"step": 4494
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.693003094086992e-06,
|
|
"loss": 0.8403,
|
|
"step": 4495
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.688888549923485e-06,
|
|
"loss": 0.7759,
|
|
"step": 4496
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.684774419054748e-06,
|
|
"loss": 0.8307,
|
|
"step": 4497
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.680660702216519e-06,
|
|
"loss": 0.8374,
|
|
"step": 4498
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.676547400144433e-06,
|
|
"loss": 0.7932,
|
|
"step": 4499
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.672434513574085e-06,
|
|
"loss": 0.81,
|
|
"step": 4500
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.66832204324097e-06,
|
|
"loss": 0.8405,
|
|
"step": 4501
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.66420998988053e-06,
|
|
"loss": 0.8269,
|
|
"step": 4502
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.660098354228112e-06,
|
|
"loss": 0.7671,
|
|
"step": 4503
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.655987137019001e-06,
|
|
"loss": 0.8223,
|
|
"step": 4504
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.6518763389884e-06,
|
|
"loss": 0.7589,
|
|
"step": 4505
|
|
},
|
|
{
|
|
"epoch": 1.76,
|
|
"learning_rate": 7.647765960871447e-06,
|
|
"loss": 0.778,
|
|
"step": 4506
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.643656003403198e-06,
|
|
"loss": 0.84,
|
|
"step": 4507
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.639546467318629e-06,
|
|
"loss": 0.8073,
|
|
"step": 4508
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.635437353352644e-06,
|
|
"loss": 0.838,
|
|
"step": 4509
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.631328662240085e-06,
|
|
"loss": 0.7843,
|
|
"step": 4510
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.6272203947157e-06,
|
|
"loss": 0.7505,
|
|
"step": 4511
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.6231125515141645e-06,
|
|
"loss": 0.8499,
|
|
"step": 4512
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.619005133370089e-06,
|
|
"loss": 0.8148,
|
|
"step": 4513
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.6148981410179966e-06,
|
|
"loss": 0.7729,
|
|
"step": 4514
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.610791575192342e-06,
|
|
"loss": 0.7621,
|
|
"step": 4515
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.606685436627494e-06,
|
|
"loss": 0.7803,
|
|
"step": 4516
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.602579726057757e-06,
|
|
"loss": 0.8712,
|
|
"step": 4517
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.59847444421735e-06,
|
|
"loss": 0.8574,
|
|
"step": 4518
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.594369591840416e-06,
|
|
"loss": 0.9028,
|
|
"step": 4519
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.590265169661032e-06,
|
|
"loss": 0.7645,
|
|
"step": 4520
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.586161178413181e-06,
|
|
"loss": 0.7132,
|
|
"step": 4521
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.582057618830778e-06,
|
|
"loss": 0.7777,
|
|
"step": 4522
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.577954491647665e-06,
|
|
"loss": 0.7977,
|
|
"step": 4523
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.573851797597602e-06,
|
|
"loss": 0.7945,
|
|
"step": 4524
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.569749537414266e-06,
|
|
"loss": 0.7132,
|
|
"step": 4525
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.565647711831263e-06,
|
|
"loss": 0.7512,
|
|
"step": 4526
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.561546321582125e-06,
|
|
"loss": 0.8519,
|
|
"step": 4527
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.557445367400302e-06,
|
|
"loss": 0.8205,
|
|
"step": 4528
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.5533448500191565e-06,
|
|
"loss": 0.8355,
|
|
"step": 4529
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.549244770171986e-06,
|
|
"loss": 0.7235,
|
|
"step": 4530
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.545145128592009e-06,
|
|
"loss": 0.8203,
|
|
"step": 4531
|
|
},
|
|
{
|
|
"epoch": 1.77,
|
|
"learning_rate": 7.5410459260123585e-06,
|
|
"loss": 0.767,
|
|
"step": 4532
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.536947163166099e-06,
|
|
"loss": 0.871,
|
|
"step": 4533
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.532848840786198e-06,
|
|
"loss": 0.791,
|
|
"step": 4534
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.528750959605565e-06,
|
|
"loss": 0.749,
|
|
"step": 4535
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.524653520357019e-06,
|
|
"loss": 0.7847,
|
|
"step": 4536
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.520556523773307e-06,
|
|
"loss": 0.7908,
|
|
"step": 4537
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.516459970587086e-06,
|
|
"loss": 0.8102,
|
|
"step": 4538
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.512363861530944e-06,
|
|
"loss": 0.7418,
|
|
"step": 4539
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.508268197337385e-06,
|
|
"loss": 0.837,
|
|
"step": 4540
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.504172978738837e-06,
|
|
"loss": 0.7612,
|
|
"step": 4541
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.500078206467642e-06,
|
|
"loss": 0.8337,
|
|
"step": 4542
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.495983881256067e-06,
|
|
"loss": 0.8296,
|
|
"step": 4543
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.491890003836295e-06,
|
|
"loss": 0.7839,
|
|
"step": 4544
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.487796574940439e-06,
|
|
"loss": 0.7735,
|
|
"step": 4545
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.483703595300522e-06,
|
|
"loss": 0.7598,
|
|
"step": 4546
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.479611065648482e-06,
|
|
"loss": 0.7892,
|
|
"step": 4547
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.475518986716193e-06,
|
|
"loss": 0.884,
|
|
"step": 4548
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.471427359235437e-06,
|
|
"loss": 0.744,
|
|
"step": 4549
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.467336183937916e-06,
|
|
"loss": 0.8527,
|
|
"step": 4550
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.46324546155525e-06,
|
|
"loss": 0.8373,
|
|
"step": 4551
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.459155192818986e-06,
|
|
"loss": 0.8486,
|
|
"step": 4552
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.45506537846058e-06,
|
|
"loss": 0.7816,
|
|
"step": 4553
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.450976019211416e-06,
|
|
"loss": 0.7677,
|
|
"step": 4554
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.446887115802785e-06,
|
|
"loss": 0.829,
|
|
"step": 4555
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.442798668965909e-06,
|
|
"loss": 0.7421,
|
|
"step": 4556
|
|
},
|
|
{
|
|
"epoch": 1.78,
|
|
"learning_rate": 7.43871067943192e-06,
|
|
"loss": 0.7919,
|
|
"step": 4557
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.434623147931873e-06,
|
|
"loss": 0.7897,
|
|
"step": 4558
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.43053607519674e-06,
|
|
"loss": 0.7702,
|
|
"step": 4559
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.426449461957406e-06,
|
|
"loss": 0.854,
|
|
"step": 4560
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.422363308944678e-06,
|
|
"loss": 0.8351,
|
|
"step": 4561
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.418277616889282e-06,
|
|
"loss": 0.8261,
|
|
"step": 4562
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.414192386521863e-06,
|
|
"loss": 0.8362,
|
|
"step": 4563
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.410107618572977e-06,
|
|
"loss": 0.6767,
|
|
"step": 4564
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.406023313773097e-06,
|
|
"loss": 0.752,
|
|
"step": 4565
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.401939472852624e-06,
|
|
"loss": 0.8748,
|
|
"step": 4566
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.397856096541869e-06,
|
|
"loss": 0.8281,
|
|
"step": 4567
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.393773185571053e-06,
|
|
"loss": 0.7817,
|
|
"step": 4568
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.389690740670325e-06,
|
|
"loss": 0.8258,
|
|
"step": 4569
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.3856087625697485e-06,
|
|
"loss": 0.8345,
|
|
"step": 4570
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.381527251999295e-06,
|
|
"loss": 0.7925,
|
|
"step": 4571
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.3774462096888674e-06,
|
|
"loss": 0.8001,
|
|
"step": 4572
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.37336563636827e-06,
|
|
"loss": 0.7701,
|
|
"step": 4573
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.369285532767232e-06,
|
|
"loss": 0.7681,
|
|
"step": 4574
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.365205899615392e-06,
|
|
"loss": 0.7243,
|
|
"step": 4575
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.361126737642316e-06,
|
|
"loss": 0.8127,
|
|
"step": 4576
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.357048047577471e-06,
|
|
"loss": 0.7223,
|
|
"step": 4577
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.352969830150251e-06,
|
|
"loss": 0.808,
|
|
"step": 4578
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.348892086089956e-06,
|
|
"loss": 0.7012,
|
|
"step": 4579
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.344814816125817e-06,
|
|
"loss": 0.7818,
|
|
"step": 4580
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.340738020986961e-06,
|
|
"loss": 0.8047,
|
|
"step": 4581
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.336661701402439e-06,
|
|
"loss": 0.82,
|
|
"step": 4582
|
|
},
|
|
{
|
|
"epoch": 1.79,
|
|
"learning_rate": 7.332585858101222e-06,
|
|
"loss": 0.8462,
|
|
"step": 4583
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.32851049181219e-06,
|
|
"loss": 0.8068,
|
|
"step": 4584
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.324435603264137e-06,
|
|
"loss": 0.8274,
|
|
"step": 4585
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.32036119318577e-06,
|
|
"loss": 0.8281,
|
|
"step": 4586
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.316287262305718e-06,
|
|
"loss": 0.8257,
|
|
"step": 4587
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.312213811352515e-06,
|
|
"loss": 0.9269,
|
|
"step": 4588
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.308140841054625e-06,
|
|
"loss": 0.8206,
|
|
"step": 4589
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.304068352140401e-06,
|
|
"loss": 0.8252,
|
|
"step": 4590
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.2999963453381315e-06,
|
|
"loss": 0.8163,
|
|
"step": 4591
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.295924821376009e-06,
|
|
"loss": 0.7753,
|
|
"step": 4592
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.2918537809821475e-06,
|
|
"loss": 0.8111,
|
|
"step": 4593
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.287783224884559e-06,
|
|
"loss": 0.8142,
|
|
"step": 4594
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.283713153811185e-06,
|
|
"loss": 0.8054,
|
|
"step": 4595
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.2796435684898716e-06,
|
|
"loss": 0.7408,
|
|
"step": 4596
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.275574469648383e-06,
|
|
"loss": 0.8178,
|
|
"step": 4597
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.271505858014396e-06,
|
|
"loss": 0.7742,
|
|
"step": 4598
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.267437734315493e-06,
|
|
"loss": 0.7432,
|
|
"step": 4599
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.263370099279173e-06,
|
|
"loss": 0.811,
|
|
"step": 4600
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.2593029536328515e-06,
|
|
"loss": 0.8287,
|
|
"step": 4601
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.25523629810386e-06,
|
|
"loss": 0.7985,
|
|
"step": 4602
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.251170133419427e-06,
|
|
"loss": 0.8007,
|
|
"step": 4603
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.247104460306706e-06,
|
|
"loss": 0.8117,
|
|
"step": 4604
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.243039279492759e-06,
|
|
"loss": 0.7896,
|
|
"step": 4605
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.238974591704563e-06,
|
|
"loss": 0.8128,
|
|
"step": 4606
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.234910397668997e-06,
|
|
"loss": 0.811,
|
|
"step": 4607
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.230846698112865e-06,
|
|
"loss": 0.8236,
|
|
"step": 4608
|
|
},
|
|
{
|
|
"epoch": 1.8,
|
|
"learning_rate": 7.226783493762872e-06,
|
|
"loss": 0.8487,
|
|
"step": 4609
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.22272078534564e-06,
|
|
"loss": 0.7673,
|
|
"step": 4610
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.218658573587705e-06,
|
|
"loss": 0.7854,
|
|
"step": 4611
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.214596859215504e-06,
|
|
"loss": 0.7395,
|
|
"step": 4612
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.210535642955391e-06,
|
|
"loss": 0.767,
|
|
"step": 4613
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.206474925533635e-06,
|
|
"loss": 0.809,
|
|
"step": 4614
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.2024147076764125e-06,
|
|
"loss": 0.8517,
|
|
"step": 4615
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.198354990109806e-06,
|
|
"loss": 0.8012,
|
|
"step": 4616
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.194295773559811e-06,
|
|
"loss": 0.7701,
|
|
"step": 4617
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.190237058752343e-06,
|
|
"loss": 0.7963,
|
|
"step": 4618
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.1861788464132145e-06,
|
|
"loss": 0.8114,
|
|
"step": 4619
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.182121137268152e-06,
|
|
"loss": 0.7424,
|
|
"step": 4620
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.178063932042794e-06,
|
|
"loss": 0.7995,
|
|
"step": 4621
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.174007231462691e-06,
|
|
"loss": 0.776,
|
|
"step": 4622
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.169951036253296e-06,
|
|
"loss": 0.8027,
|
|
"step": 4623
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.165895347139986e-06,
|
|
"loss": 0.8226,
|
|
"step": 4624
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.161840164848024e-06,
|
|
"loss": 0.7518,
|
|
"step": 4625
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.1577854901026045e-06,
|
|
"loss": 0.8589,
|
|
"step": 4626
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.153731323628819e-06,
|
|
"loss": 0.9094,
|
|
"step": 4627
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.1496776661516795e-06,
|
|
"loss": 0.7362,
|
|
"step": 4628
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.145624518396086e-06,
|
|
"loss": 0.7758,
|
|
"step": 4629
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.141571881086873e-06,
|
|
"loss": 0.8445,
|
|
"step": 4630
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.137519754948762e-06,
|
|
"loss": 0.7812,
|
|
"step": 4631
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.133468140706401e-06,
|
|
"loss": 0.8051,
|
|
"step": 4632
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.1294170390843335e-06,
|
|
"loss": 0.7733,
|
|
"step": 4633
|
|
},
|
|
{
|
|
"epoch": 1.81,
|
|
"learning_rate": 7.1253664508070145e-06,
|
|
"loss": 0.7769,
|
|
"step": 4634
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.121316376598809e-06,
|
|
"loss": 0.756,
|
|
"step": 4635
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.117266817183994e-06,
|
|
"loss": 0.8455,
|
|
"step": 4636
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.113217773286744e-06,
|
|
"loss": 0.814,
|
|
"step": 4637
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.109169245631149e-06,
|
|
"loss": 0.8263,
|
|
"step": 4638
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.105121234941207e-06,
|
|
"loss": 0.7974,
|
|
"step": 4639
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.101073741940821e-06,
|
|
"loss": 0.8534,
|
|
"step": 4640
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.097026767353801e-06,
|
|
"loss": 0.8164,
|
|
"step": 4641
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.092980311903863e-06,
|
|
"loss": 0.8016,
|
|
"step": 4642
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.088934376314636e-06,
|
|
"loss": 0.7937,
|
|
"step": 4643
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.0848889613096505e-06,
|
|
"loss": 0.7363,
|
|
"step": 4644
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.080844067612349e-06,
|
|
"loss": 0.7903,
|
|
"step": 4645
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.0767996959460714e-06,
|
|
"loss": 0.7704,
|
|
"step": 4646
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.072755847034076e-06,
|
|
"loss": 0.77,
|
|
"step": 4647
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.068712521599517e-06,
|
|
"loss": 0.7913,
|
|
"step": 4648
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.06466972036547e-06,
|
|
"loss": 0.778,
|
|
"step": 4649
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.0606274440548935e-06,
|
|
"loss": 0.8247,
|
|
"step": 4650
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.056585693390674e-06,
|
|
"loss": 0.7747,
|
|
"step": 4651
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.052544469095593e-06,
|
|
"loss": 0.8213,
|
|
"step": 4652
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.048503771892341e-06,
|
|
"loss": 0.8056,
|
|
"step": 4653
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.044463602503515e-06,
|
|
"loss": 0.8798,
|
|
"step": 4654
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.040423961651614e-06,
|
|
"loss": 0.8628,
|
|
"step": 4655
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.036384850059044e-06,
|
|
"loss": 0.8331,
|
|
"step": 4656
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.032346268448118e-06,
|
|
"loss": 0.7527,
|
|
"step": 4657
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.028308217541058e-06,
|
|
"loss": 0.7302,
|
|
"step": 4658
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.024270698059981e-06,
|
|
"loss": 0.7635,
|
|
"step": 4659
|
|
},
|
|
{
|
|
"epoch": 1.82,
|
|
"learning_rate": 7.020233710726912e-06,
|
|
"loss": 0.8015,
|
|
"step": 4660
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 7.016197256263788e-06,
|
|
"loss": 0.8302,
|
|
"step": 4661
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 7.0121613353924486e-06,
|
|
"loss": 0.7948,
|
|
"step": 4662
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 7.008125948834627e-06,
|
|
"loss": 0.7513,
|
|
"step": 4663
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 7.004091097311974e-06,
|
|
"loss": 0.7906,
|
|
"step": 4664
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 7.0000567815460406e-06,
|
|
"loss": 0.7555,
|
|
"step": 4665
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.996023002258278e-06,
|
|
"loss": 0.7218,
|
|
"step": 4666
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.991989760170051e-06,
|
|
"loss": 0.7666,
|
|
"step": 4667
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.987957056002615e-06,
|
|
"loss": 0.7283,
|
|
"step": 4668
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.983924890477138e-06,
|
|
"loss": 0.7912,
|
|
"step": 4669
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.97989326431469e-06,
|
|
"loss": 0.805,
|
|
"step": 4670
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.97586217823625e-06,
|
|
"loss": 0.7239,
|
|
"step": 4671
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.971831632962687e-06,
|
|
"loss": 0.8024,
|
|
"step": 4672
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.967801629214783e-06,
|
|
"loss": 0.8232,
|
|
"step": 4673
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.963772167713223e-06,
|
|
"loss": 0.8256,
|
|
"step": 4674
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.959743249178596e-06,
|
|
"loss": 0.7758,
|
|
"step": 4675
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.955714874331388e-06,
|
|
"loss": 0.7955,
|
|
"step": 4676
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.951687043891988e-06,
|
|
"loss": 0.7619,
|
|
"step": 4677
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.947659758580696e-06,
|
|
"loss": 0.7267,
|
|
"step": 4678
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.9436330191177074e-06,
|
|
"loss": 0.7984,
|
|
"step": 4679
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.939606826223123e-06,
|
|
"loss": 0.7756,
|
|
"step": 4680
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.935581180616939e-06,
|
|
"loss": 0.7675,
|
|
"step": 4681
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.931556083019067e-06,
|
|
"loss": 0.8268,
|
|
"step": 4682
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.927531534149305e-06,
|
|
"loss": 0.7476,
|
|
"step": 4683
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.923507534727374e-06,
|
|
"loss": 0.8908,
|
|
"step": 4684
|
|
},
|
|
{
|
|
"epoch": 1.83,
|
|
"learning_rate": 6.919484085472867e-06,
|
|
"loss": 0.7513,
|
|
"step": 4685
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.915461187105306e-06,
|
|
"loss": 0.8708,
|
|
"step": 4686
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.911438840344098e-06,
|
|
"loss": 0.8773,
|
|
"step": 4687
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.907417045908564e-06,
|
|
"loss": 0.7725,
|
|
"step": 4688
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.9033958045179135e-06,
|
|
"loss": 0.7354,
|
|
"step": 4689
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.8993751168912645e-06,
|
|
"loss": 0.8001,
|
|
"step": 4690
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.895354983747632e-06,
|
|
"loss": 0.8653,
|
|
"step": 4691
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.8913354058059374e-06,
|
|
"loss": 0.8332,
|
|
"step": 4692
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.887316383785002e-06,
|
|
"loss": 0.8449,
|
|
"step": 4693
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.88329791840354e-06,
|
|
"loss": 0.7655,
|
|
"step": 4694
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.879280010380169e-06,
|
|
"loss": 0.7991,
|
|
"step": 4695
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.875262660433417e-06,
|
|
"loss": 0.8617,
|
|
"step": 4696
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.871245869281703e-06,
|
|
"loss": 0.7548,
|
|
"step": 4697
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.867229637643341e-06,
|
|
"loss": 0.7957,
|
|
"step": 4698
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.863213966236558e-06,
|
|
"loss": 0.8336,
|
|
"step": 4699
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.859198855779472e-06,
|
|
"loss": 0.7826,
|
|
"step": 4700
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.855184306990106e-06,
|
|
"loss": 0.7926,
|
|
"step": 4701
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.851170320586371e-06,
|
|
"loss": 0.7875,
|
|
"step": 4702
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.847156897286094e-06,
|
|
"loss": 0.7689,
|
|
"step": 4703
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.8431440378069905e-06,
|
|
"loss": 0.7545,
|
|
"step": 4704
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.839131742866677e-06,
|
|
"loss": 0.7653,
|
|
"step": 4705
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.835120013182674e-06,
|
|
"loss": 0.8256,
|
|
"step": 4706
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.8311088494723944e-06,
|
|
"loss": 0.8513,
|
|
"step": 4707
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.8270982524531485e-06,
|
|
"loss": 0.8583,
|
|
"step": 4708
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.823088222842154e-06,
|
|
"loss": 0.7878,
|
|
"step": 4709
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.819078761356524e-06,
|
|
"loss": 0.8409,
|
|
"step": 4710
|
|
},
|
|
{
|
|
"epoch": 1.84,
|
|
"learning_rate": 6.815069868713263e-06,
|
|
"loss": 0.7974,
|
|
"step": 4711
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.811061545629278e-06,
|
|
"loss": 0.8686,
|
|
"step": 4712
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.807053792821383e-06,
|
|
"loss": 0.767,
|
|
"step": 4713
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.803046611006278e-06,
|
|
"loss": 0.8256,
|
|
"step": 4714
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.799040000900563e-06,
|
|
"loss": 0.7287,
|
|
"step": 4715
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.795033963220737e-06,
|
|
"loss": 0.8559,
|
|
"step": 4716
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.791028498683203e-06,
|
|
"loss": 0.7501,
|
|
"step": 4717
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.787023608004251e-06,
|
|
"loss": 0.796,
|
|
"step": 4718
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.783019291900079e-06,
|
|
"loss": 0.8388,
|
|
"step": 4719
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.779015551086769e-06,
|
|
"loss": 0.8112,
|
|
"step": 4720
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.775012386280314e-06,
|
|
"loss": 0.8223,
|
|
"step": 4721
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.771009798196592e-06,
|
|
"loss": 0.7891,
|
|
"step": 4722
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.767007787551394e-06,
|
|
"loss": 0.7866,
|
|
"step": 4723
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.763006355060386e-06,
|
|
"loss": 0.8312,
|
|
"step": 4724
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.759005501439146e-06,
|
|
"loss": 0.7749,
|
|
"step": 4725
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.755005227403144e-06,
|
|
"loss": 0.8028,
|
|
"step": 4726
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.7510055336677515e-06,
|
|
"loss": 0.8236,
|
|
"step": 4727
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.747006420948225e-06,
|
|
"loss": 0.7888,
|
|
"step": 4728
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.743007889959727e-06,
|
|
"loss": 0.8182,
|
|
"step": 4729
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.739009941417309e-06,
|
|
"loss": 0.8456,
|
|
"step": 4730
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.735012576035927e-06,
|
|
"loss": 0.8148,
|
|
"step": 4731
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.731015794530426e-06,
|
|
"loss": 0.8142,
|
|
"step": 4732
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.727019597615545e-06,
|
|
"loss": 0.742,
|
|
"step": 4733
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.723023986005925e-06,
|
|
"loss": 0.8123,
|
|
"step": 4734
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.7190289604160986e-06,
|
|
"loss": 0.7555,
|
|
"step": 4735
|
|
},
|
|
{
|
|
"epoch": 1.85,
|
|
"learning_rate": 6.715034521560495e-06,
|
|
"loss": 0.834,
|
|
"step": 4736
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.7110406701534325e-06,
|
|
"loss": 0.7574,
|
|
"step": 4737
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.707047406909135e-06,
|
|
"loss": 0.8212,
|
|
"step": 4738
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.703054732541713e-06,
|
|
"loss": 0.7944,
|
|
"step": 4739
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.699062647765176e-06,
|
|
"loss": 0.7676,
|
|
"step": 4740
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.695071153293422e-06,
|
|
"loss": 0.7846,
|
|
"step": 4741
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.691080249840252e-06,
|
|
"loss": 0.8015,
|
|
"step": 4742
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.687089938119353e-06,
|
|
"loss": 0.8388,
|
|
"step": 4743
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.683100218844316e-06,
|
|
"loss": 0.8143,
|
|
"step": 4744
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.679111092728619e-06,
|
|
"loss": 0.7898,
|
|
"step": 4745
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.67512256048563e-06,
|
|
"loss": 0.775,
|
|
"step": 4746
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.671134622828619e-06,
|
|
"loss": 0.81,
|
|
"step": 4747
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.6671472804707475e-06,
|
|
"loss": 0.7636,
|
|
"step": 4748
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.663160534125073e-06,
|
|
"loss": 0.8117,
|
|
"step": 4749
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.659174384504535e-06,
|
|
"loss": 0.7691,
|
|
"step": 4750
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.655188832321978e-06,
|
|
"loss": 0.8351,
|
|
"step": 4751
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.651203878290139e-06,
|
|
"loss": 0.7622,
|
|
"step": 4752
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.647219523121644e-06,
|
|
"loss": 0.7399,
|
|
"step": 4753
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.643235767529012e-06,
|
|
"loss": 0.7719,
|
|
"step": 4754
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.6392526122246515e-06,
|
|
"loss": 0.7667,
|
|
"step": 4755
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.635270057920875e-06,
|
|
"loss": 0.7327,
|
|
"step": 4756
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.631288105329879e-06,
|
|
"loss": 0.8302,
|
|
"step": 4757
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.627306755163749e-06,
|
|
"loss": 0.7957,
|
|
"step": 4758
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.623326008134474e-06,
|
|
"loss": 0.8346,
|
|
"step": 4759
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.619345864953926e-06,
|
|
"loss": 0.8548,
|
|
"step": 4760
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.615366326333868e-06,
|
|
"loss": 0.8074,
|
|
"step": 4761
|
|
},
|
|
{
|
|
"epoch": 1.86,
|
|
"learning_rate": 6.611387392985967e-06,
|
|
"loss": 0.7818,
|
|
"step": 4762
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.607409065621766e-06,
|
|
"loss": 0.7416,
|
|
"step": 4763
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.603431344952711e-06,
|
|
"loss": 0.7683,
|
|
"step": 4764
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.5994542316901326e-06,
|
|
"loss": 0.7754,
|
|
"step": 4765
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.595477726545261e-06,
|
|
"loss": 0.7888,
|
|
"step": 4766
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.591501830229207e-06,
|
|
"loss": 0.7702,
|
|
"step": 4767
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.587526543452977e-06,
|
|
"loss": 0.8792,
|
|
"step": 4768
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.583551866927475e-06,
|
|
"loss": 0.7792,
|
|
"step": 4769
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.5795778013634905e-06,
|
|
"loss": 0.8628,
|
|
"step": 4770
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.575604347471696e-06,
|
|
"loss": 0.861,
|
|
"step": 4771
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.571631505962666e-06,
|
|
"loss": 0.7772,
|
|
"step": 4772
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.5676592775468625e-06,
|
|
"loss": 0.8374,
|
|
"step": 4773
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.5636876629346355e-06,
|
|
"loss": 0.772,
|
|
"step": 4774
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.559716662836234e-06,
|
|
"loss": 0.771,
|
|
"step": 4775
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.555746277961776e-06,
|
|
"loss": 0.83,
|
|
"step": 4776
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.551776509021293e-06,
|
|
"loss": 0.8052,
|
|
"step": 4777
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.547807356724691e-06,
|
|
"loss": 0.7093,
|
|
"step": 4778
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.543838821781781e-06,
|
|
"loss": 0.7782,
|
|
"step": 4779
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.539870904902241e-06,
|
|
"loss": 0.7899,
|
|
"step": 4780
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.53590360679566e-06,
|
|
"loss": 0.8362,
|
|
"step": 4781
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.531936928171501e-06,
|
|
"loss": 0.8642,
|
|
"step": 4782
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.527970869739133e-06,
|
|
"loss": 0.7359,
|
|
"step": 4783
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.524005432207795e-06,
|
|
"loss": 0.7312,
|
|
"step": 4784
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.520040616286628e-06,
|
|
"loss": 0.7625,
|
|
"step": 4785
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.516076422684654e-06,
|
|
"loss": 0.8067,
|
|
"step": 4786
|
|
},
|
|
{
|
|
"epoch": 1.87,
|
|
"learning_rate": 6.512112852110793e-06,
|
|
"loss": 0.7673,
|
|
"step": 4787
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.508149905273849e-06,
|
|
"loss": 0.789,
|
|
"step": 4788
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.504187582882506e-06,
|
|
"loss": 0.8271,
|
|
"step": 4789
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.500225885645346e-06,
|
|
"loss": 0.9034,
|
|
"step": 4790
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.496264814270843e-06,
|
|
"loss": 0.7755,
|
|
"step": 4791
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.49230436946735e-06,
|
|
"loss": 0.8402,
|
|
"step": 4792
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.4883445519431065e-06,
|
|
"loss": 0.8427,
|
|
"step": 4793
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.48438536240625e-06,
|
|
"loss": 0.7678,
|
|
"step": 4794
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.480426801564799e-06,
|
|
"loss": 0.7338,
|
|
"step": 4795
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.476468870126663e-06,
|
|
"loss": 0.8222,
|
|
"step": 4796
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.472511568799629e-06,
|
|
"loss": 0.8254,
|
|
"step": 4797
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.468554898291386e-06,
|
|
"loss": 0.814,
|
|
"step": 4798
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.4645988593095e-06,
|
|
"loss": 0.826,
|
|
"step": 4799
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.4606434525614305e-06,
|
|
"loss": 0.727,
|
|
"step": 4800
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.456688678754519e-06,
|
|
"loss": 0.8113,
|
|
"step": 4801
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.452734538595995e-06,
|
|
"loss": 0.7286,
|
|
"step": 4802
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.4487810327929726e-06,
|
|
"loss": 0.7676,
|
|
"step": 4803
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.444828162052459e-06,
|
|
"loss": 0.7823,
|
|
"step": 4804
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.440875927081345e-06,
|
|
"loss": 0.7334,
|
|
"step": 4805
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.436924328586402e-06,
|
|
"loss": 0.8824,
|
|
"step": 4806
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.432973367274292e-06,
|
|
"loss": 0.8446,
|
|
"step": 4807
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.4290230438515676e-06,
|
|
"loss": 0.7044,
|
|
"step": 4808
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.425073359024664e-06,
|
|
"loss": 0.8653,
|
|
"step": 4809
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.421124313499895e-06,
|
|
"loss": 0.8232,
|
|
"step": 4810
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.417175907983467e-06,
|
|
"loss": 0.7693,
|
|
"step": 4811
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.413228143181476e-06,
|
|
"loss": 0.7455,
|
|
"step": 4812
|
|
},
|
|
{
|
|
"epoch": 1.88,
|
|
"learning_rate": 6.4092810197998935e-06,
|
|
"loss": 0.7853,
|
|
"step": 4813
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.40533453854459e-06,
|
|
"loss": 0.7947,
|
|
"step": 4814
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.4013887001213e-06,
|
|
"loss": 0.7999,
|
|
"step": 4815
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.397443505235663e-06,
|
|
"loss": 0.8505,
|
|
"step": 4816
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.393498954593193e-06,
|
|
"loss": 0.7934,
|
|
"step": 4817
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.389555048899299e-06,
|
|
"loss": 0.7376,
|
|
"step": 4818
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.38561178885926e-06,
|
|
"loss": 0.7463,
|
|
"step": 4819
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.381669175178249e-06,
|
|
"loss": 0.7675,
|
|
"step": 4820
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.37772720856132e-06,
|
|
"loss": 0.8412,
|
|
"step": 4821
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.3737858897134175e-06,
|
|
"loss": 0.7693,
|
|
"step": 4822
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.3698452193393614e-06,
|
|
"loss": 0.7976,
|
|
"step": 4823
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.365905198143856e-06,
|
|
"loss": 0.7769,
|
|
"step": 4824
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.361965826831501e-06,
|
|
"loss": 0.8078,
|
|
"step": 4825
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.3580271061067675e-06,
|
|
"loss": 0.7666,
|
|
"step": 4826
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.354089036674018e-06,
|
|
"loss": 0.7765,
|
|
"step": 4827
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.350151619237489e-06,
|
|
"loss": 0.7694,
|
|
"step": 4828
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.3462148545013124e-06,
|
|
"loss": 0.7885,
|
|
"step": 4829
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.3422787431694954e-06,
|
|
"loss": 0.8055,
|
|
"step": 4830
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.3383432859459335e-06,
|
|
"loss": 0.6949,
|
|
"step": 4831
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.334408483534396e-06,
|
|
"loss": 0.785,
|
|
"step": 4832
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.330474336638546e-06,
|
|
"loss": 0.8056,
|
|
"step": 4833
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.326540845961923e-06,
|
|
"loss": 0.7839,
|
|
"step": 4834
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.322608012207958e-06,
|
|
"loss": 0.7882,
|
|
"step": 4835
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.318675836079944e-06,
|
|
"loss": 0.7889,
|
|
"step": 4836
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.314744318281081e-06,
|
|
"loss": 0.8042,
|
|
"step": 4837
|
|
},
|
|
{
|
|
"epoch": 1.89,
|
|
"learning_rate": 6.310813459514433e-06,
|
|
"loss": 0.7756,
|
|
"step": 4838
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.30688326048296e-06,
|
|
"loss": 0.8112,
|
|
"step": 4839
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.302953721889495e-06,
|
|
"loss": 0.8132,
|
|
"step": 4840
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.299024844436752e-06,
|
|
"loss": 0.7917,
|
|
"step": 4841
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.295096628827331e-06,
|
|
"loss": 0.7682,
|
|
"step": 4842
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.291169075763716e-06,
|
|
"loss": 0.845,
|
|
"step": 4843
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.287242185948268e-06,
|
|
"loss": 0.7584,
|
|
"step": 4844
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.283315960083227e-06,
|
|
"loss": 0.8005,
|
|
"step": 4845
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.279390398870719e-06,
|
|
"loss": 0.8038,
|
|
"step": 4846
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.275465503012752e-06,
|
|
"loss": 0.8536,
|
|
"step": 4847
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.271541273211215e-06,
|
|
"loss": 0.7489,
|
|
"step": 4848
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.267617710167869e-06,
|
|
"loss": 0.7221,
|
|
"step": 4849
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.263694814584369e-06,
|
|
"loss": 0.858,
|
|
"step": 4850
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.25977258716224e-06,
|
|
"loss": 0.7845,
|
|
"step": 4851
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.255851028602893e-06,
|
|
"loss": 0.8199,
|
|
"step": 4852
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.251930139607621e-06,
|
|
"loss": 0.6913,
|
|
"step": 4853
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.248009920877591e-06,
|
|
"loss": 0.7807,
|
|
"step": 4854
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.244090373113854e-06,
|
|
"loss": 0.7901,
|
|
"step": 4855
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.24017149701734e-06,
|
|
"loss": 0.7799,
|
|
"step": 4856
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.236253293288863e-06,
|
|
"loss": 0.8101,
|
|
"step": 4857
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.2323357626291095e-06,
|
|
"loss": 0.8347,
|
|
"step": 4858
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.228418905738648e-06,
|
|
"loss": 0.7323,
|
|
"step": 4859
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.224502723317932e-06,
|
|
"loss": 0.7759,
|
|
"step": 4860
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.220587216067291e-06,
|
|
"loss": 0.8353,
|
|
"step": 4861
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.216672384686928e-06,
|
|
"loss": 0.8258,
|
|
"step": 4862
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.2127582298769316e-06,
|
|
"loss": 0.7657,
|
|
"step": 4863
|
|
},
|
|
{
|
|
"epoch": 1.9,
|
|
"learning_rate": 6.208844752337271e-06,
|
|
"loss": 0.7816,
|
|
"step": 4864
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.2049319527677875e-06,
|
|
"loss": 0.722,
|
|
"step": 4865
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.201019831868209e-06,
|
|
"loss": 0.7109,
|
|
"step": 4866
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.197108390338133e-06,
|
|
"loss": 0.8037,
|
|
"step": 4867
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.193197628877043e-06,
|
|
"loss": 0.7894,
|
|
"step": 4868
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.189287548184296e-06,
|
|
"loss": 0.8262,
|
|
"step": 4869
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.185378148959138e-06,
|
|
"loss": 0.7736,
|
|
"step": 4870
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.181469431900673e-06,
|
|
"loss": 0.8087,
|
|
"step": 4871
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.177561397707902e-06,
|
|
"loss": 0.7261,
|
|
"step": 4872
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.173654047079691e-06,
|
|
"loss": 0.8199,
|
|
"step": 4873
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.169747380714799e-06,
|
|
"loss": 0.7673,
|
|
"step": 4874
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.165841399311844e-06,
|
|
"loss": 0.7623,
|
|
"step": 4875
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.161936103569333e-06,
|
|
"loss": 0.8281,
|
|
"step": 4876
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.158031494185647e-06,
|
|
"loss": 0.7392,
|
|
"step": 4877
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.15412757185905e-06,
|
|
"loss": 0.8927,
|
|
"step": 4878
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.1502243372876715e-06,
|
|
"loss": 0.8232,
|
|
"step": 4879
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.146321791169529e-06,
|
|
"loss": 0.8091,
|
|
"step": 4880
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.142419934202509e-06,
|
|
"loss": 0.782,
|
|
"step": 4881
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.138518767084384e-06,
|
|
"loss": 0.7678,
|
|
"step": 4882
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.134618290512795e-06,
|
|
"loss": 0.7955,
|
|
"step": 4883
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.130718505185258e-06,
|
|
"loss": 0.842,
|
|
"step": 4884
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.126819411799175e-06,
|
|
"loss": 0.6896,
|
|
"step": 4885
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.122921011051817e-06,
|
|
"loss": 0.8344,
|
|
"step": 4886
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.119023303640335e-06,
|
|
"loss": 0.7583,
|
|
"step": 4887
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.115126290261746e-06,
|
|
"loss": 0.8207,
|
|
"step": 4888
|
|
},
|
|
{
|
|
"epoch": 1.91,
|
|
"learning_rate": 6.1112299716129575e-06,
|
|
"loss": 0.7733,
|
|
"step": 4889
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.107334348390745e-06,
|
|
"loss": 0.7314,
|
|
"step": 4890
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.1034394212917635e-06,
|
|
"loss": 0.828,
|
|
"step": 4891
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.099545191012532e-06,
|
|
"loss": 0.8336,
|
|
"step": 4892
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.095651658249461e-06,
|
|
"loss": 0.8425,
|
|
"step": 4893
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.091758823698824e-06,
|
|
"loss": 0.7787,
|
|
"step": 4894
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.087866688056781e-06,
|
|
"loss": 0.8204,
|
|
"step": 4895
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.0839752520193565e-06,
|
|
"loss": 0.7852,
|
|
"step": 4896
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.080084516282451e-06,
|
|
"loss": 0.7824,
|
|
"step": 4897
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.076194481541846e-06,
|
|
"loss": 0.8168,
|
|
"step": 4898
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.072305148493195e-06,
|
|
"loss": 0.7309,
|
|
"step": 4899
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.068416517832025e-06,
|
|
"loss": 0.8397,
|
|
"step": 4900
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.064528590253735e-06,
|
|
"loss": 0.7758,
|
|
"step": 4901
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.060641366453601e-06,
|
|
"loss": 0.804,
|
|
"step": 4902
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.056754847126777e-06,
|
|
"loss": 0.8161,
|
|
"step": 4903
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.052869032968285e-06,
|
|
"loss": 0.7505,
|
|
"step": 4904
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.048983924673022e-06,
|
|
"loss": 0.7359,
|
|
"step": 4905
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.045099522935758e-06,
|
|
"loss": 0.7285,
|
|
"step": 4906
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.0412158284511426e-06,
|
|
"loss": 0.8514,
|
|
"step": 4907
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.03733284191369e-06,
|
|
"loss": 0.7421,
|
|
"step": 4908
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.033450564017801e-06,
|
|
"loss": 0.8096,
|
|
"step": 4909
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.029568995457733e-06,
|
|
"loss": 0.7551,
|
|
"step": 4910
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.025688136927628e-06,
|
|
"loss": 0.7807,
|
|
"step": 4911
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.021807989121493e-06,
|
|
"loss": 0.8045,
|
|
"step": 4912
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.017928552733223e-06,
|
|
"loss": 0.8517,
|
|
"step": 4913
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.0140498284565654e-06,
|
|
"loss": 0.7853,
|
|
"step": 4914
|
|
},
|
|
{
|
|
"epoch": 1.92,
|
|
"learning_rate": 6.010171816985155e-06,
|
|
"loss": 0.8079,
|
|
"step": 4915
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 6.006294519012493e-06,
|
|
"loss": 0.7745,
|
|
"step": 4916
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 6.0024179352319564e-06,
|
|
"loss": 0.7783,
|
|
"step": 4917
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.99854206633679e-06,
|
|
"loss": 0.7503,
|
|
"step": 4918
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.994666913020113e-06,
|
|
"loss": 0.8479,
|
|
"step": 4919
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.99079247597492e-06,
|
|
"loss": 0.7372,
|
|
"step": 4920
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.9869187558940735e-06,
|
|
"loss": 0.7441,
|
|
"step": 4921
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.983045753470308e-06,
|
|
"loss": 0.7615,
|
|
"step": 4922
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.979173469396226e-06,
|
|
"loss": 0.7739,
|
|
"step": 4923
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.975301904364314e-06,
|
|
"loss": 0.8736,
|
|
"step": 4924
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.971431059066919e-06,
|
|
"loss": 0.8365,
|
|
"step": 4925
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.967560934196261e-06,
|
|
"loss": 0.8137,
|
|
"step": 4926
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.963691530444429e-06,
|
|
"loss": 0.8111,
|
|
"step": 4927
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.959822848503392e-06,
|
|
"loss": 0.7976,
|
|
"step": 4928
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.955954889064979e-06,
|
|
"loss": 0.7708,
|
|
"step": 4929
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.9520876528209035e-06,
|
|
"loss": 0.7595,
|
|
"step": 4930
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.948221140462731e-06,
|
|
"loss": 0.8187,
|
|
"step": 4931
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.944355352681914e-06,
|
|
"loss": 0.7908,
|
|
"step": 4932
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.940490290169765e-06,
|
|
"loss": 0.7707,
|
|
"step": 4933
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.9366259536174765e-06,
|
|
"loss": 0.8231,
|
|
"step": 4934
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.932762343716103e-06,
|
|
"loss": 0.8173,
|
|
"step": 4935
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.928899461156571e-06,
|
|
"loss": 0.8198,
|
|
"step": 4936
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.9250373066296754e-06,
|
|
"loss": 0.8562,
|
|
"step": 4937
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.9211758808260874e-06,
|
|
"loss": 0.7919,
|
|
"step": 4938
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.917315184436345e-06,
|
|
"loss": 0.7611,
|
|
"step": 4939
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.913455218150848e-06,
|
|
"loss": 0.8262,
|
|
"step": 4940
|
|
},
|
|
{
|
|
"epoch": 1.93,
|
|
"learning_rate": 5.909595982659874e-06,
|
|
"loss": 0.7779,
|
|
"step": 4941
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.905737478653573e-06,
|
|
"loss": 0.7398,
|
|
"step": 4942
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.901879706821956e-06,
|
|
"loss": 0.8503,
|
|
"step": 4943
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.898022667854901e-06,
|
|
"loss": 0.8876,
|
|
"step": 4944
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.894166362442167e-06,
|
|
"loss": 0.8497,
|
|
"step": 4945
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.890310791273372e-06,
|
|
"loss": 0.7631,
|
|
"step": 4946
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.886455955038006e-06,
|
|
"loss": 0.827,
|
|
"step": 4947
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.882601854425431e-06,
|
|
"loss": 0.8012,
|
|
"step": 4948
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.878748490124868e-06,
|
|
"loss": 0.7443,
|
|
"step": 4949
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.874895862825415e-06,
|
|
"loss": 0.7225,
|
|
"step": 4950
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.871043973216031e-06,
|
|
"loss": 0.8621,
|
|
"step": 4951
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.8671928219855555e-06,
|
|
"loss": 0.7449,
|
|
"step": 4952
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.863342409822681e-06,
|
|
"loss": 0.8625,
|
|
"step": 4953
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.859492737415973e-06,
|
|
"loss": 0.7222,
|
|
"step": 4954
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.855643805453875e-06,
|
|
"loss": 0.765,
|
|
"step": 4955
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.8517956146246826e-06,
|
|
"loss": 0.8052,
|
|
"step": 4956
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.847948165616566e-06,
|
|
"loss": 0.7446,
|
|
"step": 4957
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.844101459117563e-06,
|
|
"loss": 0.8422,
|
|
"step": 4958
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.840255495815579e-06,
|
|
"loss": 0.7579,
|
|
"step": 4959
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.8364102763983835e-06,
|
|
"loss": 0.8329,
|
|
"step": 4960
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.8325658015536205e-06,
|
|
"loss": 0.7683,
|
|
"step": 4961
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.82872207196879e-06,
|
|
"loss": 0.8429,
|
|
"step": 4962
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.824879088331261e-06,
|
|
"loss": 0.8493,
|
|
"step": 4963
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.8210368513282776e-06,
|
|
"loss": 0.8238,
|
|
"step": 4964
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.817195361646948e-06,
|
|
"loss": 0.7443,
|
|
"step": 4965
|
|
},
|
|
{
|
|
"epoch": 1.94,
|
|
"learning_rate": 5.813354619974239e-06,
|
|
"loss": 0.7998,
|
|
"step": 4966
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.809514626996983e-06,
|
|
"loss": 0.7603,
|
|
"step": 4967
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.8056753834018895e-06,
|
|
"loss": 0.8009,
|
|
"step": 4968
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.801836889875534e-06,
|
|
"loss": 0.8178,
|
|
"step": 4969
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.797999147104345e-06,
|
|
"loss": 0.8136,
|
|
"step": 4970
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.7941621557746205e-06,
|
|
"loss": 0.8373,
|
|
"step": 4971
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.790325916572533e-06,
|
|
"loss": 0.8078,
|
|
"step": 4972
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.786490430184115e-06,
|
|
"loss": 0.7604,
|
|
"step": 4973
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.782655697295267e-06,
|
|
"loss": 0.8308,
|
|
"step": 4974
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.778821718591747e-06,
|
|
"loss": 0.7827,
|
|
"step": 4975
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.7749884947591835e-06,
|
|
"loss": 0.841,
|
|
"step": 4976
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.77115602648307e-06,
|
|
"loss": 0.7569,
|
|
"step": 4977
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.7673243144487705e-06,
|
|
"loss": 0.7948,
|
|
"step": 4978
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.7634933593415035e-06,
|
|
"loss": 0.8306,
|
|
"step": 4979
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.759663161846352e-06,
|
|
"loss": 0.7704,
|
|
"step": 4980
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.755833722648273e-06,
|
|
"loss": 0.7543,
|
|
"step": 4981
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.752005042432086e-06,
|
|
"loss": 0.806,
|
|
"step": 4982
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.748177121882469e-06,
|
|
"loss": 0.7976,
|
|
"step": 4983
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.744349961683962e-06,
|
|
"loss": 0.8059,
|
|
"step": 4984
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.740523562520978e-06,
|
|
"loss": 0.7491,
|
|
"step": 4985
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.736697925077793e-06,
|
|
"loss": 0.7594,
|
|
"step": 4986
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.732873050038541e-06,
|
|
"loss": 0.8689,
|
|
"step": 4987
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.729048938087218e-06,
|
|
"loss": 0.7958,
|
|
"step": 4988
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.725225589907693e-06,
|
|
"loss": 0.8132,
|
|
"step": 4989
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.72140300618369e-06,
|
|
"loss": 0.7773,
|
|
"step": 4990
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.71758118759881e-06,
|
|
"loss": 0.7649,
|
|
"step": 4991
|
|
},
|
|
{
|
|
"epoch": 1.95,
|
|
"learning_rate": 5.71376013483649e-06,
|
|
"loss": 0.8498,
|
|
"step": 4992
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.709939848580054e-06,
|
|
"loss": 0.7895,
|
|
"step": 4993
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.7061203295126835e-06,
|
|
"loss": 0.8363,
|
|
"step": 4994
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.702301578317427e-06,
|
|
"loss": 0.8062,
|
|
"step": 4995
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.698483595677175e-06,
|
|
"loss": 0.7522,
|
|
"step": 4996
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.694666382274703e-06,
|
|
"loss": 0.7424,
|
|
"step": 4997
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.690849938792641e-06,
|
|
"loss": 0.8041,
|
|
"step": 4998
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.687034265913484e-06,
|
|
"loss": 0.8205,
|
|
"step": 4999
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.683219364319585e-06,
|
|
"loss": 0.806,
|
|
"step": 5000
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.679405234693156e-06,
|
|
"loss": 0.7908,
|
|
"step": 5001
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.67559187771628e-06,
|
|
"loss": 0.7714,
|
|
"step": 5002
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.671779294070899e-06,
|
|
"loss": 0.7087,
|
|
"step": 5003
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.667967484438815e-06,
|
|
"loss": 0.8047,
|
|
"step": 5004
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.664156449501685e-06,
|
|
"loss": 0.8066,
|
|
"step": 5005
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.660346189941039e-06,
|
|
"loss": 0.795,
|
|
"step": 5006
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.656536706438267e-06,
|
|
"loss": 0.7968,
|
|
"step": 5007
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.652727999674614e-06,
|
|
"loss": 0.661,
|
|
"step": 5008
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.648920070331185e-06,
|
|
"loss": 0.7442,
|
|
"step": 5009
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.645112919088954e-06,
|
|
"loss": 0.8375,
|
|
"step": 5010
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.641306546628755e-06,
|
|
"loss": 0.8,
|
|
"step": 5011
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.637500953631274e-06,
|
|
"loss": 0.7856,
|
|
"step": 5012
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.633696140777062e-06,
|
|
"loss": 0.7929,
|
|
"step": 5013
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.629892108746533e-06,
|
|
"loss": 0.7627,
|
|
"step": 5014
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.626088858219962e-06,
|
|
"loss": 0.846,
|
|
"step": 5015
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.622286389877485e-06,
|
|
"loss": 0.8217,
|
|
"step": 5016
|
|
},
|
|
{
|
|
"epoch": 1.96,
|
|
"learning_rate": 5.618484704399092e-06,
|
|
"loss": 0.7997,
|
|
"step": 5017
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.614683802464631e-06,
|
|
"loss": 0.8144,
|
|
"step": 5018
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.610883684753821e-06,
|
|
"loss": 0.7791,
|
|
"step": 5019
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.607084351946237e-06,
|
|
"loss": 0.7895,
|
|
"step": 5020
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.603285804721309e-06,
|
|
"loss": 0.7209,
|
|
"step": 5021
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.599488043758324e-06,
|
|
"loss": 0.8048,
|
|
"step": 5022
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.595691069736439e-06,
|
|
"loss": 0.8203,
|
|
"step": 5023
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.591894883334668e-06,
|
|
"loss": 0.84,
|
|
"step": 5024
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.588099485231876e-06,
|
|
"loss": 0.7666,
|
|
"step": 5025
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.584304876106788e-06,
|
|
"loss": 0.7305,
|
|
"step": 5026
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.580511056637998e-06,
|
|
"loss": 0.7644,
|
|
"step": 5027
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.576718027503955e-06,
|
|
"loss": 0.7791,
|
|
"step": 5028
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.572925789382956e-06,
|
|
"loss": 0.8279,
|
|
"step": 5029
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.569134342953173e-06,
|
|
"loss": 0.8456,
|
|
"step": 5030
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.56534368889262e-06,
|
|
"loss": 0.8529,
|
|
"step": 5031
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.561553827879187e-06,
|
|
"loss": 0.8176,
|
|
"step": 5032
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.557764760590603e-06,
|
|
"loss": 0.6798,
|
|
"step": 5033
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.5539764877044735e-06,
|
|
"loss": 0.815,
|
|
"step": 5034
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.5501890098982455e-06,
|
|
"loss": 0.8064,
|
|
"step": 5035
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.546402327849241e-06,
|
|
"loss": 0.7386,
|
|
"step": 5036
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.542616442234618e-06,
|
|
"loss": 0.7841,
|
|
"step": 5037
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.538831353731415e-06,
|
|
"loss": 0.7402,
|
|
"step": 5038
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.535047063016511e-06,
|
|
"loss": 0.8097,
|
|
"step": 5039
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.531263570766649e-06,
|
|
"loss": 0.7769,
|
|
"step": 5040
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.527480877658437e-06,
|
|
"loss": 0.8708,
|
|
"step": 5041
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.5236989843683194e-06,
|
|
"loss": 0.7829,
|
|
"step": 5042
|
|
},
|
|
{
|
|
"epoch": 1.97,
|
|
"learning_rate": 5.519917891572621e-06,
|
|
"loss": 0.7194,
|
|
"step": 5043
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.516137599947502e-06,
|
|
"loss": 0.7675,
|
|
"step": 5044
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.512358110169001e-06,
|
|
"loss": 0.7413,
|
|
"step": 5045
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.508579422912992e-06,
|
|
"loss": 0.8272,
|
|
"step": 5046
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.504801538855224e-06,
|
|
"loss": 0.7513,
|
|
"step": 5047
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.501024458671287e-06,
|
|
"loss": 0.8388,
|
|
"step": 5048
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.497248183036639e-06,
|
|
"loss": 0.7708,
|
|
"step": 5049
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.493472712626585e-06,
|
|
"loss": 0.8051,
|
|
"step": 5050
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.489698048116297e-06,
|
|
"loss": 0.7983,
|
|
"step": 5051
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.485924190180787e-06,
|
|
"loss": 0.7345,
|
|
"step": 5052
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.48215113949494e-06,
|
|
"loss": 0.8044,
|
|
"step": 5053
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.478378896733482e-06,
|
|
"loss": 0.7867,
|
|
"step": 5054
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.474607462571005e-06,
|
|
"loss": 0.8409,
|
|
"step": 5055
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.470836837681955e-06,
|
|
"loss": 0.8575,
|
|
"step": 5056
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.467067022740627e-06,
|
|
"loss": 0.7206,
|
|
"step": 5057
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.463298018421171e-06,
|
|
"loss": 0.8156,
|
|
"step": 5058
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.4595298253976024e-06,
|
|
"loss": 0.8457,
|
|
"step": 5059
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.455762444343785e-06,
|
|
"loss": 0.8073,
|
|
"step": 5060
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.451995875933437e-06,
|
|
"loss": 0.7688,
|
|
"step": 5061
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.448230120840126e-06,
|
|
"loss": 0.8192,
|
|
"step": 5062
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.444465179737284e-06,
|
|
"loss": 0.82,
|
|
"step": 5063
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.440701053298196e-06,
|
|
"loss": 0.7794,
|
|
"step": 5064
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.4369377421959936e-06,
|
|
"loss": 0.8855,
|
|
"step": 5065
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.433175247103674e-06,
|
|
"loss": 0.7634,
|
|
"step": 5066
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.429413568694073e-06,
|
|
"loss": 0.8555,
|
|
"step": 5067
|
|
},
|
|
{
|
|
"epoch": 1.98,
|
|
"learning_rate": 5.425652707639895e-06,
|
|
"loss": 0.8268,
|
|
"step": 5068
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.421892664613697e-06,
|
|
"loss": 0.7476,
|
|
"step": 5069
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.41813344028788e-06,
|
|
"loss": 0.8085,
|
|
"step": 5070
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.414375035334702e-06,
|
|
"loss": 0.8045,
|
|
"step": 5071
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.4106174504262785e-06,
|
|
"loss": 0.777,
|
|
"step": 5072
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.4068606862345804e-06,
|
|
"loss": 0.8093,
|
|
"step": 5073
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.403104743431424e-06,
|
|
"loss": 0.7301,
|
|
"step": 5074
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.399349622688479e-06,
|
|
"loss": 0.7532,
|
|
"step": 5075
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.3955953246772766e-06,
|
|
"loss": 0.8096,
|
|
"step": 5076
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.391841850069197e-06,
|
|
"loss": 0.8127,
|
|
"step": 5077
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.388089199535471e-06,
|
|
"loss": 0.8153,
|
|
"step": 5078
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.3843373737471764e-06,
|
|
"loss": 0.8272,
|
|
"step": 5079
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.380586373375255e-06,
|
|
"loss": 0.724,
|
|
"step": 5080
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.376836199090501e-06,
|
|
"loss": 0.8164,
|
|
"step": 5081
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.3730868515635515e-06,
|
|
"loss": 0.8233,
|
|
"step": 5082
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.369338331464895e-06,
|
|
"loss": 0.7906,
|
|
"step": 5083
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.365590639464883e-06,
|
|
"loss": 0.8222,
|
|
"step": 5084
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.361843776233713e-06,
|
|
"loss": 0.7725,
|
|
"step": 5085
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.358097742441441e-06,
|
|
"loss": 0.7931,
|
|
"step": 5086
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.354352538757953e-06,
|
|
"loss": 0.8087,
|
|
"step": 5087
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.35060816585301e-06,
|
|
"loss": 0.7935,
|
|
"step": 5088
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.346864624396215e-06,
|
|
"loss": 0.827,
|
|
"step": 5089
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.3431219150570304e-06,
|
|
"loss": 0.8031,
|
|
"step": 5090
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.339380038504755e-06,
|
|
"loss": 0.7749,
|
|
"step": 5091
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.3356389954085455e-06,
|
|
"loss": 0.7951,
|
|
"step": 5092
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.331898786437413e-06,
|
|
"loss": 0.7109,
|
|
"step": 5093
|
|
},
|
|
{
|
|
"epoch": 1.99,
|
|
"learning_rate": 5.32815941226022e-06,
|
|
"loss": 0.763,
|
|
"step": 5094
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.324420873545675e-06,
|
|
"loss": 0.7682,
|
|
"step": 5095
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.320683170962333e-06,
|
|
"loss": 0.7482,
|
|
"step": 5096
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.31694630517861e-06,
|
|
"loss": 0.817,
|
|
"step": 5097
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.313210276862772e-06,
|
|
"loss": 0.8083,
|
|
"step": 5098
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.309475086682927e-06,
|
|
"loss": 0.8289,
|
|
"step": 5099
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.3057407353070324e-06,
|
|
"loss": 0.7813,
|
|
"step": 5100
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.302007223402903e-06,
|
|
"loss": 0.8331,
|
|
"step": 5101
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.298274551638207e-06,
|
|
"loss": 0.8113,
|
|
"step": 5102
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.294542720680451e-06,
|
|
"loss": 0.8181,
|
|
"step": 5103
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.290811731196992e-06,
|
|
"loss": 0.7622,
|
|
"step": 5104
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.2870815838550446e-06,
|
|
"loss": 0.7897,
|
|
"step": 5105
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.283352279321673e-06,
|
|
"loss": 0.7408,
|
|
"step": 5106
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.27962381826378e-06,
|
|
"loss": 0.7182,
|
|
"step": 5107
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.2758962013481295e-06,
|
|
"loss": 0.5111,
|
|
"step": 5108
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.272169429241325e-06,
|
|
"loss": 0.5193,
|
|
"step": 5109
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.268443502609824e-06,
|
|
"loss": 0.5466,
|
|
"step": 5110
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.2647184221199355e-06,
|
|
"loss": 0.505,
|
|
"step": 5111
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.260994188437812e-06,
|
|
"loss": 0.531,
|
|
"step": 5112
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.2572708022294504e-06,
|
|
"loss": 0.4913,
|
|
"step": 5113
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.253548264160706e-06,
|
|
"loss": 0.4953,
|
|
"step": 5114
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.249826574897283e-06,
|
|
"loss": 0.4361,
|
|
"step": 5115
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.246105735104725e-06,
|
|
"loss": 0.4667,
|
|
"step": 5116
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.242385745448424e-06,
|
|
"loss": 0.5108,
|
|
"step": 5117
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.238666606593625e-06,
|
|
"loss": 0.4994,
|
|
"step": 5118
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"learning_rate": 5.234948319205428e-06,
|
|
"loss": 0.4598,
|
|
"step": 5119
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.231230883948759e-06,
|
|
"loss": 0.4491,
|
|
"step": 5120
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.227514301488417e-06,
|
|
"loss": 0.489,
|
|
"step": 5121
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.223798572489027e-06,
|
|
"loss": 0.5031,
|
|
"step": 5122
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.22008369761508e-06,
|
|
"loss": 0.4561,
|
|
"step": 5123
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.216369677530893e-06,
|
|
"loss": 0.5403,
|
|
"step": 5124
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.212656512900653e-06,
|
|
"loss": 0.4512,
|
|
"step": 5125
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.208944204388377e-06,
|
|
"loss": 0.4387,
|
|
"step": 5126
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.20523275265794e-06,
|
|
"loss": 0.4765,
|
|
"step": 5127
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.201522158373054e-06,
|
|
"loss": 0.4582,
|
|
"step": 5128
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.197812422197286e-06,
|
|
"loss": 0.4797,
|
|
"step": 5129
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.19410354479404e-06,
|
|
"loss": 0.4804,
|
|
"step": 5130
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.190395526826582e-06,
|
|
"loss": 0.4714,
|
|
"step": 5131
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.186688368958006e-06,
|
|
"loss": 0.4356,
|
|
"step": 5132
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.182982071851268e-06,
|
|
"loss": 0.5015,
|
|
"step": 5133
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.179276636169157e-06,
|
|
"loss": 0.5232,
|
|
"step": 5134
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.175572062574317e-06,
|
|
"loss": 0.4085,
|
|
"step": 5135
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.1718683517292366e-06,
|
|
"loss": 0.4037,
|
|
"step": 5136
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.168165504296243e-06,
|
|
"loss": 0.4315,
|
|
"step": 5137
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.164463520937522e-06,
|
|
"loss": 0.4262,
|
|
"step": 5138
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.160762402315089e-06,
|
|
"loss": 0.4473,
|
|
"step": 5139
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.157062149090821e-06,
|
|
"loss": 0.4405,
|
|
"step": 5140
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.153362761926424e-06,
|
|
"loss": 0.4638,
|
|
"step": 5141
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.1496642414834654e-06,
|
|
"loss": 0.4474,
|
|
"step": 5142
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.145966588423341e-06,
|
|
"loss": 0.5015,
|
|
"step": 5143
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.142269803407309e-06,
|
|
"loss": 0.4743,
|
|
"step": 5144
|
|
},
|
|
{
|
|
"epoch": 2.01,
|
|
"learning_rate": 5.138573887096456e-06,
|
|
"loss": 0.3949,
|
|
"step": 5145
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.1348788401517265e-06,
|
|
"loss": 0.4827,
|
|
"step": 5146
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.131184663233898e-06,
|
|
"loss": 0.4092,
|
|
"step": 5147
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.1274913570036045e-06,
|
|
"loss": 0.5547,
|
|
"step": 5148
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.1237989221213104e-06,
|
|
"loss": 0.413,
|
|
"step": 5149
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.120107359247335e-06,
|
|
"loss": 0.4962,
|
|
"step": 5150
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.1164166690418435e-06,
|
|
"loss": 0.4309,
|
|
"step": 5151
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.112726852164836e-06,
|
|
"loss": 0.4239,
|
|
"step": 5152
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.1090379092761555e-06,
|
|
"loss": 0.4869,
|
|
"step": 5153
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.105349841035498e-06,
|
|
"loss": 0.4024,
|
|
"step": 5154
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.1016626481024036e-06,
|
|
"loss": 0.4427,
|
|
"step": 5155
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.097976331136245e-06,
|
|
"loss": 0.4257,
|
|
"step": 5156
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.094290890796243e-06,
|
|
"loss": 0.5059,
|
|
"step": 5157
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.0906063277414645e-06,
|
|
"loss": 0.5064,
|
|
"step": 5158
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.086922642630823e-06,
|
|
"loss": 0.4431,
|
|
"step": 5159
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.0832398361230595e-06,
|
|
"loss": 0.4525,
|
|
"step": 5160
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.07955790887678e-06,
|
|
"loss": 0.5195,
|
|
"step": 5161
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.07587686155041e-06,
|
|
"loss": 0.4399,
|
|
"step": 5162
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.072196694802237e-06,
|
|
"loss": 0.4899,
|
|
"step": 5163
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.068517409290383e-06,
|
|
"loss": 0.3918,
|
|
"step": 5164
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.064839005672813e-06,
|
|
"loss": 0.4474,
|
|
"step": 5165
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.0611614846073256e-06,
|
|
"loss": 0.4334,
|
|
"step": 5166
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.057484846751575e-06,
|
|
"loss": 0.4939,
|
|
"step": 5167
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.053809092763058e-06,
|
|
"loss": 0.4388,
|
|
"step": 5168
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.050134223299102e-06,
|
|
"loss": 0.4111,
|
|
"step": 5169
|
|
},
|
|
{
|
|
"epoch": 2.02,
|
|
"learning_rate": 5.046460239016879e-06,
|
|
"loss": 0.4716,
|
|
"step": 5170
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.04278714057341e-06,
|
|
"loss": 0.4448,
|
|
"step": 5171
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.039114928625555e-06,
|
|
"loss": 0.5283,
|
|
"step": 5172
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.03544360383001e-06,
|
|
"loss": 0.4456,
|
|
"step": 5173
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.031773166843315e-06,
|
|
"loss": 0.4672,
|
|
"step": 5174
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.028103618321851e-06,
|
|
"loss": 0.4437,
|
|
"step": 5175
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.024434958921847e-06,
|
|
"loss": 0.4744,
|
|
"step": 5176
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.020767189299369e-06,
|
|
"loss": 0.4916,
|
|
"step": 5177
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.017100310110312e-06,
|
|
"loss": 0.4509,
|
|
"step": 5178
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.013434322010429e-06,
|
|
"loss": 0.4347,
|
|
"step": 5179
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.009769225655303e-06,
|
|
"loss": 0.4787,
|
|
"step": 5180
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.006105021700374e-06,
|
|
"loss": 0.4132,
|
|
"step": 5181
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 5.00244171080089e-06,
|
|
"loss": 0.4239,
|
|
"step": 5182
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.9987792936119665e-06,
|
|
"loss": 0.51,
|
|
"step": 5183
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.9951177707885535e-06,
|
|
"loss": 0.4733,
|
|
"step": 5184
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.991457142985442e-06,
|
|
"loss": 0.4481,
|
|
"step": 5185
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.987797410857257e-06,
|
|
"loss": 0.4507,
|
|
"step": 5186
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.984138575058462e-06,
|
|
"loss": 0.4476,
|
|
"step": 5187
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.980480636243369e-06,
|
|
"loss": 0.4804,
|
|
"step": 5188
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.976823595066128e-06,
|
|
"loss": 0.4307,
|
|
"step": 5189
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.9731674521807214e-06,
|
|
"loss": 0.4863,
|
|
"step": 5190
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.969512208240973e-06,
|
|
"loss": 0.4387,
|
|
"step": 5191
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.965857863900551e-06,
|
|
"loss": 0.4118,
|
|
"step": 5192
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.962204419812964e-06,
|
|
"loss": 0.4548,
|
|
"step": 5193
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.95855187663155e-06,
|
|
"loss": 0.5122,
|
|
"step": 5194
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.95490023500949e-06,
|
|
"loss": 0.4419,
|
|
"step": 5195
|
|
},
|
|
{
|
|
"epoch": 2.03,
|
|
"learning_rate": 4.951249495599806e-06,
|
|
"loss": 0.4867,
|
|
"step": 5196
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.947599659055364e-06,
|
|
"loss": 0.436,
|
|
"step": 5197
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.9439507260288565e-06,
|
|
"loss": 0.4641,
|
|
"step": 5198
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.940302697172818e-06,
|
|
"loss": 0.4629,
|
|
"step": 5199
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.936655573139627e-06,
|
|
"loss": 0.4957,
|
|
"step": 5200
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.933009354581495e-06,
|
|
"loss": 0.4482,
|
|
"step": 5201
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.929364042150478e-06,
|
|
"loss": 0.4329,
|
|
"step": 5202
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.9257196364984626e-06,
|
|
"loss": 0.4116,
|
|
"step": 5203
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.922076138277171e-06,
|
|
"loss": 0.4202,
|
|
"step": 5204
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.918433548138172e-06,
|
|
"loss": 0.4495,
|
|
"step": 5205
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.91479186673287e-06,
|
|
"loss": 0.5351,
|
|
"step": 5206
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.911151094712504e-06,
|
|
"loss": 0.4501,
|
|
"step": 5207
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.907511232728145e-06,
|
|
"loss": 0.521,
|
|
"step": 5208
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.903872281430712e-06,
|
|
"loss": 0.4576,
|
|
"step": 5209
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.900234241470962e-06,
|
|
"loss": 0.4269,
|
|
"step": 5210
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.896597113499479e-06,
|
|
"loss": 0.481,
|
|
"step": 5211
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.892960898166683e-06,
|
|
"loss": 0.4633,
|
|
"step": 5212
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.889325596122843e-06,
|
|
"loss": 0.4905,
|
|
"step": 5213
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.885691208018059e-06,
|
|
"loss": 0.3689,
|
|
"step": 5214
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.8820577345022605e-06,
|
|
"loss": 0.4609,
|
|
"step": 5215
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.878425176225227e-06,
|
|
"loss": 0.5063,
|
|
"step": 5216
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.874793533836561e-06,
|
|
"loss": 0.4254,
|
|
"step": 5217
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.871162807985711e-06,
|
|
"loss": 0.511,
|
|
"step": 5218
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.8675329993219525e-06,
|
|
"loss": 0.5012,
|
|
"step": 5219
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.863904108494409e-06,
|
|
"loss": 0.4785,
|
|
"step": 5220
|
|
},
|
|
{
|
|
"epoch": 2.04,
|
|
"learning_rate": 4.860276136152026e-06,
|
|
"loss": 0.461,
|
|
"step": 5221
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.856649082943598e-06,
|
|
"loss": 0.5164,
|
|
"step": 5222
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.853022949517745e-06,
|
|
"loss": 0.4523,
|
|
"step": 5223
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.849397736522929e-06,
|
|
"loss": 0.4615,
|
|
"step": 5224
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.845773444607441e-06,
|
|
"loss": 0.4654,
|
|
"step": 5225
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.8421500744194125e-06,
|
|
"loss": 0.5041,
|
|
"step": 5226
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.838527626606815e-06,
|
|
"loss": 0.4835,
|
|
"step": 5227
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.8349061018174385e-06,
|
|
"loss": 0.4304,
|
|
"step": 5228
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.831285500698928e-06,
|
|
"loss": 0.4928,
|
|
"step": 5229
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.827665823898745e-06,
|
|
"loss": 0.485,
|
|
"step": 5230
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.824047072064202e-06,
|
|
"loss": 0.425,
|
|
"step": 5231
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.82042924584243e-06,
|
|
"loss": 0.4164,
|
|
"step": 5232
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.8168123458804115e-06,
|
|
"loss": 0.4912,
|
|
"step": 5233
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.813196372824947e-06,
|
|
"loss": 0.4293,
|
|
"step": 5234
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.809581327322688e-06,
|
|
"loss": 0.4683,
|
|
"step": 5235
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.805967210020101e-06,
|
|
"loss": 0.4871,
|
|
"step": 5236
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.802354021563503e-06,
|
|
"loss": 0.4926,
|
|
"step": 5237
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.798741762599035e-06,
|
|
"loss": 0.4138,
|
|
"step": 5238
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.79513043377268e-06,
|
|
"loss": 0.4494,
|
|
"step": 5239
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.791520035730244e-06,
|
|
"loss": 0.4463,
|
|
"step": 5240
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.787910569117374e-06,
|
|
"loss": 0.4814,
|
|
"step": 5241
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.784302034579556e-06,
|
|
"loss": 0.4673,
|
|
"step": 5242
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.780694432762095e-06,
|
|
"loss": 0.4735,
|
|
"step": 5243
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.777087764310134e-06,
|
|
"loss": 0.5029,
|
|
"step": 5244
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.773482029868657e-06,
|
|
"loss": 0.4839,
|
|
"step": 5245
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.769877230082476e-06,
|
|
"loss": 0.4874,
|
|
"step": 5246
|
|
},
|
|
{
|
|
"epoch": 2.05,
|
|
"learning_rate": 4.7662733655962335e-06,
|
|
"loss": 0.5362,
|
|
"step": 5247
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.762670437054403e-06,
|
|
"loss": 0.4872,
|
|
"step": 5248
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.759068445101298e-06,
|
|
"loss": 0.3966,
|
|
"step": 5249
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.755467390381063e-06,
|
|
"loss": 0.4534,
|
|
"step": 5250
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.751867273537667e-06,
|
|
"loss": 0.394,
|
|
"step": 5251
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.748268095214923e-06,
|
|
"loss": 0.4652,
|
|
"step": 5252
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.744669856056463e-06,
|
|
"loss": 0.4758,
|
|
"step": 5253
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.741072556705766e-06,
|
|
"loss": 0.4985,
|
|
"step": 5254
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.737476197806127e-06,
|
|
"loss": 0.4395,
|
|
"step": 5255
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.7338807800006885e-06,
|
|
"loss": 0.4656,
|
|
"step": 5256
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.730286303932409e-06,
|
|
"loss": 0.4578,
|
|
"step": 5257
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.726692770244091e-06,
|
|
"loss": 0.3675,
|
|
"step": 5258
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.723100179578368e-06,
|
|
"loss": 0.4811,
|
|
"step": 5259
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.719508532577698e-06,
|
|
"loss": 0.4771,
|
|
"step": 5260
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.715917829884368e-06,
|
|
"loss": 0.488,
|
|
"step": 5261
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.712328072140505e-06,
|
|
"loss": 0.4768,
|
|
"step": 5262
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.708739259988069e-06,
|
|
"loss": 0.4455,
|
|
"step": 5263
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.70515139406884e-06,
|
|
"loss": 0.4528,
|
|
"step": 5264
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.70156447502443e-06,
|
|
"loss": 0.5027,
|
|
"step": 5265
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.697978503496291e-06,
|
|
"loss": 0.3646,
|
|
"step": 5266
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.694393480125703e-06,
|
|
"loss": 0.4391,
|
|
"step": 5267
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.69080940555377e-06,
|
|
"loss": 0.4269,
|
|
"step": 5268
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.687226280421426e-06,
|
|
"loss": 0.4196,
|
|
"step": 5269
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.683644105369443e-06,
|
|
"loss": 0.4406,
|
|
"step": 5270
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.680062881038422e-06,
|
|
"loss": 0.5042,
|
|
"step": 5271
|
|
},
|
|
{
|
|
"epoch": 2.06,
|
|
"learning_rate": 4.676482608068796e-06,
|
|
"loss": 0.4408,
|
|
"step": 5272
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.6729032871008074e-06,
|
|
"loss": 0.4746,
|
|
"step": 5273
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.669324918774555e-06,
|
|
"loss": 0.503,
|
|
"step": 5274
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.665747503729954e-06,
|
|
"loss": 0.4634,
|
|
"step": 5275
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.662171042606755e-06,
|
|
"loss": 0.4534,
|
|
"step": 5276
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.658595536044532e-06,
|
|
"loss": 0.4444,
|
|
"step": 5277
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.655020984682688e-06,
|
|
"loss": 0.4703,
|
|
"step": 5278
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.6514473891604584e-06,
|
|
"loss": 0.5277,
|
|
"step": 5279
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.647874750116915e-06,
|
|
"loss": 0.4498,
|
|
"step": 5280
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.644303068190945e-06,
|
|
"loss": 0.458,
|
|
"step": 5281
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.640732344021266e-06,
|
|
"loss": 0.4932,
|
|
"step": 5282
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.637162578246434e-06,
|
|
"loss": 0.4987,
|
|
"step": 5283
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.63359377150483e-06,
|
|
"loss": 0.4982,
|
|
"step": 5284
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.630025924434661e-06,
|
|
"loss": 0.4134,
|
|
"step": 5285
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.626459037673957e-06,
|
|
"loss": 0.5169,
|
|
"step": 5286
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.622893111860586e-06,
|
|
"loss": 0.5137,
|
|
"step": 5287
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.6193281476322446e-06,
|
|
"loss": 0.502,
|
|
"step": 5288
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.61576414562645e-06,
|
|
"loss": 0.4152,
|
|
"step": 5289
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.612201106480548e-06,
|
|
"loss": 0.4279,
|
|
"step": 5290
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.608639030831717e-06,
|
|
"loss": 0.4523,
|
|
"step": 5291
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.605077919316965e-06,
|
|
"loss": 0.4833,
|
|
"step": 5292
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.60151777257312e-06,
|
|
"loss": 0.4665,
|
|
"step": 5293
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.597958591236837e-06,
|
|
"loss": 0.4576,
|
|
"step": 5294
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.594400375944607e-06,
|
|
"loss": 0.453,
|
|
"step": 5295
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.590843127332744e-06,
|
|
"loss": 0.3957,
|
|
"step": 5296
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.58728684603739e-06,
|
|
"loss": 0.5423,
|
|
"step": 5297
|
|
},
|
|
{
|
|
"epoch": 2.07,
|
|
"learning_rate": 4.583731532694512e-06,
|
|
"loss": 0.4654,
|
|
"step": 5298
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.580177187939899e-06,
|
|
"loss": 0.436,
|
|
"step": 5299
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.576623812409178e-06,
|
|
"loss": 0.4808,
|
|
"step": 5300
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.573071406737799e-06,
|
|
"loss": 0.3849,
|
|
"step": 5301
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.569519971561035e-06,
|
|
"loss": 0.4354,
|
|
"step": 5302
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.565969507513981e-06,
|
|
"loss": 0.4862,
|
|
"step": 5303
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.56242001523157e-06,
|
|
"loss": 0.4158,
|
|
"step": 5304
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.558871495348559e-06,
|
|
"loss": 0.4601,
|
|
"step": 5305
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.555323948499525e-06,
|
|
"loss": 0.3906,
|
|
"step": 5306
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.551777375318868e-06,
|
|
"loss": 0.4452,
|
|
"step": 5307
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.548231776440826e-06,
|
|
"loss": 0.4573,
|
|
"step": 5308
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.544687152499457e-06,
|
|
"loss": 0.4669,
|
|
"step": 5309
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.54114350412864e-06,
|
|
"loss": 0.4485,
|
|
"step": 5310
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.537600831962089e-06,
|
|
"loss": 0.4323,
|
|
"step": 5311
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.534059136633331e-06,
|
|
"loss": 0.4522,
|
|
"step": 5312
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.530518418775734e-06,
|
|
"loss": 0.4865,
|
|
"step": 5313
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.526978679022474e-06,
|
|
"loss": 0.4472,
|
|
"step": 5314
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.523439918006567e-06,
|
|
"loss": 0.4687,
|
|
"step": 5315
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.5199021363608435e-06,
|
|
"loss": 0.4753,
|
|
"step": 5316
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.516365334717967e-06,
|
|
"loss": 0.4791,
|
|
"step": 5317
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.5128295137104176e-06,
|
|
"loss": 0.4871,
|
|
"step": 5318
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.509294673970508e-06,
|
|
"loss": 0.4463,
|
|
"step": 5319
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.505760816130368e-06,
|
|
"loss": 0.4502,
|
|
"step": 5320
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.502227940821956e-06,
|
|
"loss": 0.4738,
|
|
"step": 5321
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.4986960486770596e-06,
|
|
"loss": 0.5138,
|
|
"step": 5322
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.4951651403272765e-06,
|
|
"loss": 0.4674,
|
|
"step": 5323
|
|
},
|
|
{
|
|
"epoch": 2.08,
|
|
"learning_rate": 4.491635216404046e-06,
|
|
"loss": 0.4918,
|
|
"step": 5324
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.488106277538614e-06,
|
|
"loss": 0.4211,
|
|
"step": 5325
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.484578324362067e-06,
|
|
"loss": 0.3918,
|
|
"step": 5326
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.481051357505297e-06,
|
|
"loss": 0.4681,
|
|
"step": 5327
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.4775253775990394e-06,
|
|
"loss": 0.5068,
|
|
"step": 5328
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.4740003852738355e-06,
|
|
"loss": 0.5166,
|
|
"step": 5329
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.470476381160065e-06,
|
|
"loss": 0.3971,
|
|
"step": 5330
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.466953365887915e-06,
|
|
"loss": 0.4593,
|
|
"step": 5331
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.463431340087413e-06,
|
|
"loss": 0.4474,
|
|
"step": 5332
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.459910304388393e-06,
|
|
"loss": 0.4736,
|
|
"step": 5333
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.456390259420528e-06,
|
|
"loss": 0.4944,
|
|
"step": 5334
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.452871205813296e-06,
|
|
"loss": 0.4965,
|
|
"step": 5335
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.449353144196015e-06,
|
|
"loss": 0.5291,
|
|
"step": 5336
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.445836075197818e-06,
|
|
"loss": 0.4827,
|
|
"step": 5337
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.442319999447659e-06,
|
|
"loss": 0.469,
|
|
"step": 5338
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.438804917574311e-06,
|
|
"loss": 0.4241,
|
|
"step": 5339
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.4352908302063776e-06,
|
|
"loss": 0.4596,
|
|
"step": 5340
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.431777737972287e-06,
|
|
"loss": 0.4788,
|
|
"step": 5341
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.428265641500278e-06,
|
|
"loss": 0.4356,
|
|
"step": 5342
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.424754541418412e-06,
|
|
"loss": 0.4039,
|
|
"step": 5343
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.421244438354584e-06,
|
|
"loss": 0.4905,
|
|
"step": 5344
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.417735332936505e-06,
|
|
"loss": 0.4746,
|
|
"step": 5345
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.414227225791699e-06,
|
|
"loss": 0.4146,
|
|
"step": 5346
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.410720117547528e-06,
|
|
"loss": 0.4531,
|
|
"step": 5347
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.407214008831157e-06,
|
|
"loss": 0.4082,
|
|
"step": 5348
|
|
},
|
|
{
|
|
"epoch": 2.09,
|
|
"learning_rate": 4.4037089002695856e-06,
|
|
"loss": 0.4176,
|
|
"step": 5349
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.4002047924896344e-06,
|
|
"loss": 0.4531,
|
|
"step": 5350
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.396701686117937e-06,
|
|
"loss": 0.4995,
|
|
"step": 5351
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.393199581780949e-06,
|
|
"loss": 0.4806,
|
|
"step": 5352
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.389698480104953e-06,
|
|
"loss": 0.4346,
|
|
"step": 5353
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.386198381716053e-06,
|
|
"loss": 0.4524,
|
|
"step": 5354
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.382699287240166e-06,
|
|
"loss": 0.3894,
|
|
"step": 5355
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.379201197303029e-06,
|
|
"loss": 0.4243,
|
|
"step": 5356
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.375704112530207e-06,
|
|
"loss": 0.4371,
|
|
"step": 5357
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.372208033547086e-06,
|
|
"loss": 0.4945,
|
|
"step": 5358
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.368712960978863e-06,
|
|
"loss": 0.4396,
|
|
"step": 5359
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.365218895450558e-06,
|
|
"loss": 0.4537,
|
|
"step": 5360
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.3617258375870156e-06,
|
|
"loss": 0.5184,
|
|
"step": 5361
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.358233788012896e-06,
|
|
"loss": 0.4745,
|
|
"step": 5362
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.354742747352691e-06,
|
|
"loss": 0.4447,
|
|
"step": 5363
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.351252716230685e-06,
|
|
"loss": 0.4224,
|
|
"step": 5364
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.347763695271004e-06,
|
|
"loss": 0.4415,
|
|
"step": 5365
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.344275685097591e-06,
|
|
"loss": 0.4544,
|
|
"step": 5366
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.3407886863342085e-06,
|
|
"loss": 0.4789,
|
|
"step": 5367
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.337302699604423e-06,
|
|
"loss": 0.4394,
|
|
"step": 5368
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.333817725531638e-06,
|
|
"loss": 0.5032,
|
|
"step": 5369
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.33033376473907e-06,
|
|
"loss": 0.5006,
|
|
"step": 5370
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.326850817849755e-06,
|
|
"loss": 0.4898,
|
|
"step": 5371
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.3233688854865455e-06,
|
|
"loss": 0.4426,
|
|
"step": 5372
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.319887968272109e-06,
|
|
"loss": 0.4774,
|
|
"step": 5373
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.316408066828939e-06,
|
|
"loss": 0.4731,
|
|
"step": 5374
|
|
},
|
|
{
|
|
"epoch": 2.1,
|
|
"learning_rate": 4.312929181779349e-06,
|
|
"loss": 0.4626,
|
|
"step": 5375
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.309451313745461e-06,
|
|
"loss": 0.4383,
|
|
"step": 5376
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.3059744633492174e-06,
|
|
"loss": 0.4633,
|
|
"step": 5377
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.302498631212385e-06,
|
|
"loss": 0.5226,
|
|
"step": 5378
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.29902381795655e-06,
|
|
"loss": 0.5356,
|
|
"step": 5379
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.295550024203106e-06,
|
|
"loss": 0.4799,
|
|
"step": 5380
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.292077250573265e-06,
|
|
"loss": 0.4191,
|
|
"step": 5381
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.2886054976880676e-06,
|
|
"loss": 0.48,
|
|
"step": 5382
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.285134766168367e-06,
|
|
"loss": 0.4511,
|
|
"step": 5383
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.2816650566348295e-06,
|
|
"loss": 0.4603,
|
|
"step": 5384
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.278196369707936e-06,
|
|
"loss": 0.4726,
|
|
"step": 5385
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.2747287060079965e-06,
|
|
"loss": 0.4104,
|
|
"step": 5386
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.2712620661551285e-06,
|
|
"loss": 0.423,
|
|
"step": 5387
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.267796450769278e-06,
|
|
"loss": 0.4546,
|
|
"step": 5388
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.2643318604701845e-06,
|
|
"loss": 0.4702,
|
|
"step": 5389
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.260868295877426e-06,
|
|
"loss": 0.4758,
|
|
"step": 5390
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.257405757610389e-06,
|
|
"loss": 0.4803,
|
|
"step": 5391
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.2539442462882816e-06,
|
|
"loss": 0.4491,
|
|
"step": 5392
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.250483762530121e-06,
|
|
"loss": 0.4447,
|
|
"step": 5393
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.247024306954739e-06,
|
|
"loss": 0.4607,
|
|
"step": 5394
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.243565880180791e-06,
|
|
"loss": 0.4096,
|
|
"step": 5395
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.24010848282675e-06,
|
|
"loss": 0.4297,
|
|
"step": 5396
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.236652115510898e-06,
|
|
"loss": 0.4559,
|
|
"step": 5397
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.2331967788513295e-06,
|
|
"loss": 0.4058,
|
|
"step": 5398
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.229742473465966e-06,
|
|
"loss": 0.4443,
|
|
"step": 5399
|
|
},
|
|
{
|
|
"epoch": 2.11,
|
|
"learning_rate": 4.226289199972541e-06,
|
|
"loss": 0.3994,
|
|
"step": 5400
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.2228369589885985e-06,
|
|
"loss": 0.4462,
|
|
"step": 5401
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.219385751131497e-06,
|
|
"loss": 0.4967,
|
|
"step": 5402
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.215935577018418e-06,
|
|
"loss": 0.4954,
|
|
"step": 5403
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.2124864372663586e-06,
|
|
"loss": 0.4991,
|
|
"step": 5404
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.209038332492118e-06,
|
|
"loss": 0.4428,
|
|
"step": 5405
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.205591263312327e-06,
|
|
"loss": 0.4518,
|
|
"step": 5406
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.202145230343415e-06,
|
|
"loss": 0.4857,
|
|
"step": 5407
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.198700234201641e-06,
|
|
"loss": 0.4757,
|
|
"step": 5408
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.195256275503066e-06,
|
|
"loss": 0.4391,
|
|
"step": 5409
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.191813354863577e-06,
|
|
"loss": 0.4157,
|
|
"step": 5410
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.188371472898865e-06,
|
|
"loss": 0.4595,
|
|
"step": 5411
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.18493063022444e-06,
|
|
"loss": 0.4175,
|
|
"step": 5412
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.181490827455631e-06,
|
|
"loss": 0.4364,
|
|
"step": 5413
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.178052065207573e-06,
|
|
"loss": 0.4437,
|
|
"step": 5414
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.1746143440952135e-06,
|
|
"loss": 0.4759,
|
|
"step": 5415
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.1711776647333224e-06,
|
|
"loss": 0.469,
|
|
"step": 5416
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.1677420277364825e-06,
|
|
"loss": 0.4579,
|
|
"step": 5417
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.1643074337190796e-06,
|
|
"loss": 0.4248,
|
|
"step": 5418
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.160873883295328e-06,
|
|
"loss": 0.4807,
|
|
"step": 5419
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.157441377079239e-06,
|
|
"loss": 0.4243,
|
|
"step": 5420
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.154009915684653e-06,
|
|
"loss": 0.4678,
|
|
"step": 5421
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.150579499725213e-06,
|
|
"loss": 0.4523,
|
|
"step": 5422
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.14715012981438e-06,
|
|
"loss": 0.4998,
|
|
"step": 5423
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.143721806565423e-06,
|
|
"loss": 0.4446,
|
|
"step": 5424
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.140294530591433e-06,
|
|
"loss": 0.4787,
|
|
"step": 5425
|
|
},
|
|
{
|
|
"epoch": 2.12,
|
|
"learning_rate": 4.136868302505301e-06,
|
|
"loss": 0.4183,
|
|
"step": 5426
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.133443122919743e-06,
|
|
"loss": 0.5005,
|
|
"step": 5427
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.130018992447276e-06,
|
|
"loss": 0.4837,
|
|
"step": 5428
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.126595911700244e-06,
|
|
"loss": 0.3971,
|
|
"step": 5429
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.123173881290787e-06,
|
|
"loss": 0.465,
|
|
"step": 5430
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.119752901830865e-06,
|
|
"loss": 0.577,
|
|
"step": 5431
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.116332973932256e-06,
|
|
"loss": 0.5264,
|
|
"step": 5432
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.1129140982065415e-06,
|
|
"loss": 0.5127,
|
|
"step": 5433
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.109496275265112e-06,
|
|
"loss": 0.4928,
|
|
"step": 5434
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.1060795057191784e-06,
|
|
"loss": 0.4758,
|
|
"step": 5435
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.102663790179764e-06,
|
|
"loss": 0.4442,
|
|
"step": 5436
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.099249129257691e-06,
|
|
"loss": 0.4949,
|
|
"step": 5437
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.095835523563609e-06,
|
|
"loss": 0.404,
|
|
"step": 5438
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.092422973707965e-06,
|
|
"loss": 0.4436,
|
|
"step": 5439
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.089011480301028e-06,
|
|
"loss": 0.3884,
|
|
"step": 5440
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.08560104395287e-06,
|
|
"loss": 0.4246,
|
|
"step": 5441
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.082191665273381e-06,
|
|
"loss": 0.4838,
|
|
"step": 5442
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.078783344872254e-06,
|
|
"loss": 0.4355,
|
|
"step": 5443
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.075376083358997e-06,
|
|
"loss": 0.405,
|
|
"step": 5444
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.071969881342936e-06,
|
|
"loss": 0.4349,
|
|
"step": 5445
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.068564739433196e-06,
|
|
"loss": 0.4741,
|
|
"step": 5446
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.0651606582387115e-06,
|
|
"loss": 0.4794,
|
|
"step": 5447
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.061757638368238e-06,
|
|
"loss": 0.4543,
|
|
"step": 5448
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.058355680430337e-06,
|
|
"loss": 0.3883,
|
|
"step": 5449
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.0549547850333756e-06,
|
|
"loss": 0.4723,
|
|
"step": 5450
|
|
},
|
|
{
|
|
"epoch": 2.13,
|
|
"learning_rate": 4.051554952785531e-06,
|
|
"loss": 0.4665,
|
|
"step": 5451
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.048156184294798e-06,
|
|
"loss": 0.4909,
|
|
"step": 5452
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.044758480168979e-06,
|
|
"loss": 0.4865,
|
|
"step": 5453
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.0413618410156785e-06,
|
|
"loss": 0.4772,
|
|
"step": 5454
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.0379662674423145e-06,
|
|
"loss": 0.4862,
|
|
"step": 5455
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.034571760056117e-06,
|
|
"loss": 0.4387,
|
|
"step": 5456
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.031178319464126e-06,
|
|
"loss": 0.4672,
|
|
"step": 5457
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.027785946273195e-06,
|
|
"loss": 0.4216,
|
|
"step": 5458
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.0243946410899645e-06,
|
|
"loss": 0.4814,
|
|
"step": 5459
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.021004404520909e-06,
|
|
"loss": 0.4547,
|
|
"step": 5460
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.0176152371723e-06,
|
|
"loss": 0.429,
|
|
"step": 5461
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.0142271396502264e-06,
|
|
"loss": 0.4582,
|
|
"step": 5462
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.010840112560575e-06,
|
|
"loss": 0.4352,
|
|
"step": 5463
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.007454156509043e-06,
|
|
"loss": 0.4952,
|
|
"step": 5464
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.004069272101142e-06,
|
|
"loss": 0.4201,
|
|
"step": 5465
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 4.000685459942193e-06,
|
|
"loss": 0.4824,
|
|
"step": 5466
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.997302720637318e-06,
|
|
"loss": 0.4641,
|
|
"step": 5467
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.993921054791445e-06,
|
|
"loss": 0.4177,
|
|
"step": 5468
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.9905404630093225e-06,
|
|
"loss": 0.4664,
|
|
"step": 5469
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.9871609458955e-06,
|
|
"loss": 0.4247,
|
|
"step": 5470
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.983782504054332e-06,
|
|
"loss": 0.4565,
|
|
"step": 5471
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.98040513808998e-06,
|
|
"loss": 0.4619,
|
|
"step": 5472
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.977028848606421e-06,
|
|
"loss": 0.4521,
|
|
"step": 5473
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.973653636207437e-06,
|
|
"loss": 0.4235,
|
|
"step": 5474
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.970279501496614e-06,
|
|
"loss": 0.4669,
|
|
"step": 5475
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.96690644507734e-06,
|
|
"loss": 0.4952,
|
|
"step": 5476
|
|
},
|
|
{
|
|
"epoch": 2.14,
|
|
"learning_rate": 3.963534467552822e-06,
|
|
"loss": 0.5388,
|
|
"step": 5477
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.960163569526073e-06,
|
|
"loss": 0.4125,
|
|
"step": 5478
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.956793751599904e-06,
|
|
"loss": 0.4365,
|
|
"step": 5479
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.953425014376936e-06,
|
|
"loss": 0.4381,
|
|
"step": 5480
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.9500573584595994e-06,
|
|
"loss": 0.4904,
|
|
"step": 5481
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.946690784450132e-06,
|
|
"loss": 0.5548,
|
|
"step": 5482
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.943325292950579e-06,
|
|
"loss": 0.4913,
|
|
"step": 5483
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.939960884562787e-06,
|
|
"loss": 0.4442,
|
|
"step": 5484
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.936597559888405e-06,
|
|
"loss": 0.4786,
|
|
"step": 5485
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.9332353195289016e-06,
|
|
"loss": 0.4134,
|
|
"step": 5486
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.929874164085546e-06,
|
|
"loss": 0.4841,
|
|
"step": 5487
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.926514094159409e-06,
|
|
"loss": 0.4299,
|
|
"step": 5488
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.923155110351365e-06,
|
|
"loss": 0.451,
|
|
"step": 5489
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.919797213262105e-06,
|
|
"loss": 0.476,
|
|
"step": 5490
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.916440403492121e-06,
|
|
"loss": 0.5037,
|
|
"step": 5491
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.913084681641708e-06,
|
|
"loss": 0.488,
|
|
"step": 5492
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.9097300483109625e-06,
|
|
"loss": 0.4909,
|
|
"step": 5493
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.906376504099796e-06,
|
|
"loss": 0.4636,
|
|
"step": 5494
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.903024049607925e-06,
|
|
"loss": 0.456,
|
|
"step": 5495
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.899672685434863e-06,
|
|
"loss": 0.4168,
|
|
"step": 5496
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.896322412179929e-06,
|
|
"loss": 0.4896,
|
|
"step": 5497
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.892973230442254e-06,
|
|
"loss": 0.4795,
|
|
"step": 5498
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.889625140820774e-06,
|
|
"loss": 0.4671,
|
|
"step": 5499
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.886278143914219e-06,
|
|
"loss": 0.4428,
|
|
"step": 5500
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.882932240321138e-06,
|
|
"loss": 0.4623,
|
|
"step": 5501
|
|
},
|
|
{
|
|
"epoch": 2.15,
|
|
"learning_rate": 3.8795874306398695e-06,
|
|
"loss": 0.4517,
|
|
"step": 5502
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.8762437154685715e-06,
|
|
"loss": 0.4721,
|
|
"step": 5503
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.87290109540519e-06,
|
|
"loss": 0.4789,
|
|
"step": 5504
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.869559571047493e-06,
|
|
"loss": 0.4939,
|
|
"step": 5505
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.8662191429930366e-06,
|
|
"loss": 0.4582,
|
|
"step": 5506
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.862879811839188e-06,
|
|
"loss": 0.5179,
|
|
"step": 5507
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.859541578183126e-06,
|
|
"loss": 0.4747,
|
|
"step": 5508
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.856204442621817e-06,
|
|
"loss": 0.4467,
|
|
"step": 5509
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.8528684057520395e-06,
|
|
"loss": 0.4044,
|
|
"step": 5510
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.849533468170376e-06,
|
|
"loss": 0.4499,
|
|
"step": 5511
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.846199630473216e-06,
|
|
"loss": 0.5423,
|
|
"step": 5512
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.842866893256742e-06,
|
|
"loss": 0.4944,
|
|
"step": 5513
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.8395352571169496e-06,
|
|
"loss": 0.4381,
|
|
"step": 5514
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.836204722649629e-06,
|
|
"loss": 0.4167,
|
|
"step": 5515
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.832875290450383e-06,
|
|
"loss": 0.4397,
|
|
"step": 5516
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.829546961114608e-06,
|
|
"loss": 0.4379,
|
|
"step": 5517
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.826219735237511e-06,
|
|
"loss": 0.46,
|
|
"step": 5518
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.822893613414091e-06,
|
|
"loss": 0.4566,
|
|
"step": 5519
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.819568596239167e-06,
|
|
"loss": 0.443,
|
|
"step": 5520
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.81624468430734e-06,
|
|
"loss": 0.4109,
|
|
"step": 5521
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.8129218782130317e-06,
|
|
"loss": 0.4356,
|
|
"step": 5522
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.809600178550449e-06,
|
|
"loss": 0.4132,
|
|
"step": 5523
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.806279585913618e-06,
|
|
"loss": 0.4505,
|
|
"step": 5524
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.8029601008963514e-06,
|
|
"loss": 0.4731,
|
|
"step": 5525
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.7996417240922747e-06,
|
|
"loss": 0.4958,
|
|
"step": 5526
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.7963244560948132e-06,
|
|
"loss": 0.4311,
|
|
"step": 5527
|
|
},
|
|
{
|
|
"epoch": 2.16,
|
|
"learning_rate": 3.793008297497189e-06,
|
|
"loss": 0.4986,
|
|
"step": 5528
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7896932488924267e-06,
|
|
"loss": 0.4238,
|
|
"step": 5529
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7863793108733583e-06,
|
|
"loss": 0.4099,
|
|
"step": 5530
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.783066484032615e-06,
|
|
"loss": 0.4306,
|
|
"step": 5531
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7797547689626212e-06,
|
|
"loss": 0.4244,
|
|
"step": 5532
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7764441662556175e-06,
|
|
"loss": 0.4442,
|
|
"step": 5533
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.773134676503629e-06,
|
|
"loss": 0.4416,
|
|
"step": 5534
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7698263002984982e-06,
|
|
"loss": 0.4993,
|
|
"step": 5535
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7665190382318516e-06,
|
|
"loss": 0.4871,
|
|
"step": 5536
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7632128908951328e-06,
|
|
"loss": 0.5025,
|
|
"step": 5537
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7599078588795714e-06,
|
|
"loss": 0.4795,
|
|
"step": 5538
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.756603942776207e-06,
|
|
"loss": 0.4999,
|
|
"step": 5539
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7533011431758815e-06,
|
|
"loss": 0.5116,
|
|
"step": 5540
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.749999460669229e-06,
|
|
"loss": 0.4265,
|
|
"step": 5541
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.746698895846683e-06,
|
|
"loss": 0.4746,
|
|
"step": 5542
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.743399449298488e-06,
|
|
"loss": 0.4707,
|
|
"step": 5543
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.740101121614682e-06,
|
|
"loss": 0.4105,
|
|
"step": 5544
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.736803913385103e-06,
|
|
"loss": 0.4264,
|
|
"step": 5545
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.733507825199383e-06,
|
|
"loss": 0.4321,
|
|
"step": 5546
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7302128576469644e-06,
|
|
"loss": 0.4355,
|
|
"step": 5547
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.726919011317087e-06,
|
|
"loss": 0.4414,
|
|
"step": 5548
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7236262867987848e-06,
|
|
"loss": 0.4637,
|
|
"step": 5549
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7203346846808898e-06,
|
|
"loss": 0.4842,
|
|
"step": 5550
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7170442055520418e-06,
|
|
"loss": 0.4818,
|
|
"step": 5551
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7137548500006736e-06,
|
|
"loss": 0.378,
|
|
"step": 5552
|
|
},
|
|
{
|
|
"epoch": 2.17,
|
|
"learning_rate": 3.7104666186150286e-06,
|
|
"loss": 0.4102,
|
|
"step": 5553
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.707179511983122e-06,
|
|
"loss": 0.4719,
|
|
"step": 5554
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.703893530692795e-06,
|
|
"loss": 0.4551,
|
|
"step": 5555
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.700608675331676e-06,
|
|
"loss": 0.4434,
|
|
"step": 5556
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6973249464871983e-06,
|
|
"loss": 0.4509,
|
|
"step": 5557
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.694042344746586e-06,
|
|
"loss": 0.4587,
|
|
"step": 5558
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.69076087069686e-06,
|
|
"loss": 0.4955,
|
|
"step": 5559
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6874805249248502e-06,
|
|
"loss": 0.5164,
|
|
"step": 5560
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6842013080171803e-06,
|
|
"loss": 0.474,
|
|
"step": 5561
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6809232205602684e-06,
|
|
"loss": 0.5203,
|
|
"step": 5562
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6776462631403296e-06,
|
|
"loss": 0.4477,
|
|
"step": 5563
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6743704363433853e-06,
|
|
"loss": 0.5292,
|
|
"step": 5564
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6710957407552494e-06,
|
|
"loss": 0.4617,
|
|
"step": 5565
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6678221769615343e-06,
|
|
"loss": 0.5282,
|
|
"step": 5566
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6645497455476442e-06,
|
|
"loss": 0.4697,
|
|
"step": 5567
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6612784470987893e-06,
|
|
"loss": 0.488,
|
|
"step": 5568
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6580082821999787e-06,
|
|
"loss": 0.4808,
|
|
"step": 5569
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6547392514360103e-06,
|
|
"loss": 0.4762,
|
|
"step": 5570
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.651471355391479e-06,
|
|
"loss": 0.4284,
|
|
"step": 5571
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.648204594650786e-06,
|
|
"loss": 0.4726,
|
|
"step": 5572
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6449389697981217e-06,
|
|
"loss": 0.4257,
|
|
"step": 5573
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6416744814174855e-06,
|
|
"loss": 0.487,
|
|
"step": 5574
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.638411130092648e-06,
|
|
"loss": 0.4322,
|
|
"step": 5575
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.635148916407202e-06,
|
|
"loss": 0.4816,
|
|
"step": 5576
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.6318878409445246e-06,
|
|
"loss": 0.4831,
|
|
"step": 5577
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.628627904287797e-06,
|
|
"loss": 0.4623,
|
|
"step": 5578
|
|
},
|
|
{
|
|
"epoch": 2.18,
|
|
"learning_rate": 3.625369107019987e-06,
|
|
"loss": 0.4537,
|
|
"step": 5579
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.622111449723863e-06,
|
|
"loss": 0.5121,
|
|
"step": 5580
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.61885493298199e-06,
|
|
"loss": 0.4772,
|
|
"step": 5581
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.6155995573767354e-06,
|
|
"loss": 0.4631,
|
|
"step": 5582
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.6123453234902507e-06,
|
|
"loss": 0.4923,
|
|
"step": 5583
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.6090922319044874e-06,
|
|
"loss": 0.4573,
|
|
"step": 5584
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.6058402832011953e-06,
|
|
"loss": 0.4749,
|
|
"step": 5585
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.602589477961923e-06,
|
|
"loss": 0.4089,
|
|
"step": 5586
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.599339816768006e-06,
|
|
"loss": 0.42,
|
|
"step": 5587
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.596091300200578e-06,
|
|
"loss": 0.4431,
|
|
"step": 5588
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.5928439288405703e-06,
|
|
"loss": 0.486,
|
|
"step": 5589
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.5895977032687146e-06,
|
|
"loss": 0.3955,
|
|
"step": 5590
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.5863526240655224e-06,
|
|
"loss": 0.4702,
|
|
"step": 5591
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.583108691811317e-06,
|
|
"loss": 0.5119,
|
|
"step": 5592
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.579865907086202e-06,
|
|
"loss": 0.4671,
|
|
"step": 5593
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.5766242704700905e-06,
|
|
"loss": 0.4709,
|
|
"step": 5594
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.5733837825426756e-06,
|
|
"loss": 0.4698,
|
|
"step": 5595
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.5701444438834565e-06,
|
|
"loss": 0.4625,
|
|
"step": 5596
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.566906255071719e-06,
|
|
"loss": 0.4752,
|
|
"step": 5597
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.563669216686548e-06,
|
|
"loss": 0.4361,
|
|
"step": 5598
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.560433329306825e-06,
|
|
"loss": 0.4185,
|
|
"step": 5599
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.5571985935112186e-06,
|
|
"loss": 0.4983,
|
|
"step": 5600
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.553965009878191e-06,
|
|
"loss": 0.4523,
|
|
"step": 5601
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.550732578986006e-06,
|
|
"loss": 0.4766,
|
|
"step": 5602
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.54750130141272e-06,
|
|
"loss": 0.4324,
|
|
"step": 5603
|
|
},
|
|
{
|
|
"epoch": 2.19,
|
|
"learning_rate": 3.544271177736176e-06,
|
|
"loss": 0.4597,
|
|
"step": 5604
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5410422085340203e-06,
|
|
"loss": 0.4397,
|
|
"step": 5605
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5378143943836794e-06,
|
|
"loss": 0.5046,
|
|
"step": 5606
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5345877358623914e-06,
|
|
"loss": 0.4541,
|
|
"step": 5607
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5313622335471696e-06,
|
|
"loss": 0.4564,
|
|
"step": 5608
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5281378880148353e-06,
|
|
"loss": 0.3731,
|
|
"step": 5609
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.52491469984199e-06,
|
|
"loss": 0.457,
|
|
"step": 5610
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5216926696050412e-06,
|
|
"loss": 0.5419,
|
|
"step": 5611
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5184717978801764e-06,
|
|
"loss": 0.4343,
|
|
"step": 5612
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.515252085243388e-06,
|
|
"loss": 0.4027,
|
|
"step": 5613
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.51203353227045e-06,
|
|
"loss": 0.451,
|
|
"step": 5614
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5088161395369414e-06,
|
|
"loss": 0.4636,
|
|
"step": 5615
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5055999076182177e-06,
|
|
"loss": 0.5152,
|
|
"step": 5616
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.5023848370894455e-06,
|
|
"loss": 0.4173,
|
|
"step": 5617
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4991709285255648e-06,
|
|
"loss": 0.5063,
|
|
"step": 5618
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4959581825013256e-06,
|
|
"loss": 0.4308,
|
|
"step": 5619
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4927465995912546e-06,
|
|
"loss": 0.4924,
|
|
"step": 5620
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4895361803696792e-06,
|
|
"loss": 0.45,
|
|
"step": 5621
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.486326925410722e-06,
|
|
"loss": 0.4751,
|
|
"step": 5622
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.483118835288285e-06,
|
|
"loss": 0.4565,
|
|
"step": 5623
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.479911910576075e-06,
|
|
"loss": 0.4303,
|
|
"step": 5624
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4767061518475798e-06,
|
|
"loss": 0.4455,
|
|
"step": 5625
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.473501559676088e-06,
|
|
"loss": 0.4805,
|
|
"step": 5626
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4702981346346698e-06,
|
|
"loss": 0.4642,
|
|
"step": 5627
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4670958772961972e-06,
|
|
"loss": 0.4066,
|
|
"step": 5628
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.463894788233323e-06,
|
|
"loss": 0.4517,
|
|
"step": 5629
|
|
},
|
|
{
|
|
"epoch": 2.2,
|
|
"learning_rate": 3.4606948680185015e-06,
|
|
"loss": 0.4222,
|
|
"step": 5630
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4574961172239675e-06,
|
|
"loss": 0.4442,
|
|
"step": 5631
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4542985364217575e-06,
|
|
"loss": 0.5324,
|
|
"step": 5632
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4511021261836865e-06,
|
|
"loss": 0.4364,
|
|
"step": 5633
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.447906887081369e-06,
|
|
"loss": 0.4819,
|
|
"step": 5634
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4447128196862134e-06,
|
|
"loss": 0.4946,
|
|
"step": 5635
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4415199245694084e-06,
|
|
"loss": 0.4394,
|
|
"step": 5636
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.438328202301935e-06,
|
|
"loss": 0.4581,
|
|
"step": 5637
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4351376534545688e-06,
|
|
"loss": 0.4658,
|
|
"step": 5638
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4319482785978796e-06,
|
|
"loss": 0.4414,
|
|
"step": 5639
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4287600783022158e-06,
|
|
"loss": 0.4343,
|
|
"step": 5640
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4255730531377198e-06,
|
|
"loss": 0.3896,
|
|
"step": 5641
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.422387203674328e-06,
|
|
"loss": 0.5128,
|
|
"step": 5642
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4192025304817677e-06,
|
|
"loss": 0.4784,
|
|
"step": 5643
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4160190341295497e-06,
|
|
"loss": 0.5203,
|
|
"step": 5644
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.412836715186971e-06,
|
|
"loss": 0.4695,
|
|
"step": 5645
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4096555742231306e-06,
|
|
"loss": 0.4418,
|
|
"step": 5646
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.4064756118069086e-06,
|
|
"loss": 0.449,
|
|
"step": 5647
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.403296828506979e-06,
|
|
"loss": 0.4753,
|
|
"step": 5648
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.400119224891798e-06,
|
|
"loss": 0.4779,
|
|
"step": 5649
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.3969428015296135e-06,
|
|
"loss": 0.4777,
|
|
"step": 5650
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.3937675589884644e-06,
|
|
"loss": 0.4531,
|
|
"step": 5651
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.3905934978361823e-06,
|
|
"loss": 0.4682,
|
|
"step": 5652
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.387420618640379e-06,
|
|
"loss": 0.4866,
|
|
"step": 5653
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.3842489219684547e-06,
|
|
"loss": 0.4479,
|
|
"step": 5654
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.381078408387608e-06,
|
|
"loss": 0.4765,
|
|
"step": 5655
|
|
},
|
|
{
|
|
"epoch": 2.21,
|
|
"learning_rate": 3.3779090784648203e-06,
|
|
"loss": 0.4404,
|
|
"step": 5656
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.37474093276686e-06,
|
|
"loss": 0.4088,
|
|
"step": 5657
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3715739718602803e-06,
|
|
"loss": 0.4775,
|
|
"step": 5658
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.36840819631143e-06,
|
|
"loss": 0.4459,
|
|
"step": 5659
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.365243606686448e-06,
|
|
"loss": 0.4305,
|
|
"step": 5660
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.362080203551251e-06,
|
|
"loss": 0.4476,
|
|
"step": 5661
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.358917987471546e-06,
|
|
"loss": 0.461,
|
|
"step": 5662
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3557569590128346e-06,
|
|
"loss": 0.43,
|
|
"step": 5663
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.352597118740404e-06,
|
|
"loss": 0.4535,
|
|
"step": 5664
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3494384672193235e-06,
|
|
"loss": 0.4483,
|
|
"step": 5665
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.34628100501445e-06,
|
|
"loss": 0.5052,
|
|
"step": 5666
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3431247326904324e-06,
|
|
"loss": 0.4989,
|
|
"step": 5667
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3399696508117087e-06,
|
|
"loss": 0.4438,
|
|
"step": 5668
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3368157599425034e-06,
|
|
"loss": 0.4705,
|
|
"step": 5669
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3336630606468136e-06,
|
|
"loss": 0.4985,
|
|
"step": 5670
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3305115534884403e-06,
|
|
"loss": 0.4388,
|
|
"step": 5671
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3273612390309673e-06,
|
|
"loss": 0.4385,
|
|
"step": 5672
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.324212117837766e-06,
|
|
"loss": 0.4456,
|
|
"step": 5673
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3210641904719885e-06,
|
|
"loss": 0.4761,
|
|
"step": 5674
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3179174574965745e-06,
|
|
"loss": 0.4176,
|
|
"step": 5675
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.3147719194742545e-06,
|
|
"loss": 0.4819,
|
|
"step": 5676
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.311627576967548e-06,
|
|
"loss": 0.4481,
|
|
"step": 5677
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.308484430538751e-06,
|
|
"loss": 0.4895,
|
|
"step": 5678
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.305342480749949e-06,
|
|
"loss": 0.4743,
|
|
"step": 5679
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.302201728163017e-06,
|
|
"loss": 0.504,
|
|
"step": 5680
|
|
},
|
|
{
|
|
"epoch": 2.22,
|
|
"learning_rate": 3.299062173339619e-06,
|
|
"loss": 0.4648,
|
|
"step": 5681
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2959238168411944e-06,
|
|
"loss": 0.4781,
|
|
"step": 5682
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2927866592289725e-06,
|
|
"loss": 0.4678,
|
|
"step": 5683
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2896507010639724e-06,
|
|
"loss": 0.4239,
|
|
"step": 5684
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.286515942906997e-06,
|
|
"loss": 0.4692,
|
|
"step": 5685
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.28338238531863e-06,
|
|
"loss": 0.5169,
|
|
"step": 5686
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2802500288592477e-06,
|
|
"loss": 0.4351,
|
|
"step": 5687
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2771188740890024e-06,
|
|
"loss": 0.5031,
|
|
"step": 5688
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2739889215678423e-06,
|
|
"loss": 0.4337,
|
|
"step": 5689
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.27086017185549e-06,
|
|
"loss": 0.4796,
|
|
"step": 5690
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2677326255114628e-06,
|
|
"loss": 0.4986,
|
|
"step": 5691
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2646062830950533e-06,
|
|
"loss": 0.4487,
|
|
"step": 5692
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.261481145165346e-06,
|
|
"loss": 0.4664,
|
|
"step": 5693
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2583572122812114e-06,
|
|
"loss": 0.4653,
|
|
"step": 5694
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.255234485001295e-06,
|
|
"loss": 0.4347,
|
|
"step": 5695
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2521129638840333e-06,
|
|
"loss": 0.4044,
|
|
"step": 5696
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2489926494876454e-06,
|
|
"loss": 0.4495,
|
|
"step": 5697
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2458735423701416e-06,
|
|
"loss": 0.4692,
|
|
"step": 5698
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2427556430893017e-06,
|
|
"loss": 0.4386,
|
|
"step": 5699
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.239638952202705e-06,
|
|
"loss": 0.4741,
|
|
"step": 5700
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.236523470267702e-06,
|
|
"loss": 0.4698,
|
|
"step": 5701
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.233409197841437e-06,
|
|
"loss": 0.4559,
|
|
"step": 5702
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.230296135480829e-06,
|
|
"loss": 0.4569,
|
|
"step": 5703
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2271842837425917e-06,
|
|
"loss": 0.5093,
|
|
"step": 5704
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2240736431832077e-06,
|
|
"loss": 0.4464,
|
|
"step": 5705
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.2209642143589583e-06,
|
|
"loss": 0.4702,
|
|
"step": 5706
|
|
},
|
|
{
|
|
"epoch": 2.23,
|
|
"learning_rate": 3.217855997825895e-06,
|
|
"loss": 0.4602,
|
|
"step": 5707
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.214748994139866e-06,
|
|
"loss": 0.4426,
|
|
"step": 5708
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.211643203856486e-06,
|
|
"loss": 0.4772,
|
|
"step": 5709
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.208538627531169e-06,
|
|
"loss": 0.4706,
|
|
"step": 5710
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.205435265719099e-06,
|
|
"loss": 0.5003,
|
|
"step": 5711
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.2023331189752516e-06,
|
|
"loss": 0.438,
|
|
"step": 5712
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.199232187854384e-06,
|
|
"loss": 0.4743,
|
|
"step": 5713
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1961324729110333e-06,
|
|
"loss": 0.4876,
|
|
"step": 5714
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1930339746995132e-06,
|
|
"loss": 0.4696,
|
|
"step": 5715
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1899366937739316e-06,
|
|
"loss": 0.4865,
|
|
"step": 5716
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1868406306881773e-06,
|
|
"loss": 0.4478,
|
|
"step": 5717
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1837457859959096e-06,
|
|
"loss": 0.4434,
|
|
"step": 5718
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.180652160250586e-06,
|
|
"loss": 0.4442,
|
|
"step": 5719
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1775597540054293e-06,
|
|
"loss": 0.532,
|
|
"step": 5720
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.174468567813461e-06,
|
|
"loss": 0.4568,
|
|
"step": 5721
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1713786022274685e-06,
|
|
"loss": 0.4618,
|
|
"step": 5722
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1682898578000355e-06,
|
|
"loss": 0.4438,
|
|
"step": 5723
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.165202335083515e-06,
|
|
"loss": 0.4325,
|
|
"step": 5724
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.162116034630048e-06,
|
|
"loss": 0.4992,
|
|
"step": 5725
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1590309569915624e-06,
|
|
"loss": 0.439,
|
|
"step": 5726
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.155947102719754e-06,
|
|
"loss": 0.4709,
|
|
"step": 5727
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.152864472366106e-06,
|
|
"loss": 0.4105,
|
|
"step": 5728
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1497830664818865e-06,
|
|
"loss": 0.3945,
|
|
"step": 5729
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.1467028856181435e-06,
|
|
"loss": 0.4572,
|
|
"step": 5730
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.143623930325701e-06,
|
|
"loss": 0.5222,
|
|
"step": 5731
|
|
},
|
|
{
|
|
"epoch": 2.24,
|
|
"learning_rate": 3.140546201155166e-06,
|
|
"loss": 0.4655,
|
|
"step": 5732
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.1374696986569266e-06,
|
|
"loss": 0.4395,
|
|
"step": 5733
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.1343944233811587e-06,
|
|
"loss": 0.4805,
|
|
"step": 5734
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.1313203758778055e-06,
|
|
"loss": 0.4551,
|
|
"step": 5735
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.1282475566965975e-06,
|
|
"loss": 0.4053,
|
|
"step": 5736
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.1251759663870462e-06,
|
|
"loss": 0.4736,
|
|
"step": 5737
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.122105605498442e-06,
|
|
"loss": 0.436,
|
|
"step": 5738
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.119036474579864e-06,
|
|
"loss": 0.5052,
|
|
"step": 5739
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.115968574180149e-06,
|
|
"loss": 0.4779,
|
|
"step": 5740
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.112901904847936e-06,
|
|
"loss": 0.3913,
|
|
"step": 5741
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.1098364671316337e-06,
|
|
"loss": 0.4216,
|
|
"step": 5742
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.106772261579437e-06,
|
|
"loss": 0.4653,
|
|
"step": 5743
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.1037092887393126e-06,
|
|
"loss": 0.4025,
|
|
"step": 5744
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.100647549159007e-06,
|
|
"loss": 0.4208,
|
|
"step": 5745
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.097587043386052e-06,
|
|
"loss": 0.4825,
|
|
"step": 5746
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.0945277719677612e-06,
|
|
"loss": 0.4747,
|
|
"step": 5747
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.091469735451217e-06,
|
|
"loss": 0.4937,
|
|
"step": 5748
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.0884129343832846e-06,
|
|
"loss": 0.4153,
|
|
"step": 5749
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.085357369310612e-06,
|
|
"loss": 0.4866,
|
|
"step": 5750
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.0823030407796283e-06,
|
|
"loss": 0.4612,
|
|
"step": 5751
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.0792499493365346e-06,
|
|
"loss": 0.4478,
|
|
"step": 5752
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.0761980955273096e-06,
|
|
"loss": 0.4605,
|
|
"step": 5753
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.073147479897718e-06,
|
|
"loss": 0.4382,
|
|
"step": 5754
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.0700981029933017e-06,
|
|
"loss": 0.4239,
|
|
"step": 5755
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.0670499653593786e-06,
|
|
"loss": 0.4774,
|
|
"step": 5756
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.06400306754104e-06,
|
|
"loss": 0.4942,
|
|
"step": 5757
|
|
},
|
|
{
|
|
"epoch": 2.25,
|
|
"learning_rate": 3.060957410083165e-06,
|
|
"loss": 0.4911,
|
|
"step": 5758
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0579129935304065e-06,
|
|
"loss": 0.483,
|
|
"step": 5759
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0548698184272017e-06,
|
|
"loss": 0.482,
|
|
"step": 5760
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0518278853177487e-06,
|
|
"loss": 0.4175,
|
|
"step": 5761
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0487871947460402e-06,
|
|
"loss": 0.4746,
|
|
"step": 5762
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0457477472558404e-06,
|
|
"loss": 0.4562,
|
|
"step": 5763
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0427095433906996e-06,
|
|
"loss": 0.4172,
|
|
"step": 5764
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0396725836939246e-06,
|
|
"loss": 0.4716,
|
|
"step": 5765
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.03663686870862e-06,
|
|
"loss": 0.4295,
|
|
"step": 5766
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0336023989776608e-06,
|
|
"loss": 0.4468,
|
|
"step": 5767
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.030569175043704e-06,
|
|
"loss": 0.4308,
|
|
"step": 5768
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.027537197449173e-06,
|
|
"loss": 0.4548,
|
|
"step": 5769
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0245064667362745e-06,
|
|
"loss": 0.4254,
|
|
"step": 5770
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0214769834469948e-06,
|
|
"loss": 0.4337,
|
|
"step": 5771
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.018448748123097e-06,
|
|
"loss": 0.431,
|
|
"step": 5772
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0154217613061165e-06,
|
|
"loss": 0.4146,
|
|
"step": 5773
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.012396023537364e-06,
|
|
"loss": 0.4232,
|
|
"step": 5774
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0093715353579334e-06,
|
|
"loss": 0.4095,
|
|
"step": 5775
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0063482973086966e-06,
|
|
"loss": 0.5011,
|
|
"step": 5776
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.003326309930292e-06,
|
|
"loss": 0.4695,
|
|
"step": 5777
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 3.0003055737631404e-06,
|
|
"loss": 0.45,
|
|
"step": 5778
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 2.997286089347439e-06,
|
|
"loss": 0.479,
|
|
"step": 5779
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 2.994267857223163e-06,
|
|
"loss": 0.418,
|
|
"step": 5780
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 2.9912508779300575e-06,
|
|
"loss": 0.5169,
|
|
"step": 5781
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 2.9882351520076513e-06,
|
|
"loss": 0.4238,
|
|
"step": 5782
|
|
},
|
|
{
|
|
"epoch": 2.26,
|
|
"learning_rate": 2.98522067999524e-06,
|
|
"loss": 0.4756,
|
|
"step": 5783
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.982207462431902e-06,
|
|
"loss": 0.4281,
|
|
"step": 5784
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.979195499856492e-06,
|
|
"loss": 0.4385,
|
|
"step": 5785
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.976184792807637e-06,
|
|
"loss": 0.495,
|
|
"step": 5786
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9731753418237353e-06,
|
|
"loss": 0.4064,
|
|
"step": 5787
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9701671474429685e-06,
|
|
"loss": 0.5029,
|
|
"step": 5788
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9671602102032926e-06,
|
|
"loss": 0.4757,
|
|
"step": 5789
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9641545306424346e-06,
|
|
"loss": 0.4808,
|
|
"step": 5790
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9611501092978954e-06,
|
|
"loss": 0.5057,
|
|
"step": 5791
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.958146946706957e-06,
|
|
"loss": 0.4423,
|
|
"step": 5792
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9551450434066752e-06,
|
|
"loss": 0.4291,
|
|
"step": 5793
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.952144399933875e-06,
|
|
"loss": 0.4493,
|
|
"step": 5794
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9491450168251635e-06,
|
|
"loss": 0.4756,
|
|
"step": 5795
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9461468946169137e-06,
|
|
"loss": 0.4393,
|
|
"step": 5796
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9431500338452833e-06,
|
|
"loss": 0.5225,
|
|
"step": 5797
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.940154435046194e-06,
|
|
"loss": 0.4648,
|
|
"step": 5798
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.937160098755354e-06,
|
|
"loss": 0.4131,
|
|
"step": 5799
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.934167025508231e-06,
|
|
"loss": 0.4652,
|
|
"step": 5800
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9311752158400817e-06,
|
|
"loss": 0.4478,
|
|
"step": 5801
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.928184670285924e-06,
|
|
"loss": 0.4466,
|
|
"step": 5802
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9251953893805597e-06,
|
|
"loss": 0.5145,
|
|
"step": 5803
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9222073736585564e-06,
|
|
"loss": 0.4431,
|
|
"step": 5804
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9192206236542642e-06,
|
|
"loss": 0.4072,
|
|
"step": 5805
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.9162351399017964e-06,
|
|
"loss": 0.4221,
|
|
"step": 5806
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.913250922935047e-06,
|
|
"loss": 0.4433,
|
|
"step": 5807
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.910267973287687e-06,
|
|
"loss": 0.4815,
|
|
"step": 5808
|
|
},
|
|
{
|
|
"epoch": 2.27,
|
|
"learning_rate": 2.907286291493149e-06,
|
|
"loss": 0.4397,
|
|
"step": 5809
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.90430587808465e-06,
|
|
"loss": 0.4308,
|
|
"step": 5810
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.901326733595171e-06,
|
|
"loss": 0.4729,
|
|
"step": 5811
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.898348858557476e-06,
|
|
"loss": 0.4845,
|
|
"step": 5812
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8953722535040916e-06,
|
|
"loss": 0.4118,
|
|
"step": 5813
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8923969189673286e-06,
|
|
"loss": 0.4218,
|
|
"step": 5814
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8894228554792583e-06,
|
|
"loss": 0.3941,
|
|
"step": 5815
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.886450063571735e-06,
|
|
"loss": 0.4619,
|
|
"step": 5816
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.883478543776378e-06,
|
|
"loss": 0.4957,
|
|
"step": 5817
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.880508296624586e-06,
|
|
"loss": 0.4708,
|
|
"step": 5818
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8775393226475225e-06,
|
|
"loss": 0.3853,
|
|
"step": 5819
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.87457162237613e-06,
|
|
"loss": 0.4244,
|
|
"step": 5820
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.871605196341124e-06,
|
|
"loss": 0.4529,
|
|
"step": 5821
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8686400450729857e-06,
|
|
"loss": 0.4434,
|
|
"step": 5822
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8656761691019673e-06,
|
|
"loss": 0.4713,
|
|
"step": 5823
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.862713568958101e-06,
|
|
"loss": 0.4755,
|
|
"step": 5824
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.85975224517119e-06,
|
|
"loss": 0.4423,
|
|
"step": 5825
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8567921982708034e-06,
|
|
"loss": 0.4871,
|
|
"step": 5826
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.853833428786281e-06,
|
|
"loss": 0.4204,
|
|
"step": 5827
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.850875937246742e-06,
|
|
"loss": 0.4653,
|
|
"step": 5828
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.847919724181074e-06,
|
|
"loss": 0.387,
|
|
"step": 5829
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8449647901179343e-06,
|
|
"loss": 0.469,
|
|
"step": 5830
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.842011135585748e-06,
|
|
"loss": 0.5157,
|
|
"step": 5831
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8390587611127183e-06,
|
|
"loss": 0.4645,
|
|
"step": 5832
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.836107667226816e-06,
|
|
"loss": 0.4192,
|
|
"step": 5833
|
|
},
|
|
{
|
|
"epoch": 2.28,
|
|
"learning_rate": 2.8331578544557893e-06,
|
|
"loss": 0.5012,
|
|
"step": 5834
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.8302093233271454e-06,
|
|
"loss": 0.4343,
|
|
"step": 5835
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.8272620743681666e-06,
|
|
"loss": 0.5139,
|
|
"step": 5836
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.824316108105911e-06,
|
|
"loss": 0.4119,
|
|
"step": 5837
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.8213714250672085e-06,
|
|
"loss": 0.438,
|
|
"step": 5838
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.818428025778649e-06,
|
|
"loss": 0.4533,
|
|
"step": 5839
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.815485910766599e-06,
|
|
"loss": 0.4802,
|
|
"step": 5840
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.8125450805571963e-06,
|
|
"loss": 0.4616,
|
|
"step": 5841
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.809605535676352e-06,
|
|
"loss": 0.4923,
|
|
"step": 5842
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.8066672766497395e-06,
|
|
"loss": 0.4916,
|
|
"step": 5843
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.803730304002803e-06,
|
|
"loss": 0.4627,
|
|
"step": 5844
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.800794618260764e-06,
|
|
"loss": 0.4478,
|
|
"step": 5845
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.7978602199486106e-06,
|
|
"loss": 0.4423,
|
|
"step": 5846
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.794927109591099e-06,
|
|
"loss": 0.444,
|
|
"step": 5847
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.79199528771275e-06,
|
|
"loss": 0.4499,
|
|
"step": 5848
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.7890647548378646e-06,
|
|
"loss": 0.4485,
|
|
"step": 5849
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.786135511490511e-06,
|
|
"loss": 0.4727,
|
|
"step": 5850
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.7832075581945207e-06,
|
|
"loss": 0.4646,
|
|
"step": 5851
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.780280895473494e-06,
|
|
"loss": 0.4817,
|
|
"step": 5852
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.77735552385081e-06,
|
|
"loss": 0.464,
|
|
"step": 5853
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.774431443849609e-06,
|
|
"loss": 0.3877,
|
|
"step": 5854
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.7715086559928097e-06,
|
|
"loss": 0.4521,
|
|
"step": 5855
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.76858716080308e-06,
|
|
"loss": 0.4445,
|
|
"step": 5856
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.765666958802876e-06,
|
|
"loss": 0.5313,
|
|
"step": 5857
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.7627480505144157e-06,
|
|
"loss": 0.4673,
|
|
"step": 5858
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.7598304364596874e-06,
|
|
"loss": 0.4931,
|
|
"step": 5859
|
|
},
|
|
{
|
|
"epoch": 2.29,
|
|
"learning_rate": 2.756914117160445e-06,
|
|
"loss": 0.4703,
|
|
"step": 5860
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.753999093138209e-06,
|
|
"loss": 0.4473,
|
|
"step": 5861
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7510853649142743e-06,
|
|
"loss": 0.451,
|
|
"step": 5862
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7481729330097042e-06,
|
|
"loss": 0.5362,
|
|
"step": 5863
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7452617979453244e-06,
|
|
"loss": 0.4809,
|
|
"step": 5864
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.742351960241728e-06,
|
|
"loss": 0.4386,
|
|
"step": 5865
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7394434204192834e-06,
|
|
"loss": 0.442,
|
|
"step": 5866
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.736536178998126e-06,
|
|
"loss": 0.4381,
|
|
"step": 5867
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7336302364981527e-06,
|
|
"loss": 0.4624,
|
|
"step": 5868
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.73072559343903e-06,
|
|
"loss": 0.4443,
|
|
"step": 5869
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7278222503401943e-06,
|
|
"loss": 0.4773,
|
|
"step": 5870
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7249202077208535e-06,
|
|
"loss": 0.4121,
|
|
"step": 5871
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7220194660999753e-06,
|
|
"loss": 0.4468,
|
|
"step": 5872
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7191200259962938e-06,
|
|
"loss": 0.4481,
|
|
"step": 5873
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7162218879283174e-06,
|
|
"loss": 0.4647,
|
|
"step": 5874
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.713325052414323e-06,
|
|
"loss": 0.4861,
|
|
"step": 5875
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7104295199723417e-06,
|
|
"loss": 0.4239,
|
|
"step": 5876
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.707535291120188e-06,
|
|
"loss": 0.4665,
|
|
"step": 5877
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.7046423663754275e-06,
|
|
"loss": 0.5013,
|
|
"step": 5878
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.701750746255405e-06,
|
|
"loss": 0.4233,
|
|
"step": 5879
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.698860431277228e-06,
|
|
"loss": 0.4518,
|
|
"step": 5880
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.6959714219577703e-06,
|
|
"loss": 0.4804,
|
|
"step": 5881
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.693083718813666e-06,
|
|
"loss": 0.4548,
|
|
"step": 5882
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.6901973223613255e-06,
|
|
"loss": 0.457,
|
|
"step": 5883
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.6873122331169243e-06,
|
|
"loss": 0.4965,
|
|
"step": 5884
|
|
},
|
|
{
|
|
"epoch": 2.3,
|
|
"learning_rate": 2.6844284515963993e-06,
|
|
"loss": 0.5165,
|
|
"step": 5885
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6815459783154505e-06,
|
|
"loss": 0.4901,
|
|
"step": 5886
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6786648137895533e-06,
|
|
"loss": 0.4485,
|
|
"step": 5887
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.675784958533948e-06,
|
|
"loss": 0.4447,
|
|
"step": 5888
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6729064130636305e-06,
|
|
"loss": 0.4435,
|
|
"step": 5889
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6700291778933764e-06,
|
|
"loss": 0.4333,
|
|
"step": 5890
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6671532535377132e-06,
|
|
"loss": 0.4899,
|
|
"step": 5891
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6642786405109477e-06,
|
|
"loss": 0.4842,
|
|
"step": 5892
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.661405339327139e-06,
|
|
"loss": 0.4146,
|
|
"step": 5893
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.658533350500123e-06,
|
|
"loss": 0.4506,
|
|
"step": 5894
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.655662674543492e-06,
|
|
"loss": 0.5,
|
|
"step": 5895
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.652793311970612e-06,
|
|
"loss": 0.5161,
|
|
"step": 5896
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.649925263294603e-06,
|
|
"loss": 0.4622,
|
|
"step": 5897
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6470585290283636e-06,
|
|
"loss": 0.4427,
|
|
"step": 5898
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.644193109684544e-06,
|
|
"loss": 0.443,
|
|
"step": 5899
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6413290057755714e-06,
|
|
"loss": 0.4947,
|
|
"step": 5900
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.638466217813627e-06,
|
|
"loss": 0.4135,
|
|
"step": 5901
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6356047463106615e-06,
|
|
"loss": 0.47,
|
|
"step": 5902
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.632744591778397e-06,
|
|
"loss": 0.4655,
|
|
"step": 5903
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.629885754728304e-06,
|
|
"loss": 0.4586,
|
|
"step": 5904
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6270282356716358e-06,
|
|
"loss": 0.4721,
|
|
"step": 5905
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.624172035119393e-06,
|
|
"loss": 0.4116,
|
|
"step": 5906
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6213171535823556e-06,
|
|
"loss": 0.4507,
|
|
"step": 5907
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.618463591571052e-06,
|
|
"loss": 0.447,
|
|
"step": 5908
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6156113495957924e-06,
|
|
"loss": 0.4389,
|
|
"step": 5909
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.612760428166633e-06,
|
|
"loss": 0.4585,
|
|
"step": 5910
|
|
},
|
|
{
|
|
"epoch": 2.31,
|
|
"learning_rate": 2.6099108277934105e-06,
|
|
"loss": 0.4605,
|
|
"step": 5911
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.6070625489857105e-06,
|
|
"loss": 0.408,
|
|
"step": 5912
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.6042155922528946e-06,
|
|
"loss": 0.467,
|
|
"step": 5913
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.6013699581040765e-06,
|
|
"loss": 0.4467,
|
|
"step": 5914
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5985256470481425e-06,
|
|
"loss": 0.456,
|
|
"step": 5915
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5956826595937433e-06,
|
|
"loss": 0.4567,
|
|
"step": 5916
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5928409962492852e-06,
|
|
"loss": 0.4158,
|
|
"step": 5917
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5900006575229365e-06,
|
|
"loss": 0.5345,
|
|
"step": 5918
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.587161643922639e-06,
|
|
"loss": 0.4598,
|
|
"step": 5919
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.584323955956094e-06,
|
|
"loss": 0.4434,
|
|
"step": 5920
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5814875941307593e-06,
|
|
"loss": 0.4208,
|
|
"step": 5921
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.578652558953859e-06,
|
|
"loss": 0.4979,
|
|
"step": 5922
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.575818850932382e-06,
|
|
"loss": 0.4566,
|
|
"step": 5923
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5729864705730835e-06,
|
|
"loss": 0.4748,
|
|
"step": 5924
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.570155418382473e-06,
|
|
"loss": 0.5014,
|
|
"step": 5925
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.567325694866821e-06,
|
|
"loss": 0.4431,
|
|
"step": 5926
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5644973005321714e-06,
|
|
"loss": 0.4256,
|
|
"step": 5927
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.561670235884323e-06,
|
|
"loss": 0.4481,
|
|
"step": 5928
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.55884450142884e-06,
|
|
"loss": 0.4575,
|
|
"step": 5929
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.556020097671046e-06,
|
|
"loss": 0.4925,
|
|
"step": 5930
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5531970251160233e-06,
|
|
"loss": 0.4706,
|
|
"step": 5931
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5503752842686237e-06,
|
|
"loss": 0.445,
|
|
"step": 5932
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5475548756334602e-06,
|
|
"loss": 0.4468,
|
|
"step": 5933
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5447357997149025e-06,
|
|
"loss": 0.4675,
|
|
"step": 5934
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.541918057017082e-06,
|
|
"loss": 0.4866,
|
|
"step": 5935
|
|
},
|
|
{
|
|
"epoch": 2.32,
|
|
"learning_rate": 2.5391016480438945e-06,
|
|
"loss": 0.469,
|
|
"step": 5936
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5362865732990027e-06,
|
|
"loss": 0.4081,
|
|
"step": 5937
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5334728332858207e-06,
|
|
"loss": 0.4808,
|
|
"step": 5938
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5306604285075254e-06,
|
|
"loss": 0.4501,
|
|
"step": 5939
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5278493594670597e-06,
|
|
"loss": 0.4603,
|
|
"step": 5940
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5250396266671286e-06,
|
|
"loss": 0.4952,
|
|
"step": 5941
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5222312306101925e-06,
|
|
"loss": 0.4882,
|
|
"step": 5942
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5194241717984726e-06,
|
|
"loss": 0.4258,
|
|
"step": 5943
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5166184507339566e-06,
|
|
"loss": 0.4206,
|
|
"step": 5944
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.51381406791839e-06,
|
|
"loss": 0.4328,
|
|
"step": 5945
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5110110238532837e-06,
|
|
"loss": 0.4315,
|
|
"step": 5946
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.508209319039895e-06,
|
|
"loss": 0.4667,
|
|
"step": 5947
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5054089539792547e-06,
|
|
"loss": 0.422,
|
|
"step": 5948
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.5026099291721517e-06,
|
|
"loss": 0.418,
|
|
"step": 5949
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.49981224511914e-06,
|
|
"loss": 0.4827,
|
|
"step": 5950
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4970159023205167e-06,
|
|
"loss": 0.4414,
|
|
"step": 5951
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.494220901276354e-06,
|
|
"loss": 0.4507,
|
|
"step": 5952
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4914272424864827e-06,
|
|
"loss": 0.4564,
|
|
"step": 5953
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.488634926450493e-06,
|
|
"loss": 0.4794,
|
|
"step": 5954
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4858439536677305e-06,
|
|
"loss": 0.5324,
|
|
"step": 5955
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4830543246373007e-06,
|
|
"loss": 0.4965,
|
|
"step": 5956
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4802660398580735e-06,
|
|
"loss": 0.4039,
|
|
"step": 5957
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4774790998286804e-06,
|
|
"loss": 0.5087,
|
|
"step": 5958
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.474693505047504e-06,
|
|
"loss": 0.4418,
|
|
"step": 5959
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.471909256012689e-06,
|
|
"loss": 0.4452,
|
|
"step": 5960
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4691263532221433e-06,
|
|
"loss": 0.4377,
|
|
"step": 5961
|
|
},
|
|
{
|
|
"epoch": 2.33,
|
|
"learning_rate": 2.4663447971735356e-06,
|
|
"loss": 0.4218,
|
|
"step": 5962
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4635645883642855e-06,
|
|
"loss": 0.3877,
|
|
"step": 5963
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.460785727291575e-06,
|
|
"loss": 0.5032,
|
|
"step": 5964
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4580082144523486e-06,
|
|
"loss": 0.4293,
|
|
"step": 5965
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.455232050343309e-06,
|
|
"loss": 0.4358,
|
|
"step": 5966
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.452457235460913e-06,
|
|
"loss": 0.4625,
|
|
"step": 5967
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.449683770301382e-06,
|
|
"loss": 0.4918,
|
|
"step": 5968
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4469116553606898e-06,
|
|
"loss": 0.4472,
|
|
"step": 5969
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4441408911345743e-06,
|
|
"loss": 0.4829,
|
|
"step": 5970
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.441371478118532e-06,
|
|
"loss": 0.4707,
|
|
"step": 5971
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.438603416807813e-06,
|
|
"loss": 0.4876,
|
|
"step": 5972
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4358367076974252e-06,
|
|
"loss": 0.4505,
|
|
"step": 5973
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4330713512821403e-06,
|
|
"loss": 0.4589,
|
|
"step": 5974
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.430307348056489e-06,
|
|
"loss": 0.4207,
|
|
"step": 5975
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.427544698514753e-06,
|
|
"loss": 0.4107,
|
|
"step": 5976
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.424783403150973e-06,
|
|
"loss": 0.4136,
|
|
"step": 5977
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4220234624589534e-06,
|
|
"loss": 0.4645,
|
|
"step": 5978
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4192648769322537e-06,
|
|
"loss": 0.4432,
|
|
"step": 5979
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4165076470641856e-06,
|
|
"loss": 0.4236,
|
|
"step": 5980
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.413751773347829e-06,
|
|
"loss": 0.4586,
|
|
"step": 5981
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4109972562760096e-06,
|
|
"loss": 0.4146,
|
|
"step": 5982
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.408244096341321e-06,
|
|
"loss": 0.4207,
|
|
"step": 5983
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.4054922940361047e-06,
|
|
"loss": 0.4535,
|
|
"step": 5984
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.402741849852468e-06,
|
|
"loss": 0.5009,
|
|
"step": 5985
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.3999927642822673e-06,
|
|
"loss": 0.4553,
|
|
"step": 5986
|
|
},
|
|
{
|
|
"epoch": 2.34,
|
|
"learning_rate": 2.3972450378171254e-06,
|
|
"loss": 0.4153,
|
|
"step": 5987
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3944986709484107e-06,
|
|
"loss": 0.4686,
|
|
"step": 5988
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.39175366416726e-06,
|
|
"loss": 0.4384,
|
|
"step": 5989
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3890100179645547e-06,
|
|
"loss": 0.4245,
|
|
"step": 5990
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.386267732830948e-06,
|
|
"loss": 0.4113,
|
|
"step": 5991
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.383526809256833e-06,
|
|
"loss": 0.4699,
|
|
"step": 5992
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3807872477323736e-06,
|
|
"loss": 0.4525,
|
|
"step": 5993
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3780490487474794e-06,
|
|
"loss": 0.4303,
|
|
"step": 5994
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3753122127918227e-06,
|
|
"loss": 0.4749,
|
|
"step": 5995
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.372576740354835e-06,
|
|
"loss": 0.5027,
|
|
"step": 5996
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3698426319256905e-06,
|
|
"loss": 0.4366,
|
|
"step": 5997
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3671098879933373e-06,
|
|
"loss": 0.4384,
|
|
"step": 5998
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.364378509046463e-06,
|
|
"loss": 0.5022,
|
|
"step": 5999
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.361648495573524e-06,
|
|
"loss": 0.4813,
|
|
"step": 6000
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3589198480627228e-06,
|
|
"loss": 0.4775,
|
|
"step": 6001
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3561925670020268e-06,
|
|
"loss": 0.4944,
|
|
"step": 6002
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.353466652879147e-06,
|
|
"loss": 0.418,
|
|
"step": 6003
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.350742106181566e-06,
|
|
"loss": 0.5069,
|
|
"step": 6004
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.348018927396505e-06,
|
|
"loss": 0.5019,
|
|
"step": 6005
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.345297117010954e-06,
|
|
"loss": 0.4248,
|
|
"step": 6006
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3425766755116486e-06,
|
|
"loss": 0.4384,
|
|
"step": 6007
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3398576033850894e-06,
|
|
"loss": 0.4592,
|
|
"step": 6008
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.337139901117519e-06,
|
|
"loss": 0.4761,
|
|
"step": 6009
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.334423569194948e-06,
|
|
"loss": 0.4666,
|
|
"step": 6010
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3317086081031372e-06,
|
|
"loss": 0.4147,
|
|
"step": 6011
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.3289950183276e-06,
|
|
"loss": 0.4449,
|
|
"step": 6012
|
|
},
|
|
{
|
|
"epoch": 2.35,
|
|
"learning_rate": 2.326282800353602e-06,
|
|
"loss": 0.4563,
|
|
"step": 6013
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.323571954666173e-06,
|
|
"loss": 0.4123,
|
|
"step": 6014
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.320862481750091e-06,
|
|
"loss": 0.5211,
|
|
"step": 6015
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.318154382089891e-06,
|
|
"loss": 0.4944,
|
|
"step": 6016
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.3154476561698546e-06,
|
|
"loss": 0.4675,
|
|
"step": 6017
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.3127423044740293e-06,
|
|
"loss": 0.4246,
|
|
"step": 6018
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.3100383274862136e-06,
|
|
"loss": 0.4296,
|
|
"step": 6019
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.307335725689952e-06,
|
|
"loss": 0.4345,
|
|
"step": 6020
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.3046344995685554e-06,
|
|
"loss": 0.4564,
|
|
"step": 6021
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.3019346496050777e-06,
|
|
"loss": 0.5076,
|
|
"step": 6022
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2992361762823333e-06,
|
|
"loss": 0.4181,
|
|
"step": 6023
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2965390800828913e-06,
|
|
"loss": 0.472,
|
|
"step": 6024
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2938433614890696e-06,
|
|
"loss": 0.4468,
|
|
"step": 6025
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.29114902098294e-06,
|
|
"loss": 0.4661,
|
|
"step": 6026
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.288456059046331e-06,
|
|
"loss": 0.4905,
|
|
"step": 6027
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.285764476160828e-06,
|
|
"loss": 0.4339,
|
|
"step": 6028
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2830742728077616e-06,
|
|
"loss": 0.4976,
|
|
"step": 6029
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2803854494682164e-06,
|
|
"loss": 0.5003,
|
|
"step": 6030
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.277698006623036e-06,
|
|
"loss": 0.4791,
|
|
"step": 6031
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2750119447528174e-06,
|
|
"loss": 0.4893,
|
|
"step": 6032
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2723272643379056e-06,
|
|
"loss": 0.4133,
|
|
"step": 6033
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.269643965858396e-06,
|
|
"loss": 0.4122,
|
|
"step": 6034
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2669620497941446e-06,
|
|
"loss": 0.4227,
|
|
"step": 6035
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2642815166247613e-06,
|
|
"loss": 0.4739,
|
|
"step": 6036
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2616023668295993e-06,
|
|
"loss": 0.4002,
|
|
"step": 6037
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2589246008877684e-06,
|
|
"loss": 0.4802,
|
|
"step": 6038
|
|
},
|
|
{
|
|
"epoch": 2.36,
|
|
"learning_rate": 2.2562482192781332e-06,
|
|
"loss": 0.517,
|
|
"step": 6039
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2535732224793118e-06,
|
|
"loss": 0.5053,
|
|
"step": 6040
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2508996109696757e-06,
|
|
"loss": 0.5032,
|
|
"step": 6041
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2482273852273352e-06,
|
|
"loss": 0.4317,
|
|
"step": 6042
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2455565457301686e-06,
|
|
"loss": 0.4837,
|
|
"step": 6043
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2428870929558012e-06,
|
|
"loss": 0.4713,
|
|
"step": 6044
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2402190273816106e-06,
|
|
"loss": 0.4417,
|
|
"step": 6045
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.237552349484725e-06,
|
|
"loss": 0.4922,
|
|
"step": 6046
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.23488705974202e-06,
|
|
"loss": 0.4228,
|
|
"step": 6047
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2322231586301323e-06,
|
|
"loss": 0.4689,
|
|
"step": 6048
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2295606466254483e-06,
|
|
"loss": 0.4328,
|
|
"step": 6049
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2268995242041004e-06,
|
|
"loss": 0.4838,
|
|
"step": 6050
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.224239791841972e-06,
|
|
"loss": 0.4263,
|
|
"step": 6051
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.221581450014706e-06,
|
|
"loss": 0.4981,
|
|
"step": 6052
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.218924499197694e-06,
|
|
"loss": 0.555,
|
|
"step": 6053
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2162689398660755e-06,
|
|
"loss": 0.4048,
|
|
"step": 6054
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2136147724947376e-06,
|
|
"loss": 0.4429,
|
|
"step": 6055
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2109619975583283e-06,
|
|
"loss": 0.4177,
|
|
"step": 6056
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.208310615531244e-06,
|
|
"loss": 0.4519,
|
|
"step": 6057
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.205660626887627e-06,
|
|
"loss": 0.4272,
|
|
"step": 6058
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2030120321013716e-06,
|
|
"loss": 0.4351,
|
|
"step": 6059
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.2003648316461245e-06,
|
|
"loss": 0.4283,
|
|
"step": 6060
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.1977190259952887e-06,
|
|
"loss": 0.3897,
|
|
"step": 6061
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.195074615622006e-06,
|
|
"loss": 0.413,
|
|
"step": 6062
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.1924316009991785e-06,
|
|
"loss": 0.4484,
|
|
"step": 6063
|
|
},
|
|
{
|
|
"epoch": 2.37,
|
|
"learning_rate": 2.189789982599453e-06,
|
|
"loss": 0.434,
|
|
"step": 6064
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1871497608952273e-06,
|
|
"loss": 0.4684,
|
|
"step": 6065
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.184510936358657e-06,
|
|
"loss": 0.4036,
|
|
"step": 6066
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1818735094616374e-06,
|
|
"loss": 0.44,
|
|
"step": 6067
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1792374806758144e-06,
|
|
"loss": 0.4925,
|
|
"step": 6068
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1766028504725923e-06,
|
|
"loss": 0.4375,
|
|
"step": 6069
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1739696193231218e-06,
|
|
"loss": 0.4471,
|
|
"step": 6070
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1713377876983e-06,
|
|
"loss": 0.4861,
|
|
"step": 6071
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1687073560687732e-06,
|
|
"loss": 0.4083,
|
|
"step": 6072
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.166078324904942e-06,
|
|
"loss": 0.4184,
|
|
"step": 6073
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1634506946769583e-06,
|
|
"loss": 0.4106,
|
|
"step": 6074
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.160824465854714e-06,
|
|
"loss": 0.4224,
|
|
"step": 6075
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.158199638907862e-06,
|
|
"loss": 0.4254,
|
|
"step": 6076
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1555762143057925e-06,
|
|
"loss": 0.4791,
|
|
"step": 6077
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1529541925176555e-06,
|
|
"loss": 0.478,
|
|
"step": 6078
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1503335740123434e-06,
|
|
"loss": 0.4513,
|
|
"step": 6079
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1477143592585016e-06,
|
|
"loss": 0.4611,
|
|
"step": 6080
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.14509654872452e-06,
|
|
"loss": 0.4069,
|
|
"step": 6081
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1424801428785447e-06,
|
|
"loss": 0.5017,
|
|
"step": 6082
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.139865142188461e-06,
|
|
"loss": 0.4699,
|
|
"step": 6083
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1372515471219134e-06,
|
|
"loss": 0.5064,
|
|
"step": 6084
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.134639358146283e-06,
|
|
"loss": 0.4617,
|
|
"step": 6085
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.132028575728714e-06,
|
|
"loss": 0.4671,
|
|
"step": 6086
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1294192003360837e-06,
|
|
"loss": 0.4648,
|
|
"step": 6087
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.1268112324350287e-06,
|
|
"loss": 0.4417,
|
|
"step": 6088
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.124204672491933e-06,
|
|
"loss": 0.5097,
|
|
"step": 6089
|
|
},
|
|
{
|
|
"epoch": 2.38,
|
|
"learning_rate": 2.121599520972921e-06,
|
|
"loss": 0.4357,
|
|
"step": 6090
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.118995778343875e-06,
|
|
"loss": 0.4872,
|
|
"step": 6091
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.1163934450704173e-06,
|
|
"loss": 0.3986,
|
|
"step": 6092
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.1137925216179256e-06,
|
|
"loss": 0.441,
|
|
"step": 6093
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.1111930084515163e-06,
|
|
"loss": 0.4542,
|
|
"step": 6094
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.1085949060360654e-06,
|
|
"loss": 0.4532,
|
|
"step": 6095
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.1059982148361836e-06,
|
|
"loss": 0.4241,
|
|
"step": 6096
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.1034029353162412e-06,
|
|
"loss": 0.4202,
|
|
"step": 6097
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.1008090679403457e-06,
|
|
"loss": 0.4097,
|
|
"step": 6098
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0982166131723624e-06,
|
|
"loss": 0.4067,
|
|
"step": 6099
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0956255714758923e-06,
|
|
"loss": 0.4952,
|
|
"step": 6100
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0930359433142934e-06,
|
|
"loss": 0.5228,
|
|
"step": 6101
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0904477291506695e-06,
|
|
"loss": 0.436,
|
|
"step": 6102
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.087860929447868e-06,
|
|
"loss": 0.4935,
|
|
"step": 6103
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0852755446684813e-06,
|
|
"loss": 0.4819,
|
|
"step": 6104
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0826915752748546e-06,
|
|
"loss": 0.4744,
|
|
"step": 6105
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.080109021729081e-06,
|
|
"loss": 0.4265,
|
|
"step": 6106
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0775278844929947e-06,
|
|
"loss": 0.4379,
|
|
"step": 6107
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0749481640281753e-06,
|
|
"loss": 0.4677,
|
|
"step": 6108
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0723698607959575e-06,
|
|
"loss": 0.4625,
|
|
"step": 6109
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0697929752574177e-06,
|
|
"loss": 0.4823,
|
|
"step": 6110
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.067217507873378e-06,
|
|
"loss": 0.4536,
|
|
"step": 6111
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.064643459104405e-06,
|
|
"loss": 0.4847,
|
|
"step": 6112
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.062070829410817e-06,
|
|
"loss": 0.4744,
|
|
"step": 6113
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.0594996192526783e-06,
|
|
"loss": 0.4482,
|
|
"step": 6114
|
|
},
|
|
{
|
|
"epoch": 2.39,
|
|
"learning_rate": 2.056929829089792e-06,
|
|
"loss": 0.4609,
|
|
"step": 6115
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0543614593817173e-06,
|
|
"loss": 0.4247,
|
|
"step": 6116
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.051794510587749e-06,
|
|
"loss": 0.471,
|
|
"step": 6117
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0492289831669366e-06,
|
|
"loss": 0.3946,
|
|
"step": 6118
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0466648775780727e-06,
|
|
"loss": 0.4319,
|
|
"step": 6119
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0441021942796947e-06,
|
|
"loss": 0.462,
|
|
"step": 6120
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0415409337300806e-06,
|
|
"loss": 0.4905,
|
|
"step": 6121
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.038981096387264e-06,
|
|
"loss": 0.4349,
|
|
"step": 6122
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0364226827090195e-06,
|
|
"loss": 0.4249,
|
|
"step": 6123
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0338656931528667e-06,
|
|
"loss": 0.4218,
|
|
"step": 6124
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.031310128176067e-06,
|
|
"loss": 0.4592,
|
|
"step": 6125
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0287559882356323e-06,
|
|
"loss": 0.4662,
|
|
"step": 6126
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.026203273788322e-06,
|
|
"loss": 0.4791,
|
|
"step": 6127
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0236519852906344e-06,
|
|
"loss": 0.4113,
|
|
"step": 6128
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0211021231988103e-06,
|
|
"loss": 0.4653,
|
|
"step": 6129
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0185536879688438e-06,
|
|
"loss": 0.4282,
|
|
"step": 6130
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0160066800564703e-06,
|
|
"loss": 0.4176,
|
|
"step": 6131
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0134610999171756e-06,
|
|
"loss": 0.4193,
|
|
"step": 6132
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.010916948006173e-06,
|
|
"loss": 0.4658,
|
|
"step": 6133
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.008374224778438e-06,
|
|
"loss": 0.4545,
|
|
"step": 6134
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0058329306886824e-06,
|
|
"loss": 0.4432,
|
|
"step": 6135
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.003293066191372e-06,
|
|
"loss": 0.4224,
|
|
"step": 6136
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 2.0007546317406967e-06,
|
|
"loss": 0.4492,
|
|
"step": 6137
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 1.9982176277906094e-06,
|
|
"loss": 0.4429,
|
|
"step": 6138
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 1.995682054794803e-06,
|
|
"loss": 0.4144,
|
|
"step": 6139
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 1.993147913206712e-06,
|
|
"loss": 0.4576,
|
|
"step": 6140
|
|
},
|
|
{
|
|
"epoch": 2.4,
|
|
"learning_rate": 1.9906152034795158e-06,
|
|
"loss": 0.5267,
|
|
"step": 6141
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.988083926066132e-06,
|
|
"loss": 0.4591,
|
|
"step": 6142
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9855540814192333e-06,
|
|
"loss": 0.4599,
|
|
"step": 6143
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.98302566999123e-06,
|
|
"loss": 0.4767,
|
|
"step": 6144
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.980498692234276e-06,
|
|
"loss": 0.4156,
|
|
"step": 6145
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9779731486002664e-06,
|
|
"loss": 0.4365,
|
|
"step": 6146
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9754490395408444e-06,
|
|
"loss": 0.4768,
|
|
"step": 6147
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.972926365507398e-06,
|
|
"loss": 0.4198,
|
|
"step": 6148
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9704051269510537e-06,
|
|
"loss": 0.4819,
|
|
"step": 6149
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9678853243226793e-06,
|
|
"loss": 0.4659,
|
|
"step": 6150
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.965366958072894e-06,
|
|
"loss": 0.4319,
|
|
"step": 6151
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.962850028652057e-06,
|
|
"loss": 0.5278,
|
|
"step": 6152
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9603345365102666e-06,
|
|
"loss": 0.4543,
|
|
"step": 6153
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9578204820973657e-06,
|
|
"loss": 0.5301,
|
|
"step": 6154
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9553078658629434e-06,
|
|
"loss": 0.4346,
|
|
"step": 6155
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9527966882563287e-06,
|
|
"loss": 0.4527,
|
|
"step": 6156
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9502869497265973e-06,
|
|
"loss": 0.4113,
|
|
"step": 6157
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9477786507225615e-06,
|
|
"loss": 0.4176,
|
|
"step": 6158
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9452717916927756e-06,
|
|
"loss": 0.419,
|
|
"step": 6159
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.942766373085544e-06,
|
|
"loss": 0.4564,
|
|
"step": 6160
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.940262395348912e-06,
|
|
"loss": 0.468,
|
|
"step": 6161
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9377598589306603e-06,
|
|
"loss": 0.4032,
|
|
"step": 6162
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9352587642783137e-06,
|
|
"loss": 0.4553,
|
|
"step": 6163
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9327591118391466e-06,
|
|
"loss": 0.4142,
|
|
"step": 6164
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.9302609020601692e-06,
|
|
"loss": 0.4654,
|
|
"step": 6165
|
|
},
|
|
{
|
|
"epoch": 2.41,
|
|
"learning_rate": 1.927764135388136e-06,
|
|
"loss": 0.4491,
|
|
"step": 6166
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.9252688122695373e-06,
|
|
"loss": 0.4682,
|
|
"step": 6167
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.922774933150614e-06,
|
|
"loss": 0.4848,
|
|
"step": 6168
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.9202824984773483e-06,
|
|
"loss": 0.522,
|
|
"step": 6169
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.9177915086954536e-06,
|
|
"loss": 0.4364,
|
|
"step": 6170
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.9153019642504e-06,
|
|
"loss": 0.4361,
|
|
"step": 6171
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.912813865587384e-06,
|
|
"loss": 0.4536,
|
|
"step": 6172
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.9103272131513573e-06,
|
|
"loss": 0.4468,
|
|
"step": 6173
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.907842007387002e-06,
|
|
"loss": 0.4017,
|
|
"step": 6174
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.905358248738749e-06,
|
|
"loss": 0.4206,
|
|
"step": 6175
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.9028759376507645e-06,
|
|
"loss": 0.5573,
|
|
"step": 6176
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.900395074566962e-06,
|
|
"loss": 0.4275,
|
|
"step": 6177
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.89791565993099e-06,
|
|
"loss": 0.4253,
|
|
"step": 6178
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.895437694186244e-06,
|
|
"loss": 0.464,
|
|
"step": 6179
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8929611777758528e-06,
|
|
"loss": 0.4846,
|
|
"step": 6180
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8904861111426942e-06,
|
|
"loss": 0.439,
|
|
"step": 6181
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8880124947293832e-06,
|
|
"loss": 0.4441,
|
|
"step": 6182
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8855403289782726e-06,
|
|
"loss": 0.5296,
|
|
"step": 6183
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8830696143314608e-06,
|
|
"loss": 0.4591,
|
|
"step": 6184
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8806003512307825e-06,
|
|
"loss": 0.4005,
|
|
"step": 6185
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8781325401178163e-06,
|
|
"loss": 0.4678,
|
|
"step": 6186
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8756661814338773e-06,
|
|
"loss": 0.4372,
|
|
"step": 6187
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8732012756200258e-06,
|
|
"loss": 0.3924,
|
|
"step": 6188
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.870737823117056e-06,
|
|
"loss": 0.4714,
|
|
"step": 6189
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8682758243655107e-06,
|
|
"loss": 0.494,
|
|
"step": 6190
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8658152798056618e-06,
|
|
"loss": 0.4382,
|
|
"step": 6191
|
|
},
|
|
{
|
|
"epoch": 2.42,
|
|
"learning_rate": 1.8633561898775331e-06,
|
|
"loss": 0.4358,
|
|
"step": 6192
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8608985550208768e-06,
|
|
"loss": 0.4248,
|
|
"step": 6193
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8584423756751957e-06,
|
|
"loss": 0.486,
|
|
"step": 6194
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8559876522797216e-06,
|
|
"loss": 0.4822,
|
|
"step": 6195
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8535343852734333e-06,
|
|
"loss": 0.4476,
|
|
"step": 6196
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8510825750950512e-06,
|
|
"loss": 0.4271,
|
|
"step": 6197
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8486322221830277e-06,
|
|
"loss": 0.5148,
|
|
"step": 6198
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8461833269755546e-06,
|
|
"loss": 0.4266,
|
|
"step": 6199
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8437358899105707e-06,
|
|
"loss": 0.4196,
|
|
"step": 6200
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.841289911425752e-06,
|
|
"loss": 0.4477,
|
|
"step": 6201
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.838845391958508e-06,
|
|
"loss": 0.441,
|
|
"step": 6202
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8364023319459888e-06,
|
|
"loss": 0.5036,
|
|
"step": 6203
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8339607318250875e-06,
|
|
"loss": 0.4282,
|
|
"step": 6204
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8315205920324375e-06,
|
|
"loss": 0.4822,
|
|
"step": 6205
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8290819130044024e-06,
|
|
"loss": 0.4777,
|
|
"step": 6206
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8266446951770945e-06,
|
|
"loss": 0.4193,
|
|
"step": 6207
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.824208938986355e-06,
|
|
"loss": 0.4095,
|
|
"step": 6208
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.821774644867772e-06,
|
|
"loss": 0.4811,
|
|
"step": 6209
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.819341813256671e-06,
|
|
"loss": 0.4898,
|
|
"step": 6210
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8169104445881103e-06,
|
|
"loss": 0.5003,
|
|
"step": 6211
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8144805392968901e-06,
|
|
"loss": 0.4653,
|
|
"step": 6212
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8120520978175494e-06,
|
|
"loss": 0.4548,
|
|
"step": 6213
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8096251205843685e-06,
|
|
"loss": 0.4332,
|
|
"step": 6214
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8071996080313602e-06,
|
|
"loss": 0.4672,
|
|
"step": 6215
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8047755605922735e-06,
|
|
"loss": 0.4608,
|
|
"step": 6216
|
|
},
|
|
{
|
|
"epoch": 2.43,
|
|
"learning_rate": 1.8023529787006044e-06,
|
|
"loss": 0.4504,
|
|
"step": 6217
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7999318627895824e-06,
|
|
"loss": 0.4626,
|
|
"step": 6218
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7975122132921718e-06,
|
|
"loss": 0.4228,
|
|
"step": 6219
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7950940306410757e-06,
|
|
"loss": 0.5011,
|
|
"step": 6220
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7926773152687394e-06,
|
|
"loss": 0.4974,
|
|
"step": 6221
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7902620676073433e-06,
|
|
"loss": 0.4718,
|
|
"step": 6222
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7878482880888037e-06,
|
|
"loss": 0.4338,
|
|
"step": 6223
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7854359771447715e-06,
|
|
"loss": 0.4175,
|
|
"step": 6224
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7830251352066442e-06,
|
|
"loss": 0.5287,
|
|
"step": 6225
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7806157627055487e-06,
|
|
"loss": 0.4608,
|
|
"step": 6226
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7782078600723573e-06,
|
|
"loss": 0.4698,
|
|
"step": 6227
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7758014277376645e-06,
|
|
"loss": 0.47,
|
|
"step": 6228
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7733964661318158e-06,
|
|
"loss": 0.4345,
|
|
"step": 6229
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7709929756848898e-06,
|
|
"loss": 0.4132,
|
|
"step": 6230
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7685909568267034e-06,
|
|
"loss": 0.4838,
|
|
"step": 6231
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7661904099868054e-06,
|
|
"loss": 0.4796,
|
|
"step": 6232
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7637913355944814e-06,
|
|
"loss": 0.4329,
|
|
"step": 6233
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7613937340787602e-06,
|
|
"loss": 0.4228,
|
|
"step": 6234
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7589976058684067e-06,
|
|
"loss": 0.3916,
|
|
"step": 6235
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7566029513919148e-06,
|
|
"loss": 0.4543,
|
|
"step": 6236
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7542097710775163e-06,
|
|
"loss": 0.4164,
|
|
"step": 6237
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.751818065353188e-06,
|
|
"loss": 0.4609,
|
|
"step": 6238
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7494278346466365e-06,
|
|
"loss": 0.4691,
|
|
"step": 6239
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7470390793853042e-06,
|
|
"loss": 0.401,
|
|
"step": 6240
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7446517999963686e-06,
|
|
"loss": 0.502,
|
|
"step": 6241
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7422659969067478e-06,
|
|
"loss": 0.45,
|
|
"step": 6242
|
|
},
|
|
{
|
|
"epoch": 2.44,
|
|
"learning_rate": 1.7398816705430966e-06,
|
|
"loss": 0.4596,
|
|
"step": 6243
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7374988213318001e-06,
|
|
"loss": 0.4785,
|
|
"step": 6244
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7351174496989775e-06,
|
|
"loss": 0.4198,
|
|
"step": 6245
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.732737556070494e-06,
|
|
"loss": 0.4616,
|
|
"step": 6246
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7303591408719444e-06,
|
|
"loss": 0.4981,
|
|
"step": 6247
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7279822045286577e-06,
|
|
"loss": 0.4227,
|
|
"step": 6248
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.725606747465699e-06,
|
|
"loss": 0.4522,
|
|
"step": 6249
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.72323277010787e-06,
|
|
"loss": 0.4685,
|
|
"step": 6250
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7208602728797098e-06,
|
|
"loss": 0.502,
|
|
"step": 6251
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7184892562054922e-06,
|
|
"loss": 0.4506,
|
|
"step": 6252
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7161197205092217e-06,
|
|
"loss": 0.5026,
|
|
"step": 6253
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7137516662146403e-06,
|
|
"loss": 0.4395,
|
|
"step": 6254
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.711385093745227e-06,
|
|
"loss": 0.4869,
|
|
"step": 6255
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7090200035241977e-06,
|
|
"loss": 0.4725,
|
|
"step": 6256
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7066563959744974e-06,
|
|
"loss": 0.391,
|
|
"step": 6257
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.704294271518807e-06,
|
|
"loss": 0.4592,
|
|
"step": 6258
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.7019336305795452e-06,
|
|
"loss": 0.469,
|
|
"step": 6259
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6995744735788677e-06,
|
|
"loss": 0.4254,
|
|
"step": 6260
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.697216800938657e-06,
|
|
"loss": 0.4277,
|
|
"step": 6261
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6948606130805334e-06,
|
|
"loss": 0.4667,
|
|
"step": 6262
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6925059104258557e-06,
|
|
"loss": 0.4662,
|
|
"step": 6263
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6901526933957147e-06,
|
|
"loss": 0.4064,
|
|
"step": 6264
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6878009624109315e-06,
|
|
"loss": 0.4157,
|
|
"step": 6265
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6854507178920687e-06,
|
|
"loss": 0.4257,
|
|
"step": 6266
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6831019602594146e-06,
|
|
"loss": 0.4708,
|
|
"step": 6267
|
|
},
|
|
{
|
|
"epoch": 2.45,
|
|
"learning_rate": 1.6807546899330018e-06,
|
|
"loss": 0.5052,
|
|
"step": 6268
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.678408907332586e-06,
|
|
"loss": 0.4788,
|
|
"step": 6269
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.676064612877667e-06,
|
|
"loss": 0.4549,
|
|
"step": 6270
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6737218069874683e-06,
|
|
"loss": 0.4987,
|
|
"step": 6271
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6713804900809583e-06,
|
|
"loss": 0.4831,
|
|
"step": 6272
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6690406625768285e-06,
|
|
"loss": 0.4202,
|
|
"step": 6273
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6667023248935122e-06,
|
|
"loss": 0.4829,
|
|
"step": 6274
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6643654774491713e-06,
|
|
"loss": 0.4291,
|
|
"step": 6275
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6620301206617018e-06,
|
|
"loss": 0.4535,
|
|
"step": 6276
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6596962549487395e-06,
|
|
"loss": 0.4433,
|
|
"step": 6277
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6573638807276416e-06,
|
|
"loss": 0.4798,
|
|
"step": 6278
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.655032998415511e-06,
|
|
"loss": 0.4707,
|
|
"step": 6279
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6527036084291736e-06,
|
|
"loss": 0.4417,
|
|
"step": 6280
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.650375711185197e-06,
|
|
"loss": 0.4039,
|
|
"step": 6281
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6480493070998738e-06,
|
|
"loss": 0.5064,
|
|
"step": 6282
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6457243965892367e-06,
|
|
"loss": 0.456,
|
|
"step": 6283
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6434009800690454e-06,
|
|
"loss": 0.4412,
|
|
"step": 6284
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6410790579547976e-06,
|
|
"loss": 0.4476,
|
|
"step": 6285
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6387586306617186e-06,
|
|
"loss": 0.3993,
|
|
"step": 6286
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6364396986047736e-06,
|
|
"loss": 0.4306,
|
|
"step": 6287
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6341222621986508e-06,
|
|
"loss": 0.4658,
|
|
"step": 6288
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6318063218577807e-06,
|
|
"loss": 0.445,
|
|
"step": 6289
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6294918779963165e-06,
|
|
"loss": 0.4791,
|
|
"step": 6290
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.6271789310281515e-06,
|
|
"loss": 0.4459,
|
|
"step": 6291
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.624867481366913e-06,
|
|
"loss": 0.4908,
|
|
"step": 6292
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.622557529425951e-06,
|
|
"loss": 0.4359,
|
|
"step": 6293
|
|
},
|
|
{
|
|
"epoch": 2.46,
|
|
"learning_rate": 1.620249075618352e-06,
|
|
"loss": 0.434,
|
|
"step": 6294
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.6179421203569368e-06,
|
|
"loss": 0.4342,
|
|
"step": 6295
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.615636664054261e-06,
|
|
"loss": 0.3952,
|
|
"step": 6296
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.6133327071226047e-06,
|
|
"loss": 0.4517,
|
|
"step": 6297
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.6110302499739806e-06,
|
|
"loss": 0.4832,
|
|
"step": 6298
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.6087292930201393e-06,
|
|
"loss": 0.4123,
|
|
"step": 6299
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.6064298366725607e-06,
|
|
"loss": 0.4248,
|
|
"step": 6300
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.6041318813424522e-06,
|
|
"loss": 0.4563,
|
|
"step": 6301
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.6018354274407588e-06,
|
|
"loss": 0.4853,
|
|
"step": 6302
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5995404753781497e-06,
|
|
"loss": 0.4514,
|
|
"step": 6303
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.597247025565033e-06,
|
|
"loss": 0.4775,
|
|
"step": 6304
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5949550784115475e-06,
|
|
"loss": 0.4272,
|
|
"step": 6305
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5926646343275576e-06,
|
|
"loss": 0.4752,
|
|
"step": 6306
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.59037569372266e-06,
|
|
"loss": 0.4638,
|
|
"step": 6307
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5880882570061884e-06,
|
|
"loss": 0.4335,
|
|
"step": 6308
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5858023245872044e-06,
|
|
"loss": 0.4554,
|
|
"step": 6309
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.583517896874498e-06,
|
|
"loss": 0.4354,
|
|
"step": 6310
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5812349742765897e-06,
|
|
"loss": 0.4622,
|
|
"step": 6311
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5789535572017368e-06,
|
|
"loss": 0.4685,
|
|
"step": 6312
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5766736460579258e-06,
|
|
"loss": 0.4562,
|
|
"step": 6313
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5743952412528685e-06,
|
|
"loss": 0.4443,
|
|
"step": 6314
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.572118343194009e-06,
|
|
"loss": 0.4486,
|
|
"step": 6315
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5698429522885273e-06,
|
|
"loss": 0.4589,
|
|
"step": 6316
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5675690689433275e-06,
|
|
"loss": 0.467,
|
|
"step": 6317
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5652966935650549e-06,
|
|
"loss": 0.4849,
|
|
"step": 6318
|
|
},
|
|
{
|
|
"epoch": 2.47,
|
|
"learning_rate": 1.5630258265600652e-06,
|
|
"loss": 0.4301,
|
|
"step": 6319
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5607564683344611e-06,
|
|
"loss": 0.4909,
|
|
"step": 6320
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.558488619294073e-06,
|
|
"loss": 0.47,
|
|
"step": 6321
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5562222798444614e-06,
|
|
"loss": 0.4689,
|
|
"step": 6322
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.553957450390906e-06,
|
|
"loss": 0.3752,
|
|
"step": 6323
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5516941313384294e-06,
|
|
"loss": 0.4608,
|
|
"step": 6324
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5494323230917807e-06,
|
|
"loss": 0.5112,
|
|
"step": 6325
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5471720260554378e-06,
|
|
"loss": 0.503,
|
|
"step": 6326
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5449132406336088e-06,
|
|
"loss": 0.4187,
|
|
"step": 6327
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5426559672302277e-06,
|
|
"loss": 0.445,
|
|
"step": 6328
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5404002062489631e-06,
|
|
"loss": 0.433,
|
|
"step": 6329
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5381459580932134e-06,
|
|
"loss": 0.4354,
|
|
"step": 6330
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5358932231661039e-06,
|
|
"loss": 0.4249,
|
|
"step": 6331
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5336420018704867e-06,
|
|
"loss": 0.4594,
|
|
"step": 6332
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5313922946089488e-06,
|
|
"loss": 0.4267,
|
|
"step": 6333
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5291441017838071e-06,
|
|
"loss": 0.3914,
|
|
"step": 6334
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5268974237971012e-06,
|
|
"loss": 0.47,
|
|
"step": 6335
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.524652261050601e-06,
|
|
"loss": 0.5268,
|
|
"step": 6336
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5224086139458116e-06,
|
|
"loss": 0.4486,
|
|
"step": 6337
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5201664828839634e-06,
|
|
"loss": 0.4756,
|
|
"step": 6338
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5179258682660136e-06,
|
|
"loss": 0.418,
|
|
"step": 6339
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5156867704926492e-06,
|
|
"loss": 0.4639,
|
|
"step": 6340
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5134491899642877e-06,
|
|
"loss": 0.4469,
|
|
"step": 6341
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.511213127081076e-06,
|
|
"loss": 0.4406,
|
|
"step": 6342
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.508978582242887e-06,
|
|
"loss": 0.4587,
|
|
"step": 6343
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.506745555849325e-06,
|
|
"loss": 0.4385,
|
|
"step": 6344
|
|
},
|
|
{
|
|
"epoch": 2.48,
|
|
"learning_rate": 1.5045140482997155e-06,
|
|
"loss": 0.4152,
|
|
"step": 6345
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.502284059993121e-06,
|
|
"loss": 0.4995,
|
|
"step": 6346
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.500055591328331e-06,
|
|
"loss": 0.4721,
|
|
"step": 6347
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4978286427038602e-06,
|
|
"loss": 0.4946,
|
|
"step": 6348
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.495603214517949e-06,
|
|
"loss": 0.4305,
|
|
"step": 6349
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.493379307168573e-06,
|
|
"loss": 0.4503,
|
|
"step": 6350
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4911569210534327e-06,
|
|
"loss": 0.4561,
|
|
"step": 6351
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4889360565699562e-06,
|
|
"loss": 0.4209,
|
|
"step": 6352
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.486716714115295e-06,
|
|
"loss": 0.4768,
|
|
"step": 6353
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.484498894086337e-06,
|
|
"loss": 0.4646,
|
|
"step": 6354
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.482282596879695e-06,
|
|
"loss": 0.4931,
|
|
"step": 6355
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4800678228917031e-06,
|
|
"loss": 0.4451,
|
|
"step": 6356
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4778545725184346e-06,
|
|
"loss": 0.5155,
|
|
"step": 6357
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4756428461556771e-06,
|
|
"loss": 0.4815,
|
|
"step": 6358
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.473432644198959e-06,
|
|
"loss": 0.4126,
|
|
"step": 6359
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4712239670435225e-06,
|
|
"loss": 0.451,
|
|
"step": 6360
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4690168150843509e-06,
|
|
"loss": 0.3949,
|
|
"step": 6361
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4668111887161418e-06,
|
|
"loss": 0.4336,
|
|
"step": 6362
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4646070883333318e-06,
|
|
"loss": 0.4762,
|
|
"step": 6363
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4624045143300735e-06,
|
|
"loss": 0.4864,
|
|
"step": 6364
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4602034671002574e-06,
|
|
"loss": 0.4281,
|
|
"step": 6365
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4580039470374906e-06,
|
|
"loss": 0.4079,
|
|
"step": 6366
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4558059545351144e-06,
|
|
"loss": 0.4051,
|
|
"step": 6367
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4536094899861964e-06,
|
|
"loss": 0.5319,
|
|
"step": 6368
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4514145537835266e-06,
|
|
"loss": 0.442,
|
|
"step": 6369
|
|
},
|
|
{
|
|
"epoch": 2.49,
|
|
"learning_rate": 1.4492211463196237e-06,
|
|
"loss": 0.4403,
|
|
"step": 6370
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4470292679867348e-06,
|
|
"loss": 0.4784,
|
|
"step": 6371
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4448389191768331e-06,
|
|
"loss": 0.4638,
|
|
"step": 6372
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4426501002816152e-06,
|
|
"loss": 0.4129,
|
|
"step": 6373
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.44046281169251e-06,
|
|
"loss": 0.4098,
|
|
"step": 6374
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.438277053800664e-06,
|
|
"loss": 0.4731,
|
|
"step": 6375
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4360928269969598e-06,
|
|
"loss": 0.3976,
|
|
"step": 6376
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4339101316719984e-06,
|
|
"loss": 0.4489,
|
|
"step": 6377
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4317289682161129e-06,
|
|
"loss": 0.4838,
|
|
"step": 6378
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4295493370193547e-06,
|
|
"loss": 0.4319,
|
|
"step": 6379
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.427371238471512e-06,
|
|
"loss": 0.4417,
|
|
"step": 6380
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.425194672962088e-06,
|
|
"loss": 0.4553,
|
|
"step": 6381
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.42301964088032e-06,
|
|
"loss": 0.4221,
|
|
"step": 6382
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4208461426151644e-06,
|
|
"loss": 0.437,
|
|
"step": 6383
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4186741785553116e-06,
|
|
"loss": 0.5249,
|
|
"step": 6384
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4165037490891676e-06,
|
|
"loss": 0.5078,
|
|
"step": 6385
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4143348546048706e-06,
|
|
"loss": 0.458,
|
|
"step": 6386
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4121674954902863e-06,
|
|
"loss": 0.4812,
|
|
"step": 6387
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.410001672133e-06,
|
|
"loss": 0.4071,
|
|
"step": 6388
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.407837384920322e-06,
|
|
"loss": 0.4571,
|
|
"step": 6389
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4056746342392924e-06,
|
|
"loss": 0.4402,
|
|
"step": 6390
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4035134204766764e-06,
|
|
"loss": 0.4713,
|
|
"step": 6391
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.4013537440189606e-06,
|
|
"loss": 0.4466,
|
|
"step": 6392
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.3991956052523613e-06,
|
|
"loss": 0.4106,
|
|
"step": 6393
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.3970390045628135e-06,
|
|
"loss": 0.4342,
|
|
"step": 6394
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.3948839423359851e-06,
|
|
"loss": 0.4991,
|
|
"step": 6395
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"learning_rate": 1.39273041895726e-06,
|
|
"loss": 0.521,
|
|
"step": 6396
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3905784348117557e-06,
|
|
"loss": 0.4611,
|
|
"step": 6397
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3884279902843068e-06,
|
|
"loss": 0.3858,
|
|
"step": 6398
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3862790857594788e-06,
|
|
"loss": 0.4845,
|
|
"step": 6399
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.38413172162156e-06,
|
|
"loss": 0.4508,
|
|
"step": 6400
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3819858982545598e-06,
|
|
"loss": 0.4843,
|
|
"step": 6401
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3798416160422135e-06,
|
|
"loss": 0.4356,
|
|
"step": 6402
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3776988753679844e-06,
|
|
"loss": 0.4322,
|
|
"step": 6403
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3755576766150592e-06,
|
|
"loss": 0.4318,
|
|
"step": 6404
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.373418020166344e-06,
|
|
"loss": 0.4608,
|
|
"step": 6405
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3712799064044725e-06,
|
|
"loss": 0.4339,
|
|
"step": 6406
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3691433357118021e-06,
|
|
"loss": 0.4404,
|
|
"step": 6407
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3670083084704189e-06,
|
|
"loss": 0.4088,
|
|
"step": 6408
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3648748250621247e-06,
|
|
"loss": 0.4582,
|
|
"step": 6409
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3627428858684488e-06,
|
|
"loss": 0.4782,
|
|
"step": 6410
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3606124912706453e-06,
|
|
"loss": 0.4601,
|
|
"step": 6411
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3584836416496917e-06,
|
|
"loss": 0.4607,
|
|
"step": 6412
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3563563373862942e-06,
|
|
"loss": 0.5192,
|
|
"step": 6413
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3542305788608678e-06,
|
|
"loss": 0.4237,
|
|
"step": 6414
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.352106366453566e-06,
|
|
"loss": 0.4354,
|
|
"step": 6415
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3499837005442595e-06,
|
|
"loss": 0.488,
|
|
"step": 6416
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3478625815125468e-06,
|
|
"loss": 0.4569,
|
|
"step": 6417
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3457430097377421e-06,
|
|
"loss": 0.4626,
|
|
"step": 6418
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3436249855988881e-06,
|
|
"loss": 0.4823,
|
|
"step": 6419
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3415085094747493e-06,
|
|
"loss": 0.4695,
|
|
"step": 6420
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3393935817438185e-06,
|
|
"loss": 0.4701,
|
|
"step": 6421
|
|
},
|
|
{
|
|
"epoch": 2.51,
|
|
"learning_rate": 1.3372802027843034e-06,
|
|
"loss": 0.463,
|
|
"step": 6422
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.335168372974136e-06,
|
|
"loss": 0.4269,
|
|
"step": 6423
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3330580926909765e-06,
|
|
"loss": 0.4544,
|
|
"step": 6424
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3309493623122072e-06,
|
|
"loss": 0.4651,
|
|
"step": 6425
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.328842182214929e-06,
|
|
"loss": 0.4715,
|
|
"step": 6426
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3267365527759658e-06,
|
|
"loss": 0.458,
|
|
"step": 6427
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3246324743718674e-06,
|
|
"loss": 0.4443,
|
|
"step": 6428
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3225299473789076e-06,
|
|
"loss": 0.4964,
|
|
"step": 6429
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.320428972173079e-06,
|
|
"loss": 0.4507,
|
|
"step": 6430
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.318329549130094e-06,
|
|
"loss": 0.4097,
|
|
"step": 6431
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3162316786253938e-06,
|
|
"loss": 0.4177,
|
|
"step": 6432
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3141353610341434e-06,
|
|
"loss": 0.3974,
|
|
"step": 6433
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.312040596731221e-06,
|
|
"loss": 0.4493,
|
|
"step": 6434
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3099473860912325e-06,
|
|
"loss": 0.4734,
|
|
"step": 6435
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3078557294885064e-06,
|
|
"loss": 0.5244,
|
|
"step": 6436
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3057656272970932e-06,
|
|
"loss": 0.4225,
|
|
"step": 6437
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3036770798907662e-06,
|
|
"loss": 0.4498,
|
|
"step": 6438
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.3015900876430165e-06,
|
|
"loss": 0.4603,
|
|
"step": 6439
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2995046509270593e-06,
|
|
"loss": 0.4439,
|
|
"step": 6440
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2974207701158337e-06,
|
|
"loss": 0.4625,
|
|
"step": 6441
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2953384455820007e-06,
|
|
"loss": 0.4809,
|
|
"step": 6442
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2932576776979379e-06,
|
|
"loss": 0.4471,
|
|
"step": 6443
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2911784668357473e-06,
|
|
"loss": 0.4462,
|
|
"step": 6444
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2891008133672557e-06,
|
|
"loss": 0.4817,
|
|
"step": 6445
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2870247176640094e-06,
|
|
"loss": 0.4394,
|
|
"step": 6446
|
|
},
|
|
{
|
|
"epoch": 2.52,
|
|
"learning_rate": 1.2849501800972742e-06,
|
|
"loss": 0.4273,
|
|
"step": 6447
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2828772010380352e-06,
|
|
"loss": 0.4236,
|
|
"step": 6448
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2808057808570052e-06,
|
|
"loss": 0.4092,
|
|
"step": 6449
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2787359199246175e-06,
|
|
"loss": 0.5541,
|
|
"step": 6450
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2766676186110193e-06,
|
|
"loss": 0.4203,
|
|
"step": 6451
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2746008772860885e-06,
|
|
"loss": 0.4419,
|
|
"step": 6452
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2725356963194146e-06,
|
|
"loss": 0.4357,
|
|
"step": 6453
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2704720760803168e-06,
|
|
"loss": 0.4339,
|
|
"step": 6454
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2684100169378278e-06,
|
|
"loss": 0.4558,
|
|
"step": 6455
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2663495192607079e-06,
|
|
"loss": 0.4559,
|
|
"step": 6456
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2642905834174313e-06,
|
|
"loss": 0.4094,
|
|
"step": 6457
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2622332097762003e-06,
|
|
"loss": 0.4457,
|
|
"step": 6458
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2601773987049281e-06,
|
|
"loss": 0.5044,
|
|
"step": 6459
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.258123150571261e-06,
|
|
"loss": 0.4656,
|
|
"step": 6460
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2560704657425538e-06,
|
|
"loss": 0.4174,
|
|
"step": 6461
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2540193445858883e-06,
|
|
"loss": 0.4129,
|
|
"step": 6462
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.251969787468068e-06,
|
|
"loss": 0.5352,
|
|
"step": 6463
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2499217947556097e-06,
|
|
"loss": 0.4668,
|
|
"step": 6464
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2478753668147603e-06,
|
|
"loss": 0.4637,
|
|
"step": 6465
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.245830504011476e-06,
|
|
"loss": 0.4408,
|
|
"step": 6466
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.243787206711442e-06,
|
|
"loss": 0.4866,
|
|
"step": 6467
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2417454752800573e-06,
|
|
"loss": 0.4116,
|
|
"step": 6468
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2397053100824463e-06,
|
|
"loss": 0.4376,
|
|
"step": 6469
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2376667114834484e-06,
|
|
"loss": 0.4948,
|
|
"step": 6470
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2356296798476274e-06,
|
|
"loss": 0.4766,
|
|
"step": 6471
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.233594215539261e-06,
|
|
"loss": 0.4061,
|
|
"step": 6472
|
|
},
|
|
{
|
|
"epoch": 2.53,
|
|
"learning_rate": 1.2315603189223545e-06,
|
|
"loss": 0.4548,
|
|
"step": 6473
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2295279903606238e-06,
|
|
"loss": 0.3883,
|
|
"step": 6474
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2274972302175136e-06,
|
|
"loss": 0.4675,
|
|
"step": 6475
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2254680388561801e-06,
|
|
"loss": 0.4695,
|
|
"step": 6476
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2234404166395032e-06,
|
|
"loss": 0.4502,
|
|
"step": 6477
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2214143639300847e-06,
|
|
"loss": 0.4597,
|
|
"step": 6478
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2193898810902395e-06,
|
|
"loss": 0.4757,
|
|
"step": 6479
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2173669684820033e-06,
|
|
"loss": 0.4624,
|
|
"step": 6480
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2153456264671337e-06,
|
|
"loss": 0.4363,
|
|
"step": 6481
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2133258554071103e-06,
|
|
"loss": 0.4614,
|
|
"step": 6482
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2113076556631231e-06,
|
|
"loss": 0.4517,
|
|
"step": 6483
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2092910275960846e-06,
|
|
"loss": 0.4467,
|
|
"step": 6484
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2072759715666293e-06,
|
|
"loss": 0.4507,
|
|
"step": 6485
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2052624879351105e-06,
|
|
"loss": 0.4819,
|
|
"step": 6486
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2032505770615943e-06,
|
|
"loss": 0.5009,
|
|
"step": 6487
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.2012402393058742e-06,
|
|
"loss": 0.4506,
|
|
"step": 6488
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1992314750274526e-06,
|
|
"loss": 0.4075,
|
|
"step": 6489
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1972242845855608e-06,
|
|
"loss": 0.5057,
|
|
"step": 6490
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1952186683391386e-06,
|
|
"loss": 0.4358,
|
|
"step": 6491
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1932146266468547e-06,
|
|
"loss": 0.4609,
|
|
"step": 6492
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1912121598670844e-06,
|
|
"loss": 0.4433,
|
|
"step": 6493
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1892112683579326e-06,
|
|
"loss": 0.4903,
|
|
"step": 6494
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1872119524772175e-06,
|
|
"loss": 0.5057,
|
|
"step": 6495
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1852142125824738e-06,
|
|
"loss": 0.3957,
|
|
"step": 6496
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.1832180490309553e-06,
|
|
"loss": 0.4172,
|
|
"step": 6497
|
|
},
|
|
{
|
|
"epoch": 2.54,
|
|
"learning_rate": 1.181223462179636e-06,
|
|
"loss": 0.4375,
|
|
"step": 6498
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1792304523852094e-06,
|
|
"loss": 0.4966,
|
|
"step": 6499
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1772390200040817e-06,
|
|
"loss": 0.4526,
|
|
"step": 6500
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.175249165392377e-06,
|
|
"loss": 0.4612,
|
|
"step": 6501
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1732608889059437e-06,
|
|
"loss": 0.406,
|
|
"step": 6502
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1712741909003444e-06,
|
|
"loss": 0.4575,
|
|
"step": 6503
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1692890717308592e-06,
|
|
"loss": 0.4392,
|
|
"step": 6504
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1673055317524817e-06,
|
|
"loss": 0.5225,
|
|
"step": 6505
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1653235713199306e-06,
|
|
"loss": 0.4771,
|
|
"step": 6506
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.163343190787638e-06,
|
|
"loss": 0.5013,
|
|
"step": 6507
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1613643905097593e-06,
|
|
"loss": 0.5072,
|
|
"step": 6508
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1593871708401528e-06,
|
|
"loss": 0.4414,
|
|
"step": 6509
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1574115321324087e-06,
|
|
"loss": 0.4721,
|
|
"step": 6510
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1554374747398288e-06,
|
|
"loss": 0.4605,
|
|
"step": 6511
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1534649990154357e-06,
|
|
"loss": 0.5141,
|
|
"step": 6512
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1514941053119622e-06,
|
|
"loss": 0.4728,
|
|
"step": 6513
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1495247939818622e-06,
|
|
"loss": 0.5003,
|
|
"step": 6514
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1475570653773073e-06,
|
|
"loss": 0.4781,
|
|
"step": 6515
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.145590919850187e-06,
|
|
"loss": 0.4484,
|
|
"step": 6516
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1436263577521057e-06,
|
|
"loss": 0.4337,
|
|
"step": 6517
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1416633794343812e-06,
|
|
"loss": 0.4503,
|
|
"step": 6518
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.139701985248055e-06,
|
|
"loss": 0.4858,
|
|
"step": 6519
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1377421755438834e-06,
|
|
"loss": 0.5007,
|
|
"step": 6520
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1357839506723356e-06,
|
|
"loss": 0.3928,
|
|
"step": 6521
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.1338273109835995e-06,
|
|
"loss": 0.4706,
|
|
"step": 6522
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.131872256827581e-06,
|
|
"loss": 0.472,
|
|
"step": 6523
|
|
},
|
|
{
|
|
"epoch": 2.55,
|
|
"learning_rate": 1.129918788553902e-06,
|
|
"loss": 0.4476,
|
|
"step": 6524
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1279669065119004e-06,
|
|
"loss": 0.4685,
|
|
"step": 6525
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1260166110506277e-06,
|
|
"loss": 0.4282,
|
|
"step": 6526
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1240679025188551e-06,
|
|
"loss": 0.4532,
|
|
"step": 6527
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1221207812650691e-06,
|
|
"loss": 0.4539,
|
|
"step": 6528
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1201752476374782e-06,
|
|
"loss": 0.4854,
|
|
"step": 6529
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1182313019839909e-06,
|
|
"loss": 0.4248,
|
|
"step": 6530
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1162889446522462e-06,
|
|
"loss": 0.4422,
|
|
"step": 6531
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1143481759895958e-06,
|
|
"loss": 0.4307,
|
|
"step": 6532
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1124089963431084e-06,
|
|
"loss": 0.5206,
|
|
"step": 6533
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1104714060595623e-06,
|
|
"loss": 0.5158,
|
|
"step": 6534
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1085354054854569e-06,
|
|
"loss": 0.4557,
|
|
"step": 6535
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1066009949670053e-06,
|
|
"loss": 0.4331,
|
|
"step": 6536
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1046681748501409e-06,
|
|
"loss": 0.4524,
|
|
"step": 6537
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1027369454805058e-06,
|
|
"loss": 0.4907,
|
|
"step": 6538
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.1008073072034586e-06,
|
|
"loss": 0.4308,
|
|
"step": 6539
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0988792603640786e-06,
|
|
"loss": 0.48,
|
|
"step": 6540
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0969528053071577e-06,
|
|
"loss": 0.3984,
|
|
"step": 6541
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0950279423772014e-06,
|
|
"loss": 0.4516,
|
|
"step": 6542
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0931046719184313e-06,
|
|
"loss": 0.5118,
|
|
"step": 6543
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0911829942747842e-06,
|
|
"loss": 0.4557,
|
|
"step": 6544
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0892629097899176e-06,
|
|
"loss": 0.4892,
|
|
"step": 6545
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0873444188071925e-06,
|
|
"loss": 0.4036,
|
|
"step": 6546
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0854275216696964e-06,
|
|
"loss": 0.4416,
|
|
"step": 6547
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.083512218720224e-06,
|
|
"loss": 0.3924,
|
|
"step": 6548
|
|
},
|
|
{
|
|
"epoch": 2.56,
|
|
"learning_rate": 1.0815985103012905e-06,
|
|
"loss": 0.4478,
|
|
"step": 6549
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0796863967551208e-06,
|
|
"loss": 0.4536,
|
|
"step": 6550
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0777758784236592e-06,
|
|
"loss": 0.4551,
|
|
"step": 6551
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.075866955648559e-06,
|
|
"loss": 0.4461,
|
|
"step": 6552
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.073959628771195e-06,
|
|
"loss": 0.4498,
|
|
"step": 6553
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0720538981326557e-06,
|
|
"loss": 0.4291,
|
|
"step": 6554
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0701497640737378e-06,
|
|
"loss": 0.427,
|
|
"step": 6555
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0682472269349564e-06,
|
|
"loss": 0.5257,
|
|
"step": 6556
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.066346287056541e-06,
|
|
"loss": 0.4423,
|
|
"step": 6557
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.064446944778441e-06,
|
|
"loss": 0.4442,
|
|
"step": 6558
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0625492004403072e-06,
|
|
"loss": 0.4167,
|
|
"step": 6559
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0606530543815185e-06,
|
|
"loss": 0.425,
|
|
"step": 6560
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0587585069411565e-06,
|
|
"loss": 0.4501,
|
|
"step": 6561
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0568655584580267e-06,
|
|
"loss": 0.4247,
|
|
"step": 6562
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0549742092706395e-06,
|
|
"loss": 0.4648,
|
|
"step": 6563
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0530844597172275e-06,
|
|
"loss": 0.4403,
|
|
"step": 6564
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0511963101357315e-06,
|
|
"loss": 0.4105,
|
|
"step": 6565
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0493097608638092e-06,
|
|
"loss": 0.4207,
|
|
"step": 6566
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0474248122388298e-06,
|
|
"loss": 0.386,
|
|
"step": 6567
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0455414645978812e-06,
|
|
"loss": 0.3821,
|
|
"step": 6568
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0436597182777553e-06,
|
|
"loss": 0.4095,
|
|
"step": 6569
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0417795736149705e-06,
|
|
"loss": 0.4535,
|
|
"step": 6570
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0399010309457459e-06,
|
|
"loss": 0.4606,
|
|
"step": 6571
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0380240906060235e-06,
|
|
"loss": 0.4353,
|
|
"step": 6572
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0361487529314573e-06,
|
|
"loss": 0.487,
|
|
"step": 6573
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0342750182574112e-06,
|
|
"loss": 0.4315,
|
|
"step": 6574
|
|
},
|
|
{
|
|
"epoch": 2.57,
|
|
"learning_rate": 1.0324028869189606e-06,
|
|
"loss": 0.4134,
|
|
"step": 6575
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.030532359250901e-06,
|
|
"loss": 0.4464,
|
|
"step": 6576
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.02866343558774e-06,
|
|
"loss": 0.4677,
|
|
"step": 6577
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0267961162636919e-06,
|
|
"loss": 0.4162,
|
|
"step": 6578
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0249304016126915e-06,
|
|
"loss": 0.4653,
|
|
"step": 6579
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.023066291968381e-06,
|
|
"loss": 0.4721,
|
|
"step": 6580
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0212037876641201e-06,
|
|
"loss": 0.4447,
|
|
"step": 6581
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.019342889032977e-06,
|
|
"loss": 0.4937,
|
|
"step": 6582
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.01748359640774e-06,
|
|
"loss": 0.4655,
|
|
"step": 6583
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0156259101208987e-06,
|
|
"loss": 0.4016,
|
|
"step": 6584
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0137698305046662e-06,
|
|
"loss": 0.4379,
|
|
"step": 6585
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0119153578909657e-06,
|
|
"loss": 0.4271,
|
|
"step": 6586
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0100624926114288e-06,
|
|
"loss": 0.5048,
|
|
"step": 6587
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0082112349974017e-06,
|
|
"loss": 0.4539,
|
|
"step": 6588
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0063615853799435e-06,
|
|
"loss": 0.4281,
|
|
"step": 6589
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0045135440898313e-06,
|
|
"loss": 0.4273,
|
|
"step": 6590
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.0026671114575437e-06,
|
|
"loss": 0.4429,
|
|
"step": 6591
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 1.000822287813278e-06,
|
|
"loss": 0.4883,
|
|
"step": 6592
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.989790734869443e-07,
|
|
"loss": 0.4268,
|
|
"step": 6593
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.971374688081648e-07,
|
|
"loss": 0.4382,
|
|
"step": 6594
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.952974741062704e-07,
|
|
"loss": 0.4859,
|
|
"step": 6595
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.934590897103058e-07,
|
|
"loss": 0.4542,
|
|
"step": 6596
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.916223159490302e-07,
|
|
"loss": 0.461,
|
|
"step": 6597
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.897871531509107e-07,
|
|
"loss": 0.4455,
|
|
"step": 6598
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.87953601644135e-07,
|
|
"loss": 0.3895,
|
|
"step": 6599
|
|
},
|
|
{
|
|
"epoch": 2.58,
|
|
"learning_rate": 9.86121661756586e-07,
|
|
"loss": 0.4543,
|
|
"step": 6600
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.842913338158732e-07,
|
|
"loss": 0.4251,
|
|
"step": 6601
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.824626181493146e-07,
|
|
"loss": 0.5291,
|
|
"step": 6602
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.806355150839386e-07,
|
|
"loss": 0.4456,
|
|
"step": 6603
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.788100249464827e-07,
|
|
"loss": 0.4854,
|
|
"step": 6604
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.76986148063398e-07,
|
|
"loss": 0.4513,
|
|
"step": 6605
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.75163884760849e-07,
|
|
"loss": 0.423,
|
|
"step": 6606
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.733432353647109e-07,
|
|
"loss": 0.4656,
|
|
"step": 6607
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.715242002005688e-07,
|
|
"loss": 0.4385,
|
|
"step": 6608
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.697067795937164e-07,
|
|
"loss": 0.4069,
|
|
"step": 6609
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.678909738691645e-07,
|
|
"loss": 0.4293,
|
|
"step": 6610
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.660767833516348e-07,
|
|
"loss": 0.4918,
|
|
"step": 6611
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.642642083655563e-07,
|
|
"loss": 0.4309,
|
|
"step": 6612
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.62453249235068e-07,
|
|
"loss": 0.391,
|
|
"step": 6613
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.606439062840256e-07,
|
|
"loss": 0.3874,
|
|
"step": 6614
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.588361798359947e-07,
|
|
"loss": 0.3755,
|
|
"step": 6615
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.570300702142465e-07,
|
|
"loss": 0.5254,
|
|
"step": 6616
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.552255777417674e-07,
|
|
"loss": 0.4208,
|
|
"step": 6617
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.534227027412535e-07,
|
|
"loss": 0.4355,
|
|
"step": 6618
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.516214455351136e-07,
|
|
"loss": 0.4017,
|
|
"step": 6619
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.498218064454656e-07,
|
|
"loss": 0.4859,
|
|
"step": 6620
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.480237857941355e-07,
|
|
"loss": 0.4205,
|
|
"step": 6621
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.462273839026625e-07,
|
|
"loss": 0.4795,
|
|
"step": 6622
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.444326010922989e-07,
|
|
"loss": 0.3943,
|
|
"step": 6623
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.426394376840054e-07,
|
|
"loss": 0.4552,
|
|
"step": 6624
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.408478939984467e-07,
|
|
"loss": 0.5115,
|
|
"step": 6625
|
|
},
|
|
{
|
|
"epoch": 2.59,
|
|
"learning_rate": 9.390579703560077e-07,
|
|
"loss": 0.4437,
|
|
"step": 6626
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.372696670767789e-07,
|
|
"loss": 0.4667,
|
|
"step": 6627
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.354829844805635e-07,
|
|
"loss": 0.3943,
|
|
"step": 6628
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.336979228868703e-07,
|
|
"loss": 0.4729,
|
|
"step": 6629
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.319144826149196e-07,
|
|
"loss": 0.4188,
|
|
"step": 6630
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.30132663983645e-07,
|
|
"loss": 0.46,
|
|
"step": 6631
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.283524673116906e-07,
|
|
"loss": 0.5181,
|
|
"step": 6632
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.265738929174051e-07,
|
|
"loss": 0.4353,
|
|
"step": 6633
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.247969411188484e-07,
|
|
"loss": 0.4793,
|
|
"step": 6634
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.230216122337932e-07,
|
|
"loss": 0.4354,
|
|
"step": 6635
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.212479065797231e-07,
|
|
"loss": 0.4657,
|
|
"step": 6636
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.194758244738278e-07,
|
|
"loss": 0.4646,
|
|
"step": 6637
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.177053662330037e-07,
|
|
"loss": 0.4524,
|
|
"step": 6638
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.159365321738655e-07,
|
|
"loss": 0.4174,
|
|
"step": 6639
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.141693226127324e-07,
|
|
"loss": 0.4189,
|
|
"step": 6640
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.124037378656303e-07,
|
|
"loss": 0.4077,
|
|
"step": 6641
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.106397782483011e-07,
|
|
"loss": 0.4619,
|
|
"step": 6642
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.088774440761905e-07,
|
|
"loss": 0.4319,
|
|
"step": 6643
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.071167356644594e-07,
|
|
"loss": 0.4962,
|
|
"step": 6644
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.053576533279685e-07,
|
|
"loss": 0.4796,
|
|
"step": 6645
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.036001973812991e-07,
|
|
"loss": 0.5206,
|
|
"step": 6646
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.018443681387312e-07,
|
|
"loss": 0.4296,
|
|
"step": 6647
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 9.000901659142614e-07,
|
|
"loss": 0.472,
|
|
"step": 6648
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 8.983375910215941e-07,
|
|
"loss": 0.4608,
|
|
"step": 6649
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 8.9658664377414e-07,
|
|
"loss": 0.4739,
|
|
"step": 6650
|
|
},
|
|
{
|
|
"epoch": 2.6,
|
|
"learning_rate": 8.948373244850173e-07,
|
|
"loss": 0.4606,
|
|
"step": 6651
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.93089633467058e-07,
|
|
"loss": 0.4321,
|
|
"step": 6652
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.913435710328034e-07,
|
|
"loss": 0.4194,
|
|
"step": 6653
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.895991374944967e-07,
|
|
"loss": 0.3861,
|
|
"step": 6654
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.878563331640965e-07,
|
|
"loss": 0.4073,
|
|
"step": 6655
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.861151583532657e-07,
|
|
"loss": 0.4303,
|
|
"step": 6656
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.843756133733805e-07,
|
|
"loss": 0.4212,
|
|
"step": 6657
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.826376985355178e-07,
|
|
"loss": 0.5089,
|
|
"step": 6658
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.809014141504734e-07,
|
|
"loss": 0.4205,
|
|
"step": 6659
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.791667605287424e-07,
|
|
"loss": 0.3992,
|
|
"step": 6660
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.774337379805342e-07,
|
|
"loss": 0.4422,
|
|
"step": 6661
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.757023468157622e-07,
|
|
"loss": 0.4573,
|
|
"step": 6662
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.739725873440519e-07,
|
|
"loss": 0.4357,
|
|
"step": 6663
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.722444598747337e-07,
|
|
"loss": 0.5088,
|
|
"step": 6664
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.705179647168494e-07,
|
|
"loss": 0.4297,
|
|
"step": 6665
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.687931021791452e-07,
|
|
"loss": 0.4643,
|
|
"step": 6666
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.670698725700776e-07,
|
|
"loss": 0.4458,
|
|
"step": 6667
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.653482761978138e-07,
|
|
"loss": 0.4182,
|
|
"step": 6668
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.636283133702228e-07,
|
|
"loss": 0.4838,
|
|
"step": 6669
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.619099843948842e-07,
|
|
"loss": 0.4449,
|
|
"step": 6670
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.601932895790877e-07,
|
|
"loss": 0.4505,
|
|
"step": 6671
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.584782292298289e-07,
|
|
"loss": 0.3654,
|
|
"step": 6672
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.56764803653809e-07,
|
|
"loss": 0.4041,
|
|
"step": 6673
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.550530131574419e-07,
|
|
"loss": 0.4292,
|
|
"step": 6674
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.533428580468428e-07,
|
|
"loss": 0.522,
|
|
"step": 6675
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.516343386278414e-07,
|
|
"loss": 0.42,
|
|
"step": 6676
|
|
},
|
|
{
|
|
"epoch": 2.61,
|
|
"learning_rate": 8.499274552059678e-07,
|
|
"loss": 0.4498,
|
|
"step": 6677
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.482222080864666e-07,
|
|
"loss": 0.4141,
|
|
"step": 6678
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.465185975742818e-07,
|
|
"loss": 0.386,
|
|
"step": 6679
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.448166239740718e-07,
|
|
"loss": 0.4418,
|
|
"step": 6680
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.431162875902021e-07,
|
|
"loss": 0.4876,
|
|
"step": 6681
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.414175887267385e-07,
|
|
"loss": 0.4805,
|
|
"step": 6682
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.397205276874598e-07,
|
|
"loss": 0.4528,
|
|
"step": 6683
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.380251047758492e-07,
|
|
"loss": 0.4632,
|
|
"step": 6684
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.363313202951029e-07,
|
|
"loss": 0.4817,
|
|
"step": 6685
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.346391745481153e-07,
|
|
"loss": 0.4379,
|
|
"step": 6686
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.32948667837491e-07,
|
|
"loss": 0.4144,
|
|
"step": 6687
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.312598004655447e-07,
|
|
"loss": 0.4034,
|
|
"step": 6688
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.295725727342962e-07,
|
|
"loss": 0.4874,
|
|
"step": 6689
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.278869849454718e-07,
|
|
"loss": 0.4369,
|
|
"step": 6690
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.262030374005003e-07,
|
|
"loss": 0.451,
|
|
"step": 6691
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.245207304005232e-07,
|
|
"loss": 0.4459,
|
|
"step": 6692
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.228400642463874e-07,
|
|
"loss": 0.4496,
|
|
"step": 6693
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.211610392386482e-07,
|
|
"loss": 0.5076,
|
|
"step": 6694
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.194836556775588e-07,
|
|
"loss": 0.3668,
|
|
"step": 6695
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.178079138630868e-07,
|
|
"loss": 0.4489,
|
|
"step": 6696
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.161338140949049e-07,
|
|
"loss": 0.4327,
|
|
"step": 6697
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.144613566723936e-07,
|
|
"loss": 0.4363,
|
|
"step": 6698
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.127905418946347e-07,
|
|
"loss": 0.4071,
|
|
"step": 6699
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.111213700604193e-07,
|
|
"loss": 0.4145,
|
|
"step": 6700
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.09453841468244e-07,
|
|
"loss": 0.4233,
|
|
"step": 6701
|
|
},
|
|
{
|
|
"epoch": 2.62,
|
|
"learning_rate": 8.077879564163138e-07,
|
|
"loss": 0.4626,
|
|
"step": 6702
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 8.061237152025381e-07,
|
|
"loss": 0.4822,
|
|
"step": 6703
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 8.044611181245299e-07,
|
|
"loss": 0.4917,
|
|
"step": 6704
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 8.028001654796114e-07,
|
|
"loss": 0.4639,
|
|
"step": 6705
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 8.011408575648117e-07,
|
|
"loss": 0.5089,
|
|
"step": 6706
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.994831946768622e-07,
|
|
"loss": 0.4184,
|
|
"step": 6707
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.978271771122015e-07,
|
|
"loss": 0.4571,
|
|
"step": 6708
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.961728051669737e-07,
|
|
"loss": 0.4956,
|
|
"step": 6709
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.945200791370334e-07,
|
|
"loss": 0.4563,
|
|
"step": 6710
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.928689993179328e-07,
|
|
"loss": 0.439,
|
|
"step": 6711
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.912195660049327e-07,
|
|
"loss": 0.4615,
|
|
"step": 6712
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.895717794930024e-07,
|
|
"loss": 0.4759,
|
|
"step": 6713
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.879256400768154e-07,
|
|
"loss": 0.4618,
|
|
"step": 6714
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.862811480507504e-07,
|
|
"loss": 0.453,
|
|
"step": 6715
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.846383037088867e-07,
|
|
"loss": 0.4217,
|
|
"step": 6716
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.829971073450171e-07,
|
|
"loss": 0.401,
|
|
"step": 6717
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.813575592526334e-07,
|
|
"loss": 0.5153,
|
|
"step": 6718
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.797196597249391e-07,
|
|
"loss": 0.4485,
|
|
"step": 6719
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.780834090548351e-07,
|
|
"loss": 0.4371,
|
|
"step": 6720
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.764488075349307e-07,
|
|
"loss": 0.4231,
|
|
"step": 6721
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.748158554575424e-07,
|
|
"loss": 0.4581,
|
|
"step": 6722
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.73184553114692e-07,
|
|
"loss": 0.431,
|
|
"step": 6723
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.715549007981026e-07,
|
|
"loss": 0.4209,
|
|
"step": 6724
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.699268987992015e-07,
|
|
"loss": 0.4514,
|
|
"step": 6725
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.683005474091265e-07,
|
|
"loss": 0.4073,
|
|
"step": 6726
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.666758469187174e-07,
|
|
"loss": 0.4477,
|
|
"step": 6727
|
|
},
|
|
{
|
|
"epoch": 2.63,
|
|
"learning_rate": 7.650527976185174e-07,
|
|
"loss": 0.5686,
|
|
"step": 6728
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.634313997987741e-07,
|
|
"loss": 0.4171,
|
|
"step": 6729
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.618116537494424e-07,
|
|
"loss": 0.4609,
|
|
"step": 6730
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.601935597601828e-07,
|
|
"loss": 0.4093,
|
|
"step": 6731
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.585771181203561e-07,
|
|
"loss": 0.4922,
|
|
"step": 6732
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.569623291190276e-07,
|
|
"loss": 0.4514,
|
|
"step": 6733
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.553491930449719e-07,
|
|
"loss": 0.4266,
|
|
"step": 6734
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.537377101866661e-07,
|
|
"loss": 0.4746,
|
|
"step": 6735
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.521278808322863e-07,
|
|
"loss": 0.4356,
|
|
"step": 6736
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.505197052697233e-07,
|
|
"loss": 0.509,
|
|
"step": 6737
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.489131837865604e-07,
|
|
"loss": 0.4416,
|
|
"step": 6738
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.473083166700946e-07,
|
|
"loss": 0.4623,
|
|
"step": 6739
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.457051042073238e-07,
|
|
"loss": 0.4713,
|
|
"step": 6740
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.441035466849489e-07,
|
|
"loss": 0.4537,
|
|
"step": 6741
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.42503644389373e-07,
|
|
"loss": 0.4482,
|
|
"step": 6742
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.409053976067071e-07,
|
|
"loss": 0.4141,
|
|
"step": 6743
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.393088066227683e-07,
|
|
"loss": 0.4909,
|
|
"step": 6744
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.377138717230703e-07,
|
|
"loss": 0.4754,
|
|
"step": 6745
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.361205931928339e-07,
|
|
"loss": 0.392,
|
|
"step": 6746
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.345289713169856e-07,
|
|
"loss": 0.4794,
|
|
"step": 6747
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.329390063801556e-07,
|
|
"loss": 0.432,
|
|
"step": 6748
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.313506986666729e-07,
|
|
"loss": 0.4912,
|
|
"step": 6749
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.297640484605783e-07,
|
|
"loss": 0.4339,
|
|
"step": 6750
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.281790560456081e-07,
|
|
"loss": 0.4476,
|
|
"step": 6751
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.265957217052078e-07,
|
|
"loss": 0.4346,
|
|
"step": 6752
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.250140457225208e-07,
|
|
"loss": 0.4455,
|
|
"step": 6753
|
|
},
|
|
{
|
|
"epoch": 2.64,
|
|
"learning_rate": 7.23434028380402e-07,
|
|
"loss": 0.4656,
|
|
"step": 6754
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.21855669961401e-07,
|
|
"loss": 0.4153,
|
|
"step": 6755
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.202789707477775e-07,
|
|
"loss": 0.483,
|
|
"step": 6756
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.187039310214894e-07,
|
|
"loss": 0.4437,
|
|
"step": 6757
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.171305510642024e-07,
|
|
"loss": 0.4851,
|
|
"step": 6758
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.155588311572803e-07,
|
|
"loss": 0.4701,
|
|
"step": 6759
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.139887715817973e-07,
|
|
"loss": 0.4651,
|
|
"step": 6760
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.124203726185219e-07,
|
|
"loss": 0.455,
|
|
"step": 6761
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.108536345479323e-07,
|
|
"loss": 0.4761,
|
|
"step": 6762
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.092885576502073e-07,
|
|
"loss": 0.3996,
|
|
"step": 6763
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.077251422052278e-07,
|
|
"loss": 0.4059,
|
|
"step": 6764
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.0616338849258e-07,
|
|
"loss": 0.473,
|
|
"step": 6765
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.046032967915484e-07,
|
|
"loss": 0.4929,
|
|
"step": 6766
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.030448673811285e-07,
|
|
"loss": 0.4192,
|
|
"step": 6767
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 7.014881005400077e-07,
|
|
"loss": 0.395,
|
|
"step": 6768
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.999329965465862e-07,
|
|
"loss": 0.4179,
|
|
"step": 6769
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.983795556789585e-07,
|
|
"loss": 0.4959,
|
|
"step": 6770
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.968277782149302e-07,
|
|
"loss": 0.4279,
|
|
"step": 6771
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.952776644320003e-07,
|
|
"loss": 0.4857,
|
|
"step": 6772
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.937292146073782e-07,
|
|
"loss": 0.497,
|
|
"step": 6773
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.921824290179679e-07,
|
|
"loss": 0.4341,
|
|
"step": 6774
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.90637307940385e-07,
|
|
"loss": 0.4246,
|
|
"step": 6775
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.890938516509416e-07,
|
|
"loss": 0.4377,
|
|
"step": 6776
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.875520604256514e-07,
|
|
"loss": 0.3994,
|
|
"step": 6777
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.860119345402327e-07,
|
|
"loss": 0.4625,
|
|
"step": 6778
|
|
},
|
|
{
|
|
"epoch": 2.65,
|
|
"learning_rate": 6.844734742701053e-07,
|
|
"loss": 0.4434,
|
|
"step": 6779
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.829366798903925e-07,
|
|
"loss": 0.4485,
|
|
"step": 6780
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.814015516759176e-07,
|
|
"loss": 0.409,
|
|
"step": 6781
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.798680899012044e-07,
|
|
"loss": 0.451,
|
|
"step": 6782
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.783362948404837e-07,
|
|
"loss": 0.5115,
|
|
"step": 6783
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.768061667676873e-07,
|
|
"loss": 0.4575,
|
|
"step": 6784
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.752777059564431e-07,
|
|
"loss": 0.4701,
|
|
"step": 6785
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.737509126800856e-07,
|
|
"loss": 0.459,
|
|
"step": 6786
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.722257872116522e-07,
|
|
"loss": 0.4018,
|
|
"step": 6787
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.707023298238791e-07,
|
|
"loss": 0.4062,
|
|
"step": 6788
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.691805407892071e-07,
|
|
"loss": 0.432,
|
|
"step": 6789
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.676604203797755e-07,
|
|
"loss": 0.4452,
|
|
"step": 6790
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.661419688674243e-07,
|
|
"loss": 0.4821,
|
|
"step": 6791
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.646251865236997e-07,
|
|
"loss": 0.473,
|
|
"step": 6792
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.631100736198492e-07,
|
|
"loss": 0.3908,
|
|
"step": 6793
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.615966304268174e-07,
|
|
"loss": 0.4317,
|
|
"step": 6794
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.600848572152497e-07,
|
|
"loss": 0.4499,
|
|
"step": 6795
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.585747542554987e-07,
|
|
"loss": 0.4327,
|
|
"step": 6796
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.570663218176177e-07,
|
|
"loss": 0.4352,
|
|
"step": 6797
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.555595601713549e-07,
|
|
"loss": 0.4628,
|
|
"step": 6798
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.540544695861628e-07,
|
|
"loss": 0.4543,
|
|
"step": 6799
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.525510503311994e-07,
|
|
"loss": 0.4906,
|
|
"step": 6800
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.510493026753195e-07,
|
|
"loss": 0.4654,
|
|
"step": 6801
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.495492268870795e-07,
|
|
"loss": 0.4637,
|
|
"step": 6802
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.480508232347349e-07,
|
|
"loss": 0.4659,
|
|
"step": 6803
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.465540919862457e-07,
|
|
"loss": 0.4557,
|
|
"step": 6804
|
|
},
|
|
{
|
|
"epoch": 2.66,
|
|
"learning_rate": 6.450590334092744e-07,
|
|
"loss": 0.4721,
|
|
"step": 6805
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.435656477711793e-07,
|
|
"loss": 0.4494,
|
|
"step": 6806
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.420739353390193e-07,
|
|
"loss": 0.4601,
|
|
"step": 6807
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.405838963795597e-07,
|
|
"loss": 0.4673,
|
|
"step": 6808
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.390955311592617e-07,
|
|
"loss": 0.4846,
|
|
"step": 6809
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.376088399442926e-07,
|
|
"loss": 0.4327,
|
|
"step": 6810
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.361238230005107e-07,
|
|
"loss": 0.3743,
|
|
"step": 6811
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.346404805934836e-07,
|
|
"loss": 0.4055,
|
|
"step": 6812
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.331588129884769e-07,
|
|
"loss": 0.4994,
|
|
"step": 6813
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.316788204504565e-07,
|
|
"loss": 0.4159,
|
|
"step": 6814
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.302005032440894e-07,
|
|
"loss": 0.4435,
|
|
"step": 6815
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.287238616337388e-07,
|
|
"loss": 0.4673,
|
|
"step": 6816
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.272488958834733e-07,
|
|
"loss": 0.4672,
|
|
"step": 6817
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.25775606257063e-07,
|
|
"loss": 0.4705,
|
|
"step": 6818
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.243039930179739e-07,
|
|
"loss": 0.5075,
|
|
"step": 6819
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.228340564293711e-07,
|
|
"loss": 0.4363,
|
|
"step": 6820
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.213657967541242e-07,
|
|
"loss": 0.4746,
|
|
"step": 6821
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.198992142548033e-07,
|
|
"loss": 0.4502,
|
|
"step": 6822
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.184343091936751e-07,
|
|
"loss": 0.4336,
|
|
"step": 6823
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.169710818327057e-07,
|
|
"loss": 0.496,
|
|
"step": 6824
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.155095324335658e-07,
|
|
"loss": 0.4096,
|
|
"step": 6825
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.140496612576241e-07,
|
|
"loss": 0.4457,
|
|
"step": 6826
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.125914685659451e-07,
|
|
"loss": 0.5514,
|
|
"step": 6827
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.111349546193002e-07,
|
|
"loss": 0.4658,
|
|
"step": 6828
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.096801196781543e-07,
|
|
"loss": 0.5095,
|
|
"step": 6829
|
|
},
|
|
{
|
|
"epoch": 2.67,
|
|
"learning_rate": 6.082269640026772e-07,
|
|
"loss": 0.4511,
|
|
"step": 6830
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 6.067754878527332e-07,
|
|
"loss": 0.4123,
|
|
"step": 6831
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 6.053256914878914e-07,
|
|
"loss": 0.4304,
|
|
"step": 6832
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 6.038775751674164e-07,
|
|
"loss": 0.3803,
|
|
"step": 6833
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 6.024311391502746e-07,
|
|
"loss": 0.4709,
|
|
"step": 6834
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 6.009863836951324e-07,
|
|
"loss": 0.4089,
|
|
"step": 6835
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.99543309060353e-07,
|
|
"loss": 0.4789,
|
|
"step": 6836
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.981019155040013e-07,
|
|
"loss": 0.436,
|
|
"step": 6837
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.966622032838398e-07,
|
|
"loss": 0.4692,
|
|
"step": 6838
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.952241726573338e-07,
|
|
"loss": 0.4078,
|
|
"step": 6839
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.93787823881643e-07,
|
|
"loss": 0.4295,
|
|
"step": 6840
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.92353157213632e-07,
|
|
"loss": 0.4725,
|
|
"step": 6841
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.90920172909858e-07,
|
|
"loss": 0.416,
|
|
"step": 6842
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.894888712265834e-07,
|
|
"loss": 0.4989,
|
|
"step": 6843
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.880592524197648e-07,
|
|
"loss": 0.447,
|
|
"step": 6844
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.866313167450622e-07,
|
|
"loss": 0.5252,
|
|
"step": 6845
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.8520506445783e-07,
|
|
"loss": 0.4495,
|
|
"step": 6846
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.837804958131277e-07,
|
|
"loss": 0.4098,
|
|
"step": 6847
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.82357611065707e-07,
|
|
"loss": 0.4417,
|
|
"step": 6848
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.809364104700244e-07,
|
|
"loss": 0.4192,
|
|
"step": 6849
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.795168942802287e-07,
|
|
"loss": 0.4449,
|
|
"step": 6850
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.780990627501748e-07,
|
|
"loss": 0.4131,
|
|
"step": 6851
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.766829161334098e-07,
|
|
"loss": 0.449,
|
|
"step": 6852
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.752684546831833e-07,
|
|
"loss": 0.4536,
|
|
"step": 6853
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.738556786524452e-07,
|
|
"loss": 0.4452,
|
|
"step": 6854
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.72444588293839e-07,
|
|
"loss": 0.408,
|
|
"step": 6855
|
|
},
|
|
{
|
|
"epoch": 2.68,
|
|
"learning_rate": 5.710351838597083e-07,
|
|
"loss": 0.4863,
|
|
"step": 6856
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.696274656020973e-07,
|
|
"loss": 0.4478,
|
|
"step": 6857
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.682214337727487e-07,
|
|
"loss": 0.439,
|
|
"step": 6858
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.668170886231006e-07,
|
|
"loss": 0.4548,
|
|
"step": 6859
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.65414430404293e-07,
|
|
"loss": 0.438,
|
|
"step": 6860
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.640134593671598e-07,
|
|
"loss": 0.4187,
|
|
"step": 6861
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.626141757622394e-07,
|
|
"loss": 0.42,
|
|
"step": 6862
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.612165798397607e-07,
|
|
"loss": 0.383,
|
|
"step": 6863
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.59820671849659e-07,
|
|
"loss": 0.4891,
|
|
"step": 6864
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.584264520415606e-07,
|
|
"loss": 0.4322,
|
|
"step": 6865
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.570339206647956e-07,
|
|
"loss": 0.4511,
|
|
"step": 6866
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.556430779683885e-07,
|
|
"loss": 0.4483,
|
|
"step": 6867
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.542539242010625e-07,
|
|
"loss": 0.4728,
|
|
"step": 6868
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.528664596112388e-07,
|
|
"loss": 0.4415,
|
|
"step": 6869
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.514806844470377e-07,
|
|
"loss": 0.3992,
|
|
"step": 6870
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.50096598956279e-07,
|
|
"loss": 0.5035,
|
|
"step": 6871
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.487142033864746e-07,
|
|
"loss": 0.4514,
|
|
"step": 6872
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.473334979848366e-07,
|
|
"loss": 0.456,
|
|
"step": 6873
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.459544829982788e-07,
|
|
"loss": 0.3932,
|
|
"step": 6874
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.445771586734094e-07,
|
|
"loss": 0.4334,
|
|
"step": 6875
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.432015252565337e-07,
|
|
"loss": 0.4676,
|
|
"step": 6876
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.418275829936537e-07,
|
|
"loss": 0.4534,
|
|
"step": 6877
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.404553321304728e-07,
|
|
"loss": 0.4251,
|
|
"step": 6878
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.390847729123905e-07,
|
|
"loss": 0.398,
|
|
"step": 6879
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.377159055845028e-07,
|
|
"loss": 0.4335,
|
|
"step": 6880
|
|
},
|
|
{
|
|
"epoch": 2.69,
|
|
"learning_rate": 5.363487303916015e-07,
|
|
"loss": 0.4458,
|
|
"step": 6881
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.34983247578178e-07,
|
|
"loss": 0.4198,
|
|
"step": 6882
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.336194573884235e-07,
|
|
"loss": 0.4404,
|
|
"step": 6883
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.322573600662228e-07,
|
|
"loss": 0.4439,
|
|
"step": 6884
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.308969558551579e-07,
|
|
"loss": 0.4435,
|
|
"step": 6885
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.295382449985099e-07,
|
|
"loss": 0.4891,
|
|
"step": 6886
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.281812277392551e-07,
|
|
"loss": 0.5361,
|
|
"step": 6887
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.268259043200707e-07,
|
|
"loss": 0.5136,
|
|
"step": 6888
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.254722749833263e-07,
|
|
"loss": 0.468,
|
|
"step": 6889
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.241203399710903e-07,
|
|
"loss": 0.5014,
|
|
"step": 6890
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.227700995251306e-07,
|
|
"loss": 0.4801,
|
|
"step": 6891
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.214215538869083e-07,
|
|
"loss": 0.4246,
|
|
"step": 6892
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.20074703297585e-07,
|
|
"loss": 0.4,
|
|
"step": 6893
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.187295479980136e-07,
|
|
"loss": 0.4001,
|
|
"step": 6894
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.173860882287507e-07,
|
|
"loss": 0.4193,
|
|
"step": 6895
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.160443242300462e-07,
|
|
"loss": 0.464,
|
|
"step": 6896
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.147042562418458e-07,
|
|
"loss": 0.4248,
|
|
"step": 6897
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.133658845037926e-07,
|
|
"loss": 0.4439,
|
|
"step": 6898
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.12029209255227e-07,
|
|
"loss": 0.471,
|
|
"step": 6899
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.10694230735188e-07,
|
|
"loss": 0.4327,
|
|
"step": 6900
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.093609491824103e-07,
|
|
"loss": 0.4695,
|
|
"step": 6901
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.080293648353185e-07,
|
|
"loss": 0.456,
|
|
"step": 6902
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.066994779320411e-07,
|
|
"loss": 0.4641,
|
|
"step": 6903
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.053712887104035e-07,
|
|
"loss": 0.4179,
|
|
"step": 6904
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.040447974079243e-07,
|
|
"loss": 0.4608,
|
|
"step": 6905
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.027200042618175e-07,
|
|
"loss": 0.3657,
|
|
"step": 6906
|
|
},
|
|
{
|
|
"epoch": 2.7,
|
|
"learning_rate": 5.013969095089944e-07,
|
|
"loss": 0.4494,
|
|
"step": 6907
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 5.000755133860657e-07,
|
|
"loss": 0.5145,
|
|
"step": 6908
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.98755816129336e-07,
|
|
"loss": 0.4942,
|
|
"step": 6909
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.974378179748052e-07,
|
|
"loss": 0.4271,
|
|
"step": 6910
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.961215191581692e-07,
|
|
"loss": 0.5305,
|
|
"step": 6911
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.94806919914822e-07,
|
|
"loss": 0.4033,
|
|
"step": 6912
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.934940204798533e-07,
|
|
"loss": 0.4129,
|
|
"step": 6913
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.921828210880475e-07,
|
|
"loss": 0.4821,
|
|
"step": 6914
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.908733219738837e-07,
|
|
"loss": 0.3959,
|
|
"step": 6915
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.895655233715413e-07,
|
|
"loss": 0.404,
|
|
"step": 6916
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.882594255148942e-07,
|
|
"loss": 0.4477,
|
|
"step": 6917
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.869550286375091e-07,
|
|
"loss": 0.425,
|
|
"step": 6918
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.856523329726503e-07,
|
|
"loss": 0.4684,
|
|
"step": 6919
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.843513387532783e-07,
|
|
"loss": 0.4271,
|
|
"step": 6920
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.830520462120514e-07,
|
|
"loss": 0.446,
|
|
"step": 6921
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.81754455581318e-07,
|
|
"loss": 0.4613,
|
|
"step": 6922
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.804585670931294e-07,
|
|
"loss": 0.4114,
|
|
"step": 6923
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.791643809792246e-07,
|
|
"loss": 0.3675,
|
|
"step": 6924
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.778718974710439e-07,
|
|
"loss": 0.4477,
|
|
"step": 6925
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.765811167997225e-07,
|
|
"loss": 0.4652,
|
|
"step": 6926
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.7529203919609025e-07,
|
|
"loss": 0.4113,
|
|
"step": 6927
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.7400466489066823e-07,
|
|
"loss": 0.4405,
|
|
"step": 6928
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.7271899411368116e-07,
|
|
"loss": 0.4497,
|
|
"step": 6929
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.71435027095043e-07,
|
|
"loss": 0.4363,
|
|
"step": 6930
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.701527640643666e-07,
|
|
"loss": 0.4499,
|
|
"step": 6931
|
|
},
|
|
{
|
|
"epoch": 2.71,
|
|
"learning_rate": 4.688722052509542e-07,
|
|
"loss": 0.4414,
|
|
"step": 6932
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.675933508838104e-07,
|
|
"loss": 0.4248,
|
|
"step": 6933
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.6631620119163336e-07,
|
|
"loss": 0.4769,
|
|
"step": 6934
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.6504075640281253e-07,
|
|
"loss": 0.4721,
|
|
"step": 6935
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.637670167454367e-07,
|
|
"loss": 0.4781,
|
|
"step": 6936
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.6249498244728573e-07,
|
|
"loss": 0.4527,
|
|
"step": 6937
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.612246537358389e-07,
|
|
"loss": 0.4293,
|
|
"step": 6938
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.599560308382678e-07,
|
|
"loss": 0.4385,
|
|
"step": 6939
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.5868911398143976e-07,
|
|
"loss": 0.4533,
|
|
"step": 6940
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.574239033919148e-07,
|
|
"loss": 0.4081,
|
|
"step": 6941
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.56160399295954e-07,
|
|
"loss": 0.4255,
|
|
"step": 6942
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.5489860191950454e-07,
|
|
"loss": 0.5453,
|
|
"step": 6943
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.536385114882169e-07,
|
|
"loss": 0.4362,
|
|
"step": 6944
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.5238012822742874e-07,
|
|
"loss": 0.488,
|
|
"step": 6945
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.511234523621799e-07,
|
|
"loss": 0.499,
|
|
"step": 6946
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.4986848411719743e-07,
|
|
"loss": 0.434,
|
|
"step": 6947
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.4861522371690723e-07,
|
|
"loss": 0.4849,
|
|
"step": 6948
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.4736367138543237e-07,
|
|
"loss": 0.4413,
|
|
"step": 6949
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.461138273465826e-07,
|
|
"loss": 0.4158,
|
|
"step": 6950
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.448656918238714e-07,
|
|
"loss": 0.4455,
|
|
"step": 6951
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.43619265040498e-07,
|
|
"loss": 0.4137,
|
|
"step": 6952
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.42374547219363e-07,
|
|
"loss": 0.4723,
|
|
"step": 6953
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.4113153858305725e-07,
|
|
"loss": 0.4523,
|
|
"step": 6954
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.398902393538684e-07,
|
|
"loss": 0.4587,
|
|
"step": 6955
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.386506497537757e-07,
|
|
"loss": 0.3767,
|
|
"step": 6956
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.374127700044561e-07,
|
|
"loss": 0.4383,
|
|
"step": 6957
|
|
},
|
|
{
|
|
"epoch": 2.72,
|
|
"learning_rate": 4.3617660032727715e-07,
|
|
"loss": 0.4324,
|
|
"step": 6958
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.349421409433041e-07,
|
|
"loss": 0.5057,
|
|
"step": 6959
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.3370939207329265e-07,
|
|
"loss": 0.4736,
|
|
"step": 6960
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.324783539376964e-07,
|
|
"loss": 0.4747,
|
|
"step": 6961
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.312490267566616e-07,
|
|
"loss": 0.465,
|
|
"step": 6962
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.3002141075002667e-07,
|
|
"loss": 0.3774,
|
|
"step": 6963
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.2879550613732504e-07,
|
|
"loss": 0.4485,
|
|
"step": 6964
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.2757131313778564e-07,
|
|
"loss": 0.4594,
|
|
"step": 6965
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.263488319703302e-07,
|
|
"loss": 0.3883,
|
|
"step": 6966
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.2512806285357475e-07,
|
|
"loss": 0.4406,
|
|
"step": 6967
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.239090060058271e-07,
|
|
"loss": 0.5016,
|
|
"step": 6968
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.226916616450916e-07,
|
|
"loss": 0.4671,
|
|
"step": 6969
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.214760299890663e-07,
|
|
"loss": 0.4846,
|
|
"step": 6970
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.2026211125514195e-07,
|
|
"loss": 0.4392,
|
|
"step": 6971
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.1904990566039915e-07,
|
|
"loss": 0.3937,
|
|
"step": 6972
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.1783941342161904e-07,
|
|
"loss": 0.4377,
|
|
"step": 6973
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.166306347552729e-07,
|
|
"loss": 0.3999,
|
|
"step": 6974
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.1542356987752775e-07,
|
|
"loss": 0.4656,
|
|
"step": 6975
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.142182190042399e-07,
|
|
"loss": 0.4116,
|
|
"step": 6976
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.1301458235096123e-07,
|
|
"loss": 0.4338,
|
|
"step": 6977
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.1181266013293864e-07,
|
|
"loss": 0.4218,
|
|
"step": 6978
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.1061245256511227e-07,
|
|
"loss": 0.4447,
|
|
"step": 6979
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.0941395986211275e-07,
|
|
"loss": 0.4459,
|
|
"step": 6980
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.082171822382652e-07,
|
|
"loss": 0.4684,
|
|
"step": 6981
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.0702211990758967e-07,
|
|
"loss": 0.4303,
|
|
"step": 6982
|
|
},
|
|
{
|
|
"epoch": 2.73,
|
|
"learning_rate": 4.0582877308380065e-07,
|
|
"loss": 0.4462,
|
|
"step": 6983
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 4.046371419803008e-07,
|
|
"loss": 0.4494,
|
|
"step": 6984
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 4.034472268101897e-07,
|
|
"loss": 0.5101,
|
|
"step": 6985
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 4.0225902778625814e-07,
|
|
"loss": 0.4409,
|
|
"step": 6986
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 4.0107254512099405e-07,
|
|
"loss": 0.4497,
|
|
"step": 6987
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.998877790265732e-07,
|
|
"loss": 0.5642,
|
|
"step": 6988
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.9870472971486605e-07,
|
|
"loss": 0.4608,
|
|
"step": 6989
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.9752339739743793e-07,
|
|
"loss": 0.4305,
|
|
"step": 6990
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.963437822855476e-07,
|
|
"loss": 0.4264,
|
|
"step": 6991
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.9516588459014184e-07,
|
|
"loss": 0.4545,
|
|
"step": 6992
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.9398970452186323e-07,
|
|
"loss": 0.4684,
|
|
"step": 6993
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.928152422910492e-07,
|
|
"loss": 0.4186,
|
|
"step": 6994
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.9164249810772823e-07,
|
|
"loss": 0.4464,
|
|
"step": 6995
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.9047147218162273e-07,
|
|
"loss": 0.4322,
|
|
"step": 6996
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.893021647221429e-07,
|
|
"loss": 0.4544,
|
|
"step": 6997
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.88134575938397e-07,
|
|
"loss": 0.4348,
|
|
"step": 6998
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.869687060391847e-07,
|
|
"loss": 0.5221,
|
|
"step": 6999
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.858045552329992e-07,
|
|
"loss": 0.4359,
|
|
"step": 7000
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.846421237280229e-07,
|
|
"loss": 0.4518,
|
|
"step": 7001
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.8348141173213285e-07,
|
|
"loss": 0.4594,
|
|
"step": 7002
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.823224194528985e-07,
|
|
"loss": 0.4544,
|
|
"step": 7003
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.811651470975841e-07,
|
|
"loss": 0.4417,
|
|
"step": 7004
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.800095948731419e-07,
|
|
"loss": 0.3977,
|
|
"step": 7005
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.7885576298621774e-07,
|
|
"loss": 0.4616,
|
|
"step": 7006
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.77703651643152e-07,
|
|
"loss": 0.4942,
|
|
"step": 7007
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.765532610499778e-07,
|
|
"loss": 0.4533,
|
|
"step": 7008
|
|
},
|
|
{
|
|
"epoch": 2.74,
|
|
"learning_rate": 3.75404591412416e-07,
|
|
"loss": 0.4599,
|
|
"step": 7009
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.742576429358835e-07,
|
|
"loss": 0.4625,
|
|
"step": 7010
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.731124158254895e-07,
|
|
"loss": 0.4692,
|
|
"step": 7011
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.7196891028603353e-07,
|
|
"loss": 0.4364,
|
|
"step": 7012
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.708271265220087e-07,
|
|
"loss": 0.4446,
|
|
"step": 7013
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.696870647375972e-07,
|
|
"loss": 0.452,
|
|
"step": 7014
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.6854872513667706e-07,
|
|
"loss": 0.4642,
|
|
"step": 7015
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.674121079228188e-07,
|
|
"loss": 0.4739,
|
|
"step": 7016
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.662772132992809e-07,
|
|
"loss": 0.4311,
|
|
"step": 7017
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.651440414690166e-07,
|
|
"loss": 0.4871,
|
|
"step": 7018
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.6401259263466937e-07,
|
|
"loss": 0.401,
|
|
"step": 7019
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.6288286699857736e-07,
|
|
"loss": 0.4589,
|
|
"step": 7020
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.6175486476276776e-07,
|
|
"loss": 0.4354,
|
|
"step": 7021
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.606285861289616e-07,
|
|
"loss": 0.4772,
|
|
"step": 7022
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.595040312985676e-07,
|
|
"loss": 0.4642,
|
|
"step": 7023
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.5838120047269274e-07,
|
|
"loss": 0.4396,
|
|
"step": 7024
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.572600938521309e-07,
|
|
"loss": 0.4897,
|
|
"step": 7025
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.561407116373694e-07,
|
|
"loss": 0.4088,
|
|
"step": 7026
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.5502305402858596e-07,
|
|
"loss": 0.4649,
|
|
"step": 7027
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.5390712122564953e-07,
|
|
"loss": 0.4254,
|
|
"step": 7028
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.5279291342812603e-07,
|
|
"loss": 0.436,
|
|
"step": 7029
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.516804308352628e-07,
|
|
"loss": 0.3984,
|
|
"step": 7030
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.5056967364601067e-07,
|
|
"loss": 0.4386,
|
|
"step": 7031
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.4946064205899966e-07,
|
|
"loss": 0.4564,
|
|
"step": 7032
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.483533362725622e-07,
|
|
"loss": 0.4638,
|
|
"step": 7033
|
|
},
|
|
{
|
|
"epoch": 2.75,
|
|
"learning_rate": 3.4724775648471434e-07,
|
|
"loss": 0.4162,
|
|
"step": 7034
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.461439028931679e-07,
|
|
"loss": 0.4226,
|
|
"step": 7035
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.4504177569532173e-07,
|
|
"loss": 0.4222,
|
|
"step": 7036
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.4394137508827254e-07,
|
|
"loss": 0.4525,
|
|
"step": 7037
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.428427012688007e-07,
|
|
"loss": 0.4771,
|
|
"step": 7038
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.417457544333846e-07,
|
|
"loss": 0.4992,
|
|
"step": 7039
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.4065053477818855e-07,
|
|
"loss": 0.4626,
|
|
"step": 7040
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.395570424990702e-07,
|
|
"loss": 0.4922,
|
|
"step": 7041
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.3846527779157866e-07,
|
|
"loss": 0.5089,
|
|
"step": 7042
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.373752408509523e-07,
|
|
"loss": 0.5158,
|
|
"step": 7043
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.362869318721251e-07,
|
|
"loss": 0.4652,
|
|
"step": 7044
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.3520035104971595e-07,
|
|
"loss": 0.4105,
|
|
"step": 7045
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.3411549857803816e-07,
|
|
"loss": 0.5158,
|
|
"step": 7046
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.330323746510955e-07,
|
|
"loss": 0.4366,
|
|
"step": 7047
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.319509794625819e-07,
|
|
"loss": 0.4744,
|
|
"step": 7048
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.3087131320588274e-07,
|
|
"loss": 0.4391,
|
|
"step": 7049
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.297933760740757e-07,
|
|
"loss": 0.4863,
|
|
"step": 7050
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.287171682599255e-07,
|
|
"loss": 0.4132,
|
|
"step": 7051
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.2764268995589156e-07,
|
|
"loss": 0.4633,
|
|
"step": 7052
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.265699413541201e-07,
|
|
"loss": 0.4204,
|
|
"step": 7053
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.254989226464522e-07,
|
|
"loss": 0.4254,
|
|
"step": 7054
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.244296340244157e-07,
|
|
"loss": 0.4792,
|
|
"step": 7055
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.233620756792322e-07,
|
|
"loss": 0.4808,
|
|
"step": 7056
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.2229624780181326e-07,
|
|
"loss": 0.457,
|
|
"step": 7057
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.212321505827598e-07,
|
|
"loss": 0.4654,
|
|
"step": 7058
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.201697842123619e-07,
|
|
"loss": 0.4328,
|
|
"step": 7059
|
|
},
|
|
{
|
|
"epoch": 2.76,
|
|
"learning_rate": 3.1910914888060307e-07,
|
|
"loss": 0.4469,
|
|
"step": 7060
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.180502447771572e-07,
|
|
"loss": 0.3916,
|
|
"step": 7061
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.169930720913872e-07,
|
|
"loss": 0.4588,
|
|
"step": 7062
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.159376310123452e-07,
|
|
"loss": 0.4315,
|
|
"step": 7063
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.1488392172877577e-07,
|
|
"loss": 0.4239,
|
|
"step": 7064
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.138319444291149e-07,
|
|
"loss": 0.4619,
|
|
"step": 7065
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.127816993014854e-07,
|
|
"loss": 0.4316,
|
|
"step": 7066
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.1173318653370145e-07,
|
|
"loss": 0.4472,
|
|
"step": 7067
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.1068640631326865e-07,
|
|
"loss": 0.4255,
|
|
"step": 7068
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.0964135882738275e-07,
|
|
"loss": 0.4389,
|
|
"step": 7069
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.085980442629288e-07,
|
|
"loss": 0.4015,
|
|
"step": 7070
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.075564628064831e-07,
|
|
"loss": 0.467,
|
|
"step": 7071
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.065166146443077e-07,
|
|
"loss": 0.4739,
|
|
"step": 7072
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.054784999623606e-07,
|
|
"loss": 0.459,
|
|
"step": 7073
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.0444211894628874e-07,
|
|
"loss": 0.4296,
|
|
"step": 7074
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.0340747178142503e-07,
|
|
"loss": 0.459,
|
|
"step": 7075
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.0237455865279484e-07,
|
|
"loss": 0.4623,
|
|
"step": 7076
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.0134337974511487e-07,
|
|
"loss": 0.5379,
|
|
"step": 7077
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 3.003139352427897e-07,
|
|
"loss": 0.4465,
|
|
"step": 7078
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 2.9928622532991446e-07,
|
|
"loss": 0.4753,
|
|
"step": 7079
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 2.9826025019027317e-07,
|
|
"loss": 0.4334,
|
|
"step": 7080
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 2.9723601000734013e-07,
|
|
"loss": 0.4356,
|
|
"step": 7081
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 2.9621350496428225e-07,
|
|
"loss": 0.4381,
|
|
"step": 7082
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 2.95192735243951e-07,
|
|
"loss": 0.4153,
|
|
"step": 7083
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 2.941737010288903e-07,
|
|
"loss": 0.4249,
|
|
"step": 7084
|
|
},
|
|
{
|
|
"epoch": 2.77,
|
|
"learning_rate": 2.931564025013334e-07,
|
|
"loss": 0.4281,
|
|
"step": 7085
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.921408398432046e-07,
|
|
"loss": 0.4561,
|
|
"step": 7086
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.911270132361166e-07,
|
|
"loss": 0.4462,
|
|
"step": 7087
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.901149228613698e-07,
|
|
"loss": 0.464,
|
|
"step": 7088
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.89104568899955e-07,
|
|
"loss": 0.4204,
|
|
"step": 7089
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.880959515325554e-07,
|
|
"loss": 0.4594,
|
|
"step": 7090
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.870890709395424e-07,
|
|
"loss": 0.4467,
|
|
"step": 7091
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.86083927300973e-07,
|
|
"loss": 0.3981,
|
|
"step": 7092
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.850805207965968e-07,
|
|
"loss": 0.4364,
|
|
"step": 7093
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.8407885160585345e-07,
|
|
"loss": 0.4282,
|
|
"step": 7094
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.8307891990787206e-07,
|
|
"loss": 0.4583,
|
|
"step": 7095
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.820807258814684e-07,
|
|
"loss": 0.3751,
|
|
"step": 7096
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.810842697051486e-07,
|
|
"loss": 0.4238,
|
|
"step": 7097
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.8008955155710784e-07,
|
|
"loss": 0.4452,
|
|
"step": 7098
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.7909657161523386e-07,
|
|
"loss": 0.4443,
|
|
"step": 7099
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.7810533005709796e-07,
|
|
"loss": 0.4743,
|
|
"step": 7100
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.771158270599639e-07,
|
|
"loss": 0.4329,
|
|
"step": 7101
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.761280628007845e-07,
|
|
"loss": 0.4109,
|
|
"step": 7102
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.7514203745620193e-07,
|
|
"loss": 0.4649,
|
|
"step": 7103
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.7415775120254507e-07,
|
|
"loss": 0.4805,
|
|
"step": 7104
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.731752042158331e-07,
|
|
"loss": 0.3997,
|
|
"step": 7105
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.721943966717744e-07,
|
|
"loss": 0.4431,
|
|
"step": 7106
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.7121532874576863e-07,
|
|
"loss": 0.4576,
|
|
"step": 7107
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.702380006128991e-07,
|
|
"loss": 0.4023,
|
|
"step": 7108
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.692624124479415e-07,
|
|
"loss": 0.4433,
|
|
"step": 7109
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.682885644253608e-07,
|
|
"loss": 0.4789,
|
|
"step": 7110
|
|
},
|
|
{
|
|
"epoch": 2.78,
|
|
"learning_rate": 2.6731645671930763e-07,
|
|
"loss": 0.4229,
|
|
"step": 7111
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.6634608950362626e-07,
|
|
"loss": 0.4663,
|
|
"step": 7112
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.653774629518457e-07,
|
|
"loss": 0.4459,
|
|
"step": 7113
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.644105772371841e-07,
|
|
"loss": 0.4022,
|
|
"step": 7114
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.634454325325497e-07,
|
|
"loss": 0.4499,
|
|
"step": 7115
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.62482029010539e-07,
|
|
"loss": 0.4431,
|
|
"step": 7116
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.615203668434374e-07,
|
|
"loss": 0.4194,
|
|
"step": 7117
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.605604462032163e-07,
|
|
"loss": 0.3604,
|
|
"step": 7118
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.5960226726153946e-07,
|
|
"loss": 0.4362,
|
|
"step": 7119
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.5864583018975763e-07,
|
|
"loss": 0.413,
|
|
"step": 7120
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.5769113515890954e-07,
|
|
"loss": 0.4901,
|
|
"step": 7121
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.567381823397219e-07,
|
|
"loss": 0.464,
|
|
"step": 7122
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.557869719026118e-07,
|
|
"loss": 0.4429,
|
|
"step": 7123
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.548375040176854e-07,
|
|
"loss": 0.4675,
|
|
"step": 7124
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.538897788547312e-07,
|
|
"loss": 0.4328,
|
|
"step": 7125
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.5294379658323374e-07,
|
|
"loss": 0.4436,
|
|
"step": 7126
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.5199955737236104e-07,
|
|
"loss": 0.4498,
|
|
"step": 7127
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.510570613909713e-07,
|
|
"loss": 0.4589,
|
|
"step": 7128
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.5011630880761084e-07,
|
|
"loss": 0.407,
|
|
"step": 7129
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.49177299790514e-07,
|
|
"loss": 0.4853,
|
|
"step": 7130
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.4824003450760193e-07,
|
|
"loss": 0.4518,
|
|
"step": 7131
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.4730451312648617e-07,
|
|
"loss": 0.4092,
|
|
"step": 7132
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.4637073581446625e-07,
|
|
"loss": 0.497,
|
|
"step": 7133
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.4543870273852857e-07,
|
|
"loss": 0.4737,
|
|
"step": 7134
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.445084140653464e-07,
|
|
"loss": 0.443,
|
|
"step": 7135
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.435798699612846e-07,
|
|
"loss": 0.4605,
|
|
"step": 7136
|
|
},
|
|
{
|
|
"epoch": 2.79,
|
|
"learning_rate": 2.426530705923946e-07,
|
|
"loss": 0.4299,
|
|
"step": 7137
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.41728016124414e-07,
|
|
"loss": 0.4864,
|
|
"step": 7138
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.4080470672277037e-07,
|
|
"loss": 0.482,
|
|
"step": 7139
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.3988314255257827e-07,
|
|
"loss": 0.3977,
|
|
"step": 7140
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.389633237786415e-07,
|
|
"loss": 0.4996,
|
|
"step": 7141
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.380452505654496e-07,
|
|
"loss": 0.4494,
|
|
"step": 7142
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.3712892307718227e-07,
|
|
"loss": 0.4633,
|
|
"step": 7143
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.3621434147770405e-07,
|
|
"loss": 0.4577,
|
|
"step": 7144
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.353015059305719e-07,
|
|
"loss": 0.5312,
|
|
"step": 7145
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.3439041659902405e-07,
|
|
"loss": 0.4893,
|
|
"step": 7146
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.334810736459936e-07,
|
|
"loss": 0.4504,
|
|
"step": 7147
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.3257347723409485e-07,
|
|
"loss": 0.4349,
|
|
"step": 7148
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.3166762752563466e-07,
|
|
"loss": 0.4594,
|
|
"step": 7149
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.3076352468260454e-07,
|
|
"loss": 0.4824,
|
|
"step": 7150
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.298611688666852e-07,
|
|
"loss": 0.4645,
|
|
"step": 7151
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.2896056023924528e-07,
|
|
"loss": 0.4898,
|
|
"step": 7152
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.2806169896134045e-07,
|
|
"loss": 0.4867,
|
|
"step": 7153
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.2716458519370988e-07,
|
|
"loss": 0.4526,
|
|
"step": 7154
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.2626921909678746e-07,
|
|
"loss": 0.4528,
|
|
"step": 7155
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.2537560083069066e-07,
|
|
"loss": 0.456,
|
|
"step": 7156
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.244837305552239e-07,
|
|
"loss": 0.4449,
|
|
"step": 7157
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.2359360842987954e-07,
|
|
"loss": 0.4705,
|
|
"step": 7158
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.227052346138381e-07,
|
|
"loss": 0.439,
|
|
"step": 7159
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.21818609265968e-07,
|
|
"loss": 0.425,
|
|
"step": 7160
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.209337325448224e-07,
|
|
"loss": 0.4855,
|
|
"step": 7161
|
|
},
|
|
{
|
|
"epoch": 2.8,
|
|
"learning_rate": 2.2005060460864369e-07,
|
|
"loss": 0.4899,
|
|
"step": 7162
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1916922561536104e-07,
|
|
"loss": 0.4939,
|
|
"step": 7163
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1828959572259168e-07,
|
|
"loss": 0.4619,
|
|
"step": 7164
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.174117150876398e-07,
|
|
"loss": 0.5322,
|
|
"step": 7165
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.165355838674943e-07,
|
|
"loss": 0.474,
|
|
"step": 7166
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1566120221883425e-07,
|
|
"loss": 0.5098,
|
|
"step": 7167
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.147885702980257e-07,
|
|
"loss": 0.4008,
|
|
"step": 7168
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1391768826111935e-07,
|
|
"loss": 0.4489,
|
|
"step": 7169
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1304855626385624e-07,
|
|
"loss": 0.5022,
|
|
"step": 7170
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1218117446166198e-07,
|
|
"loss": 0.4598,
|
|
"step": 7171
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1131554300964807e-07,
|
|
"loss": 0.4435,
|
|
"step": 7172
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.1045166206261847e-07,
|
|
"loss": 0.3975,
|
|
"step": 7173
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0958953177505958e-07,
|
|
"loss": 0.4879,
|
|
"step": 7174
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0872915230114254e-07,
|
|
"loss": 0.3926,
|
|
"step": 7175
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0787052379473204e-07,
|
|
"loss": 0.4646,
|
|
"step": 7176
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0701364640937527e-07,
|
|
"loss": 0.4439,
|
|
"step": 7177
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.061585202983063e-07,
|
|
"loss": 0.4299,
|
|
"step": 7178
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0530514561444725e-07,
|
|
"loss": 0.4368,
|
|
"step": 7179
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0445352251040718e-07,
|
|
"loss": 0.4134,
|
|
"step": 7180
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.036036511384798e-07,
|
|
"loss": 0.4918,
|
|
"step": 7181
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.027555316506502e-07,
|
|
"loss": 0.4549,
|
|
"step": 7182
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0190916419858486e-07,
|
|
"loss": 0.3967,
|
|
"step": 7183
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0106454893363824e-07,
|
|
"loss": 0.4555,
|
|
"step": 7184
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 2.0022168600685398e-07,
|
|
"loss": 0.4283,
|
|
"step": 7185
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 1.993805755689615e-07,
|
|
"loss": 0.4613,
|
|
"step": 7186
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 1.9854121777037493e-07,
|
|
"loss": 0.4655,
|
|
"step": 7187
|
|
},
|
|
{
|
|
"epoch": 2.81,
|
|
"learning_rate": 1.9770361276119532e-07,
|
|
"loss": 0.4685,
|
|
"step": 7188
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9686776069121284e-07,
|
|
"loss": 0.5198,
|
|
"step": 7189
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9603366170990234e-07,
|
|
"loss": 0.4337,
|
|
"step": 7190
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9520131596642455e-07,
|
|
"loss": 0.4911,
|
|
"step": 7191
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9437072360962704e-07,
|
|
"loss": 0.4821,
|
|
"step": 7192
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9354188478804436e-07,
|
|
"loss": 0.4948,
|
|
"step": 7193
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9271479964989903e-07,
|
|
"loss": 0.4279,
|
|
"step": 7194
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9188946834309608e-07,
|
|
"loss": 0.4593,
|
|
"step": 7195
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9106589101522965e-07,
|
|
"loss": 0.4272,
|
|
"step": 7196
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.9024406781357864e-07,
|
|
"loss": 0.4032,
|
|
"step": 7197
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.894239988851121e-07,
|
|
"loss": 0.4246,
|
|
"step": 7198
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8860568437648053e-07,
|
|
"loss": 0.4926,
|
|
"step": 7199
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8778912443402242e-07,
|
|
"loss": 0.5081,
|
|
"step": 7200
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8697431920376207e-07,
|
|
"loss": 0.4882,
|
|
"step": 7201
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8616126883141404e-07,
|
|
"loss": 0.4499,
|
|
"step": 7202
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8534997346237094e-07,
|
|
"loss": 0.4336,
|
|
"step": 7203
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.845404332417211e-07,
|
|
"loss": 0.4196,
|
|
"step": 7204
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.837326483142299e-07,
|
|
"loss": 0.4367,
|
|
"step": 7205
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8292661882435504e-07,
|
|
"loss": 0.3861,
|
|
"step": 7206
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8212234491624014e-07,
|
|
"loss": 0.4883,
|
|
"step": 7207
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8131982673371017e-07,
|
|
"loss": 0.3883,
|
|
"step": 7208
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.8051906442027922e-07,
|
|
"loss": 0.468,
|
|
"step": 7209
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.7972005811914717e-07,
|
|
"loss": 0.4953,
|
|
"step": 7210
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.7892280797320304e-07,
|
|
"loss": 0.4972,
|
|
"step": 7211
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.7812731412501504e-07,
|
|
"loss": 0.5134,
|
|
"step": 7212
|
|
},
|
|
{
|
|
"epoch": 2.82,
|
|
"learning_rate": 1.773335767168416e-07,
|
|
"loss": 0.4604,
|
|
"step": 7213
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.765415958906269e-07,
|
|
"loss": 0.4422,
|
|
"step": 7214
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.7575137178800107e-07,
|
|
"loss": 0.4612,
|
|
"step": 7215
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.7496290455027765e-07,
|
|
"loss": 0.4677,
|
|
"step": 7216
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.7417619431845945e-07,
|
|
"loss": 0.4939,
|
|
"step": 7217
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.7339124123323282e-07,
|
|
"loss": 0.4873,
|
|
"step": 7218
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.7260804543497102e-07,
|
|
"loss": 0.4457,
|
|
"step": 7219
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.718266070637309e-07,
|
|
"loss": 0.4199,
|
|
"step": 7220
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.7104692625925845e-07,
|
|
"loss": 0.5455,
|
|
"step": 7221
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.7026900316098217e-07,
|
|
"loss": 0.4455,
|
|
"step": 7222
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6949283790801962e-07,
|
|
"loss": 0.4841,
|
|
"step": 7223
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.687184306391687e-07,
|
|
"loss": 0.4466,
|
|
"step": 7224
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6794578149291975e-07,
|
|
"loss": 0.4291,
|
|
"step": 7225
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6717489060744328e-07,
|
|
"loss": 0.4258,
|
|
"step": 7226
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.66405758120598e-07,
|
|
"loss": 0.4212,
|
|
"step": 7227
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6563838416992607e-07,
|
|
"loss": 0.3755,
|
|
"step": 7228
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.648727688926577e-07,
|
|
"loss": 0.4503,
|
|
"step": 7229
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.641089124257078e-07,
|
|
"loss": 0.4692,
|
|
"step": 7230
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6334681490567494e-07,
|
|
"loss": 0.4515,
|
|
"step": 7231
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6258647646884674e-07,
|
|
"loss": 0.4257,
|
|
"step": 7232
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6182789725119218e-07,
|
|
"loss": 0.4702,
|
|
"step": 7233
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6107107738836835e-07,
|
|
"loss": 0.503,
|
|
"step": 7234
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.6031601701571696e-07,
|
|
"loss": 0.3864,
|
|
"step": 7235
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.5956271626826448e-07,
|
|
"loss": 0.4364,
|
|
"step": 7236
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.588111752807242e-07,
|
|
"loss": 0.4909,
|
|
"step": 7237
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.580613941874931e-07,
|
|
"loss": 0.5135,
|
|
"step": 7238
|
|
},
|
|
{
|
|
"epoch": 2.83,
|
|
"learning_rate": 1.5731337312265503e-07,
|
|
"loss": 0.4587,
|
|
"step": 7239
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5656711221997635e-07,
|
|
"loss": 0.4066,
|
|
"step": 7240
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5582261161291246e-07,
|
|
"loss": 0.4261,
|
|
"step": 7241
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5507987143460024e-07,
|
|
"loss": 0.4234,
|
|
"step": 7242
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5433889181786454e-07,
|
|
"loss": 0.457,
|
|
"step": 7243
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5359967289521493e-07,
|
|
"loss": 0.478,
|
|
"step": 7244
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5286221479884343e-07,
|
|
"loss": 0.454,
|
|
"step": 7245
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5212651766063124e-07,
|
|
"loss": 0.4685,
|
|
"step": 7246
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.513925816121431e-07,
|
|
"loss": 0.4375,
|
|
"step": 7247
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.5066040678462622e-07,
|
|
"loss": 0.4536,
|
|
"step": 7248
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4992999330901593e-07,
|
|
"loss": 0.4629,
|
|
"step": 7249
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4920134131593101e-07,
|
|
"loss": 0.5393,
|
|
"step": 7250
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4847445093567836e-07,
|
|
"loss": 0.4379,
|
|
"step": 7251
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4774932229824513e-07,
|
|
"loss": 0.4356,
|
|
"step": 7252
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4702595553330645e-07,
|
|
"loss": 0.4548,
|
|
"step": 7253
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.463043507702211e-07,
|
|
"loss": 0.4174,
|
|
"step": 7254
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.455845081380336e-07,
|
|
"loss": 0.4028,
|
|
"step": 7255
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.448664277654721e-07,
|
|
"loss": 0.4232,
|
|
"step": 7256
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4415010978095279e-07,
|
|
"loss": 0.5058,
|
|
"step": 7257
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4343555431257206e-07,
|
|
"loss": 0.4841,
|
|
"step": 7258
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4272276148811436e-07,
|
|
"loss": 0.4168,
|
|
"step": 7259
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4201173143504887e-07,
|
|
"loss": 0.4669,
|
|
"step": 7260
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.413024642805283e-07,
|
|
"loss": 0.4026,
|
|
"step": 7261
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.4059496015138896e-07,
|
|
"loss": 0.4079,
|
|
"step": 7262
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.398892191741541e-07,
|
|
"loss": 0.4112,
|
|
"step": 7263
|
|
},
|
|
{
|
|
"epoch": 2.84,
|
|
"learning_rate": 1.3918524147503277e-07,
|
|
"loss": 0.3721,
|
|
"step": 7264
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3848302717991536e-07,
|
|
"loss": 0.4496,
|
|
"step": 7265
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3778257641437808e-07,
|
|
"loss": 0.4314,
|
|
"step": 7266
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.37083889303683e-07,
|
|
"loss": 0.4242,
|
|
"step": 7267
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3638696597277678e-07,
|
|
"loss": 0.425,
|
|
"step": 7268
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3569180654628755e-07,
|
|
"loss": 0.4473,
|
|
"step": 7269
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.349984111485314e-07,
|
|
"loss": 0.4559,
|
|
"step": 7270
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3430677990350805e-07,
|
|
"loss": 0.4558,
|
|
"step": 7271
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3361691293490187e-07,
|
|
"loss": 0.4676,
|
|
"step": 7272
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3292881036608196e-07,
|
|
"loss": 0.4222,
|
|
"step": 7273
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.322424723200988e-07,
|
|
"loss": 0.4396,
|
|
"step": 7274
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.3155789891969084e-07,
|
|
"loss": 0.4903,
|
|
"step": 7275
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.308750902872813e-07,
|
|
"loss": 0.4912,
|
|
"step": 7276
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.301940465449758e-07,
|
|
"loss": 0.4324,
|
|
"step": 7277
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2951476781456473e-07,
|
|
"loss": 0.4078,
|
|
"step": 7278
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2883725421752203e-07,
|
|
"loss": 0.4142,
|
|
"step": 7279
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2816150587500964e-07,
|
|
"loss": 0.435,
|
|
"step": 7280
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2748752290786981e-07,
|
|
"loss": 0.4844,
|
|
"step": 7281
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2681530543663056e-07,
|
|
"loss": 0.4342,
|
|
"step": 7282
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2614485358150352e-07,
|
|
"loss": 0.4616,
|
|
"step": 7283
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.254761674623861e-07,
|
|
"loss": 0.4123,
|
|
"step": 7284
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2480924719885934e-07,
|
|
"loss": 0.442,
|
|
"step": 7285
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.241440929101878e-07,
|
|
"loss": 0.5441,
|
|
"step": 7286
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.234807047153208e-07,
|
|
"loss": 0.433,
|
|
"step": 7287
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2281908273289122e-07,
|
|
"loss": 0.4304,
|
|
"step": 7288
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.2215922708121665e-07,
|
|
"loss": 0.4329,
|
|
"step": 7289
|
|
},
|
|
{
|
|
"epoch": 2.85,
|
|
"learning_rate": 1.215011378782993e-07,
|
|
"loss": 0.4451,
|
|
"step": 7290
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.20844815241824e-07,
|
|
"loss": 0.4342,
|
|
"step": 7291
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.2019025928916128e-07,
|
|
"loss": 0.4852,
|
|
"step": 7292
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1953747013736416e-07,
|
|
"loss": 0.4255,
|
|
"step": 7293
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1888644790317261e-07,
|
|
"loss": 0.4256,
|
|
"step": 7294
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1823719270300571e-07,
|
|
"loss": 0.4647,
|
|
"step": 7295
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.175897046529717e-07,
|
|
"loss": 0.3439,
|
|
"step": 7296
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1694398386885908e-07,
|
|
"loss": 0.4082,
|
|
"step": 7297
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1630003046614324e-07,
|
|
"loss": 0.48,
|
|
"step": 7298
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1565784455997985e-07,
|
|
"loss": 0.4644,
|
|
"step": 7299
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1501742626521262e-07,
|
|
"loss": 0.4336,
|
|
"step": 7300
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1437877569636547e-07,
|
|
"loss": 0.4017,
|
|
"step": 7301
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1374189296765037e-07,
|
|
"loss": 0.4593,
|
|
"step": 7302
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1310677819295957e-07,
|
|
"loss": 0.4277,
|
|
"step": 7303
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1247343148586886e-07,
|
|
"loss": 0.3847,
|
|
"step": 7304
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.118418529596399e-07,
|
|
"loss": 0.3856,
|
|
"step": 7305
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1121204272721897e-07,
|
|
"loss": 0.4214,
|
|
"step": 7306
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.1058400090123378e-07,
|
|
"loss": 0.406,
|
|
"step": 7307
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.0995772759399559e-07,
|
|
"loss": 0.4392,
|
|
"step": 7308
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.093332229175026e-07,
|
|
"loss": 0.4424,
|
|
"step": 7309
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.0871048698343323e-07,
|
|
"loss": 0.483,
|
|
"step": 7310
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.0808951990315064e-07,
|
|
"loss": 0.453,
|
|
"step": 7311
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.0747032178770378e-07,
|
|
"loss": 0.4306,
|
|
"step": 7312
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.0685289274782296e-07,
|
|
"loss": 0.4661,
|
|
"step": 7313
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.0623723289392097e-07,
|
|
"loss": 0.4429,
|
|
"step": 7314
|
|
},
|
|
{
|
|
"epoch": 2.86,
|
|
"learning_rate": 1.056233423360975e-07,
|
|
"loss": 0.4532,
|
|
"step": 7315
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0501122118413477e-07,
|
|
"loss": 0.4008,
|
|
"step": 7316
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0440086954749517e-07,
|
|
"loss": 0.4825,
|
|
"step": 7317
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0379228753533144e-07,
|
|
"loss": 0.4704,
|
|
"step": 7318
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0318547525647316e-07,
|
|
"loss": 0.4577,
|
|
"step": 7319
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0258043281943797e-07,
|
|
"loss": 0.4345,
|
|
"step": 7320
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0197716033242377e-07,
|
|
"loss": 0.4457,
|
|
"step": 7321
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0137565790331316e-07,
|
|
"loss": 0.4806,
|
|
"step": 7322
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0077592563967564e-07,
|
|
"loss": 0.408,
|
|
"step": 7323
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 1.0017796364875653e-07,
|
|
"loss": 0.402,
|
|
"step": 7324
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.958177203749364e-08,
|
|
"loss": 0.4191,
|
|
"step": 7325
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.898735091249945e-08,
|
|
"loss": 0.4348,
|
|
"step": 7326
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.83947003800767e-08,
|
|
"loss": 0.4456,
|
|
"step": 7327
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.780382054620729e-08,
|
|
"loss": 0.4337,
|
|
"step": 7328
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.721471151655892e-08,
|
|
"loss": 0.4362,
|
|
"step": 7329
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.662737339648065e-08,
|
|
"loss": 0.4269,
|
|
"step": 7330
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.604180629100735e-08,
|
|
"loss": 0.4511,
|
|
"step": 7331
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.545801030485302e-08,
|
|
"loss": 0.4055,
|
|
"step": 7332
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.487598554242084e-08,
|
|
"loss": 0.4682,
|
|
"step": 7333
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.429573210779087e-08,
|
|
"loss": 0.4357,
|
|
"step": 7334
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.371725010473231e-08,
|
|
"loss": 0.4494,
|
|
"step": 7335
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.314053963669245e-08,
|
|
"loss": 0.5252,
|
|
"step": 7336
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.256560080680543e-08,
|
|
"loss": 0.4247,
|
|
"step": 7337
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.19924337178868e-08,
|
|
"loss": 0.491,
|
|
"step": 7338
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.142103847243677e-08,
|
|
"loss": 0.4542,
|
|
"step": 7339
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.085141517263584e-08,
|
|
"loss": 0.443,
|
|
"step": 7340
|
|
},
|
|
{
|
|
"epoch": 2.87,
|
|
"learning_rate": 9.02835639203492e-08,
|
|
"loss": 0.4578,
|
|
"step": 7341
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.971748481712894e-08,
|
|
"loss": 0.4051,
|
|
"step": 7342
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.915317796420297e-08,
|
|
"loss": 0.4784,
|
|
"step": 7343
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.859064346248613e-08,
|
|
"loss": 0.437,
|
|
"step": 7344
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.802988141257795e-08,
|
|
"loss": 0.4363,
|
|
"step": 7345
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.747089191475932e-08,
|
|
"loss": 0.4989,
|
|
"step": 7346
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.69136750689914e-08,
|
|
"loss": 0.457,
|
|
"step": 7347
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.635823097492447e-08,
|
|
"loss": 0.5326,
|
|
"step": 7348
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.580455973188462e-08,
|
|
"loss": 0.5121,
|
|
"step": 7349
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.525266143888822e-08,
|
|
"loss": 0.488,
|
|
"step": 7350
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.470253619462854e-08,
|
|
"loss": 0.4532,
|
|
"step": 7351
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.415418409748466e-08,
|
|
"loss": 0.4049,
|
|
"step": 7352
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.360760524551814e-08,
|
|
"loss": 0.4967,
|
|
"step": 7353
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.306279973647414e-08,
|
|
"loss": 0.4709,
|
|
"step": 7354
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.251976766777914e-08,
|
|
"loss": 0.423,
|
|
"step": 7355
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.197850913654326e-08,
|
|
"loss": 0.4446,
|
|
"step": 7356
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.143902423955908e-08,
|
|
"loss": 0.4518,
|
|
"step": 7357
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.090131307330274e-08,
|
|
"loss": 0.3917,
|
|
"step": 7358
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 8.036537573393289e-08,
|
|
"loss": 0.432,
|
|
"step": 7359
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 7.983121231729174e-08,
|
|
"loss": 0.4328,
|
|
"step": 7360
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 7.929882291890289e-08,
|
|
"loss": 0.4235,
|
|
"step": 7361
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 7.876820763397241e-08,
|
|
"loss": 0.4739,
|
|
"step": 7362
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 7.823936655738995e-08,
|
|
"loss": 0.4635,
|
|
"step": 7363
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 7.771229978372985e-08,
|
|
"loss": 0.4087,
|
|
"step": 7364
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 7.71870074072445e-08,
|
|
"loss": 0.5189,
|
|
"step": 7365
|
|
},
|
|
{
|
|
"epoch": 2.88,
|
|
"learning_rate": 7.66634895218743e-08,
|
|
"loss": 0.4,
|
|
"step": 7366
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.61417462212366e-08,
|
|
"loss": 0.4217,
|
|
"step": 7367
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.562177759863787e-08,
|
|
"loss": 0.4532,
|
|
"step": 7368
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.51035837470615e-08,
|
|
"loss": 0.4506,
|
|
"step": 7369
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.45871647591756e-08,
|
|
"loss": 0.4547,
|
|
"step": 7370
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.407252072733184e-08,
|
|
"loss": 0.4425,
|
|
"step": 7371
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.35596517435655e-08,
|
|
"loss": 0.5397,
|
|
"step": 7372
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.304855789958986e-08,
|
|
"loss": 0.4507,
|
|
"step": 7373
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.253923928680406e-08,
|
|
"loss": 0.4528,
|
|
"step": 7374
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.203169599629078e-08,
|
|
"loss": 0.5078,
|
|
"step": 7375
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.152592811881187e-08,
|
|
"loss": 0.4332,
|
|
"step": 7376
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.102193574481497e-08,
|
|
"loss": 0.4329,
|
|
"step": 7377
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.0519718964428e-08,
|
|
"loss": 0.4269,
|
|
"step": 7378
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 7.001927786746243e-08,
|
|
"loss": 0.4508,
|
|
"step": 7379
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.952061254341225e-08,
|
|
"loss": 0.4467,
|
|
"step": 7380
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.902372308145167e-08,
|
|
"loss": 0.4884,
|
|
"step": 7381
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.852860957044072e-08,
|
|
"loss": 0.4639,
|
|
"step": 7382
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.80352720989208e-08,
|
|
"loss": 0.4444,
|
|
"step": 7383
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.754371075511357e-08,
|
|
"loss": 0.5068,
|
|
"step": 7384
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.705392562692648e-08,
|
|
"loss": 0.4757,
|
|
"step": 7385
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.656591680194613e-08,
|
|
"loss": 0.4856,
|
|
"step": 7386
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.607968436744272e-08,
|
|
"loss": 0.5009,
|
|
"step": 7387
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.559522841037003e-08,
|
|
"loss": 0.446,
|
|
"step": 7388
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.511254901736208e-08,
|
|
"loss": 0.4427,
|
|
"step": 7389
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.46316462747365e-08,
|
|
"loss": 0.4754,
|
|
"step": 7390
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.415252026849339e-08,
|
|
"loss": 0.5148,
|
|
"step": 7391
|
|
},
|
|
{
|
|
"epoch": 2.89,
|
|
"learning_rate": 6.36751710843131e-08,
|
|
"loss": 0.4298,
|
|
"step": 7392
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 6.319959880756176e-08,
|
|
"loss": 0.371,
|
|
"step": 7393
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 6.272580352328473e-08,
|
|
"loss": 0.4931,
|
|
"step": 7394
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 6.225378531620973e-08,
|
|
"loss": 0.4367,
|
|
"step": 7395
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 6.178354427074928e-08,
|
|
"loss": 0.4203,
|
|
"step": 7396
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 6.131508047099499e-08,
|
|
"loss": 0.4595,
|
|
"step": 7397
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 6.084839400072207e-08,
|
|
"loss": 0.4668,
|
|
"step": 7398
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 6.038348494338819e-08,
|
|
"loss": 0.423,
|
|
"step": 7399
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.992035338213242e-08,
|
|
"loss": 0.4712,
|
|
"step": 7400
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.9458999399776284e-08,
|
|
"loss": 0.5246,
|
|
"step": 7401
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.899942307882489e-08,
|
|
"loss": 0.4618,
|
|
"step": 7402
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.8541624501461393e-08,
|
|
"loss": 0.4288,
|
|
"step": 7403
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.808560374955585e-08,
|
|
"loss": 0.4211,
|
|
"step": 7404
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.763136090465748e-08,
|
|
"loss": 0.4416,
|
|
"step": 7405
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.717889604799798e-08,
|
|
"loss": 0.4655,
|
|
"step": 7406
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.672820926049261e-08,
|
|
"loss": 0.4323,
|
|
"step": 7407
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.627930062273468e-08,
|
|
"loss": 0.4467,
|
|
"step": 7408
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.583217021500553e-08,
|
|
"loss": 0.4058,
|
|
"step": 7409
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.538681811726343e-08,
|
|
"loss": 0.4502,
|
|
"step": 7410
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.4943244409151334e-08,
|
|
"loss": 0.4721,
|
|
"step": 7411
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.450144916999134e-08,
|
|
"loss": 0.5141,
|
|
"step": 7412
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.4061432478791364e-08,
|
|
"loss": 0.4803,
|
|
"step": 7413
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.3623194414239576e-08,
|
|
"loss": 0.5094,
|
|
"step": 7414
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.3186735054704395e-08,
|
|
"loss": 0.462,
|
|
"step": 7415
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.2752054478238943e-08,
|
|
"loss": 0.5019,
|
|
"step": 7416
|
|
},
|
|
{
|
|
"epoch": 2.9,
|
|
"learning_rate": 5.2319152762575486e-08,
|
|
"loss": 0.4452,
|
|
"step": 7417
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 5.188802998513098e-08,
|
|
"loss": 0.4974,
|
|
"step": 7418
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 5.1458686223001544e-08,
|
|
"loss": 0.4461,
|
|
"step": 7419
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 5.103112155296797e-08,
|
|
"loss": 0.4031,
|
|
"step": 7420
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 5.06053360514902e-08,
|
|
"loss": 0.3963,
|
|
"step": 7421
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 5.0181329794711795e-08,
|
|
"loss": 0.4743,
|
|
"step": 7422
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.9759102858457643e-08,
|
|
"loss": 0.4834,
|
|
"step": 7423
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.9338655318234006e-08,
|
|
"loss": 0.4157,
|
|
"step": 7424
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.891998724922964e-08,
|
|
"loss": 0.4863,
|
|
"step": 7425
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.8503098726314645e-08,
|
|
"loss": 0.4454,
|
|
"step": 7426
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.808798982403939e-08,
|
|
"loss": 0.4621,
|
|
"step": 7427
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.767466061664006e-08,
|
|
"loss": 0.4466,
|
|
"step": 7428
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.726311117803084e-08,
|
|
"loss": 0.4358,
|
|
"step": 7429
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.685334158180954e-08,
|
|
"loss": 0.3946,
|
|
"step": 7430
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.644535190125421e-08,
|
|
"loss": 0.4411,
|
|
"step": 7431
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.603914220932537e-08,
|
|
"loss": 0.4677,
|
|
"step": 7432
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.563471257866714e-08,
|
|
"loss": 0.4545,
|
|
"step": 7433
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.5232063081600555e-08,
|
|
"loss": 0.4084,
|
|
"step": 7434
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.483119379013357e-08,
|
|
"loss": 0.45,
|
|
"step": 7435
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.443210477595106e-08,
|
|
"loss": 0.4693,
|
|
"step": 7436
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.4034796110424825e-08,
|
|
"loss": 0.4087,
|
|
"step": 7437
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.3639267864603594e-08,
|
|
"loss": 0.4944,
|
|
"step": 7438
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.324552010921857e-08,
|
|
"loss": 0.4523,
|
|
"step": 7439
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.2853552914685625e-08,
|
|
"loss": 0.4501,
|
|
"step": 7440
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.246336635109871e-08,
|
|
"loss": 0.478,
|
|
"step": 7441
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.207496048823423e-08,
|
|
"loss": 0.3808,
|
|
"step": 7442
|
|
},
|
|
{
|
|
"epoch": 2.91,
|
|
"learning_rate": 4.168833539555217e-08,
|
|
"loss": 0.4466,
|
|
"step": 7443
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 4.1303491142190564e-08,
|
|
"loss": 0.4347,
|
|
"step": 7444
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 4.092042779697325e-08,
|
|
"loss": 0.4375,
|
|
"step": 7445
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 4.053914542840098e-08,
|
|
"loss": 0.4481,
|
|
"step": 7446
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 4.015964410466033e-08,
|
|
"loss": 0.5176,
|
|
"step": 7447
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.9781923893614793e-08,
|
|
"loss": 0.4125,
|
|
"step": 7448
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.9405984862813665e-08,
|
|
"loss": 0.3931,
|
|
"step": 7449
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.9031827079486494e-08,
|
|
"loss": 0.4727,
|
|
"step": 7450
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.865945061054199e-08,
|
|
"loss": 0.402,
|
|
"step": 7451
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.828885552257244e-08,
|
|
"loss": 0.4519,
|
|
"step": 7452
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.79200418818515e-08,
|
|
"loss": 0.449,
|
|
"step": 7453
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.755300975433418e-08,
|
|
"loss": 0.4219,
|
|
"step": 7454
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.718775920565687e-08,
|
|
"loss": 0.4156,
|
|
"step": 7455
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.682429030113621e-08,
|
|
"loss": 0.4449,
|
|
"step": 7456
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.646260310577021e-08,
|
|
"loss": 0.4206,
|
|
"step": 7457
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.6102697684241574e-08,
|
|
"loss": 0.4414,
|
|
"step": 7458
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.574457410091215e-08,
|
|
"loss": 0.499,
|
|
"step": 7459
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.538823241982181e-08,
|
|
"loss": 0.4843,
|
|
"step": 7460
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.5033672704697375e-08,
|
|
"loss": 0.4593,
|
|
"step": 7461
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.468089501894478e-08,
|
|
"loss": 0.4461,
|
|
"step": 7462
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.432989942565024e-08,
|
|
"loss": 0.403,
|
|
"step": 7463
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.398068598758131e-08,
|
|
"loss": 0.4613,
|
|
"step": 7464
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.363325476718915e-08,
|
|
"loss": 0.4635,
|
|
"step": 7465
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.328760582660295e-08,
|
|
"loss": 0.4509,
|
|
"step": 7466
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.2943739227637716e-08,
|
|
"loss": 0.4277,
|
|
"step": 7467
|
|
},
|
|
{
|
|
"epoch": 2.92,
|
|
"learning_rate": 3.260165503178425e-08,
|
|
"loss": 0.5112,
|
|
"step": 7468
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 3.226135330021918e-08,
|
|
"loss": 0.4017,
|
|
"step": 7469
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 3.192283409379604e-08,
|
|
"loss": 0.3954,
|
|
"step": 7470
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 3.15860974730553e-08,
|
|
"loss": 0.5226,
|
|
"step": 7471
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 3.125114349821212e-08,
|
|
"loss": 0.5081,
|
|
"step": 7472
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 3.091797222916859e-08,
|
|
"loss": 0.4768,
|
|
"step": 7473
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 3.0586583725503715e-08,
|
|
"loss": 0.4763,
|
|
"step": 7474
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 3.02569780464812e-08,
|
|
"loss": 0.4746,
|
|
"step": 7475
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.992915525104389e-08,
|
|
"loss": 0.4427,
|
|
"step": 7476
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.9603115397814907e-08,
|
|
"loss": 0.4828,
|
|
"step": 7477
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.927885854509982e-08,
|
|
"loss": 0.4115,
|
|
"step": 7478
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.895638475088669e-08,
|
|
"loss": 0.4332,
|
|
"step": 7479
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.8635694072843835e-08,
|
|
"loss": 0.4626,
|
|
"step": 7480
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.83167865683176e-08,
|
|
"loss": 0.43,
|
|
"step": 7481
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.7999662294340145e-08,
|
|
"loss": 0.4798,
|
|
"step": 7482
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.7684321307621665e-08,
|
|
"loss": 0.4917,
|
|
"step": 7483
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.737076366455482e-08,
|
|
"loss": 0.4431,
|
|
"step": 7484
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.7058989421212545e-08,
|
|
"loss": 0.4282,
|
|
"step": 7485
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.674899863335023e-08,
|
|
"loss": 0.363,
|
|
"step": 7486
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.6440791356402428e-08,
|
|
"loss": 0.4764,
|
|
"step": 7487
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.613436764548505e-08,
|
|
"loss": 0.432,
|
|
"step": 7488
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.5829727555397587e-08,
|
|
"loss": 0.489,
|
|
"step": 7489
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.552687114061758e-08,
|
|
"loss": 0.4578,
|
|
"step": 7490
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.5225798455303928e-08,
|
|
"loss": 0.4432,
|
|
"step": 7491
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.4926509553299118e-08,
|
|
"loss": 0.4169,
|
|
"step": 7492
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.4629004488123663e-08,
|
|
"loss": 0.4462,
|
|
"step": 7493
|
|
},
|
|
{
|
|
"epoch": 2.93,
|
|
"learning_rate": 2.4333283312979463e-08,
|
|
"loss": 0.426,
|
|
"step": 7494
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.403934608075087e-08,
|
|
"loss": 0.4562,
|
|
"step": 7495
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.374719284400362e-08,
|
|
"loss": 0.4419,
|
|
"step": 7496
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.3456823654982586e-08,
|
|
"loss": 0.4542,
|
|
"step": 7497
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.31682385656129e-08,
|
|
"loss": 0.4688,
|
|
"step": 7498
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.288143762750439e-08,
|
|
"loss": 0.4729,
|
|
"step": 7499
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.2596420891943803e-08,
|
|
"loss": 0.441,
|
|
"step": 7500
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.2313188409901487e-08,
|
|
"loss": 0.422,
|
|
"step": 7501
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.2031740232026923e-08,
|
|
"loss": 0.4149,
|
|
"step": 7502
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.1752076408652067e-08,
|
|
"loss": 0.3947,
|
|
"step": 7503
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.147419698978914e-08,
|
|
"loss": 0.4745,
|
|
"step": 7504
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.119810202513062e-08,
|
|
"loss": 0.4935,
|
|
"step": 7505
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.0923791564050333e-08,
|
|
"loss": 0.537,
|
|
"step": 7506
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.065126565560349e-08,
|
|
"loss": 0.4525,
|
|
"step": 7507
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.0380524348525556e-08,
|
|
"loss": 0.475,
|
|
"step": 7508
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 2.0111567691233346e-08,
|
|
"loss": 0.4979,
|
|
"step": 7509
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.9844395731823952e-08,
|
|
"loss": 0.4282,
|
|
"step": 7510
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.9579008518074704e-08,
|
|
"loss": 0.4888,
|
|
"step": 7511
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.9315406097446533e-08,
|
|
"loss": 0.4533,
|
|
"step": 7512
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.905358851707839e-08,
|
|
"loss": 0.3951,
|
|
"step": 7513
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.879355582379061e-08,
|
|
"loss": 0.4586,
|
|
"step": 7514
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.8535308064084877e-08,
|
|
"loss": 0.5017,
|
|
"step": 7515
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.8278845284143143e-08,
|
|
"loss": 0.4177,
|
|
"step": 7516
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.802416752982983e-08,
|
|
"loss": 0.4125,
|
|
"step": 7517
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.7771274846687393e-08,
|
|
"loss": 0.3945,
|
|
"step": 7518
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.752016727994188e-08,
|
|
"loss": 0.564,
|
|
"step": 7519
|
|
},
|
|
{
|
|
"epoch": 2.94,
|
|
"learning_rate": 1.727084487449737e-08,
|
|
"loss": 0.4301,
|
|
"step": 7520
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.7023307674940424e-08,
|
|
"loss": 0.4975,
|
|
"step": 7521
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.6777555725537854e-08,
|
|
"loss": 0.4628,
|
|
"step": 7522
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.653358907023783e-08,
|
|
"loss": 0.412,
|
|
"step": 7523
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.6291407752668796e-08,
|
|
"loss": 0.4872,
|
|
"step": 7524
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.6051011816139438e-08,
|
|
"loss": 0.4445,
|
|
"step": 7525
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.5812401303639813e-08,
|
|
"loss": 0.5377,
|
|
"step": 7526
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.557557625784023e-08,
|
|
"loss": 0.4597,
|
|
"step": 7527
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.534053672109348e-08,
|
|
"loss": 0.4367,
|
|
"step": 7528
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.5107282735429274e-08,
|
|
"loss": 0.4835,
|
|
"step": 7529
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.4875814342562023e-08,
|
|
"loss": 0.4372,
|
|
"step": 7530
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.464613158388417e-08,
|
|
"loss": 0.4637,
|
|
"step": 7531
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.441823450046953e-08,
|
|
"loss": 0.4185,
|
|
"step": 7532
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.4192123133074387e-08,
|
|
"loss": 0.4535,
|
|
"step": 7533
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.3967797522131953e-08,
|
|
"loss": 0.4392,
|
|
"step": 7534
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.374525770776014e-08,
|
|
"loss": 0.4605,
|
|
"step": 7535
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.352450372975489e-08,
|
|
"loss": 0.3966,
|
|
"step": 7536
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.3305535627592404e-08,
|
|
"loss": 0.512,
|
|
"step": 7537
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.3088353440433577e-08,
|
|
"loss": 0.4442,
|
|
"step": 7538
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.2872957207112901e-08,
|
|
"loss": 0.4589,
|
|
"step": 7539
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.2659346966152897e-08,
|
|
"loss": 0.4472,
|
|
"step": 7540
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.2447522755751896e-08,
|
|
"loss": 0.4342,
|
|
"step": 7541
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.2237484613790707e-08,
|
|
"loss": 0.4919,
|
|
"step": 7542
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.2029232577830396e-08,
|
|
"loss": 0.4437,
|
|
"step": 7543
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.1822766685112285e-08,
|
|
"loss": 0.4787,
|
|
"step": 7544
|
|
},
|
|
{
|
|
"epoch": 2.95,
|
|
"learning_rate": 1.1618086972559062e-08,
|
|
"loss": 0.4525,
|
|
"step": 7545
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.1415193476772557e-08,
|
|
"loss": 0.4932,
|
|
"step": 7546
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.121408623403708e-08,
|
|
"loss": 0.4669,
|
|
"step": 7547
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.1014765280314977e-08,
|
|
"loss": 0.4452,
|
|
"step": 7548
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.0817230651253285e-08,
|
|
"loss": 0.3762,
|
|
"step": 7549
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.0621482382174863e-08,
|
|
"loss": 0.4405,
|
|
"step": 7550
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.0427520508085042e-08,
|
|
"loss": 0.4836,
|
|
"step": 7551
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.0235345063671632e-08,
|
|
"loss": 0.4579,
|
|
"step": 7552
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 1.0044956083300472e-08,
|
|
"loss": 0.4388,
|
|
"step": 7553
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 9.856353601018775e-09,
|
|
"loss": 0.4851,
|
|
"step": 7554
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 9.669537650554006e-09,
|
|
"loss": 0.4746,
|
|
"step": 7555
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 9.484508265314996e-09,
|
|
"loss": 0.4611,
|
|
"step": 7556
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 9.30126547838861e-09,
|
|
"loss": 0.4416,
|
|
"step": 7557
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 9.119809322546413e-09,
|
|
"loss": 0.4165,
|
|
"step": 7558
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 8.940139830236893e-09,
|
|
"loss": 0.5094,
|
|
"step": 7559
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 8.762257033591016e-09,
|
|
"loss": 0.4424,
|
|
"step": 7560
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 8.586160964417779e-09,
|
|
"loss": 0.3938,
|
|
"step": 7561
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 8.411851654209768e-09,
|
|
"loss": 0.428,
|
|
"step": 7562
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 8.239329134138718e-09,
|
|
"loss": 0.4744,
|
|
"step": 7563
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 8.068593435055504e-09,
|
|
"loss": 0.4538,
|
|
"step": 7564
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 7.899644587494593e-09,
|
|
"loss": 0.4844,
|
|
"step": 7565
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 7.73248262166737e-09,
|
|
"loss": 0.448,
|
|
"step": 7566
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 7.567107567467702e-09,
|
|
"loss": 0.445,
|
|
"step": 7567
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 7.40351945446971e-09,
|
|
"loss": 0.4252,
|
|
"step": 7568
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 7.24171831192666e-09,
|
|
"loss": 0.4209,
|
|
"step": 7569
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 7.081704168776515e-09,
|
|
"loss": 0.4566,
|
|
"step": 7570
|
|
},
|
|
{
|
|
"epoch": 2.96,
|
|
"learning_rate": 6.923477053630834e-09,
|
|
"loss": 0.4433,
|
|
"step": 7571
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 6.76703699478809e-09,
|
|
"loss": 0.4801,
|
|
"step": 7572
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 6.612384020222573e-09,
|
|
"loss": 0.4353,
|
|
"step": 7573
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 6.4595181575910496e-09,
|
|
"loss": 0.4591,
|
|
"step": 7574
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 6.30843943423054e-09,
|
|
"loss": 0.4678,
|
|
"step": 7575
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 6.159147877159432e-09,
|
|
"loss": 0.4548,
|
|
"step": 7576
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 6.0116435130741504e-09,
|
|
"loss": 0.4566,
|
|
"step": 7577
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 5.865926368353592e-09,
|
|
"loss": 0.4419,
|
|
"step": 7578
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 5.7219964690558015e-09,
|
|
"loss": 0.4472,
|
|
"step": 7579
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 5.5798538409201905e-09,
|
|
"loss": 0.4719,
|
|
"step": 7580
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 5.439498509365315e-09,
|
|
"loss": 0.4165,
|
|
"step": 7581
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 5.300930499491097e-09,
|
|
"loss": 0.4743,
|
|
"step": 7582
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 5.1641498360777146e-09,
|
|
"loss": 0.4767,
|
|
"step": 7583
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 5.029156543585601e-09,
|
|
"loss": 0.4082,
|
|
"step": 7584
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.895950646155445e-09,
|
|
"loss": 0.4644,
|
|
"step": 7585
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.7645321676081934e-09,
|
|
"loss": 0.528,
|
|
"step": 7586
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.634901131446157e-09,
|
|
"loss": 0.4687,
|
|
"step": 7587
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.507057560850792e-09,
|
|
"loss": 0.4506,
|
|
"step": 7588
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.3810014786827004e-09,
|
|
"loss": 0.3688,
|
|
"step": 7589
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.256732907487182e-09,
|
|
"loss": 0.4808,
|
|
"step": 7590
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.13425186948535e-09,
|
|
"loss": 0.4535,
|
|
"step": 7591
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 4.013558386580796e-09,
|
|
"loss": 0.5051,
|
|
"step": 7592
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 3.894652480357363e-09,
|
|
"loss": 0.443,
|
|
"step": 7593
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 3.777534172078046e-09,
|
|
"loss": 0.486,
|
|
"step": 7594
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 3.662203482688309e-09,
|
|
"loss": 0.5036,
|
|
"step": 7595
|
|
},
|
|
{
|
|
"epoch": 2.97,
|
|
"learning_rate": 3.548660432811657e-09,
|
|
"loss": 0.4742,
|
|
"step": 7596
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 3.4369050427540686e-09,
|
|
"loss": 0.4061,
|
|
"step": 7597
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 3.3269373324984477e-09,
|
|
"loss": 0.4193,
|
|
"step": 7598
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 3.218757321713506e-09,
|
|
"loss": 0.4833,
|
|
"step": 7599
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 3.1123650297415487e-09,
|
|
"loss": 0.4524,
|
|
"step": 7600
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 3.007760475611798e-09,
|
|
"loss": 0.4544,
|
|
"step": 7601
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.9049436780281825e-09,
|
|
"loss": 0.4335,
|
|
"step": 7602
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.803914655379325e-09,
|
|
"loss": 0.4175,
|
|
"step": 7603
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.704673425729665e-09,
|
|
"loss": 0.4265,
|
|
"step": 7604
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.607220006829447e-09,
|
|
"loss": 0.4468,
|
|
"step": 7605
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.5115544161047332e-09,
|
|
"loss": 0.4054,
|
|
"step": 7606
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.4176766706629494e-09,
|
|
"loss": 0.4977,
|
|
"step": 7607
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.3255867872928885e-09,
|
|
"loss": 0.4561,
|
|
"step": 7608
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.235284782462488e-09,
|
|
"loss": 0.4992,
|
|
"step": 7609
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.146770672319942e-09,
|
|
"loss": 0.4831,
|
|
"step": 7610
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 2.060044472695921e-09,
|
|
"loss": 0.4519,
|
|
"step": 7611
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.97510619909802e-09,
|
|
"loss": 0.4697,
|
|
"step": 7612
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.8919558667174208e-09,
|
|
"loss": 0.4661,
|
|
"step": 7613
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.8105934904211198e-09,
|
|
"loss": 0.4346,
|
|
"step": 7614
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.73101908476081e-09,
|
|
"loss": 0.4083,
|
|
"step": 7615
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.6532326639673302e-09,
|
|
"loss": 0.4257,
|
|
"step": 7616
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.5772342419506648e-09,
|
|
"loss": 0.4573,
|
|
"step": 7617
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.5030238323010537e-09,
|
|
"loss": 0.4819,
|
|
"step": 7618
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.4306014482889929e-09,
|
|
"loss": 0.4102,
|
|
"step": 7619
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.3599671028674544e-09,
|
|
"loss": 0.4368,
|
|
"step": 7620
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.2911208086663351e-09,
|
|
"loss": 0.4225,
|
|
"step": 7621
|
|
},
|
|
{
|
|
"epoch": 2.98,
|
|
"learning_rate": 1.2240625779980086e-09,
|
|
"loss": 0.458,
|
|
"step": 7622
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 1.158792422855104e-09,
|
|
"loss": 0.4266,
|
|
"step": 7623
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 1.0953103549093956e-09,
|
|
"loss": 0.4816,
|
|
"step": 7624
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 1.0336163855129143e-09,
|
|
"loss": 0.4575,
|
|
"step": 7625
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 9.737105256990564e-10,
|
|
"loss": 0.5447,
|
|
"step": 7626
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 9.155927861792535e-10,
|
|
"loss": 0.4704,
|
|
"step": 7627
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 8.592631773485238e-10,
|
|
"loss": 0.4492,
|
|
"step": 7628
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 8.047217092799209e-10,
|
|
"loss": 0.4538,
|
|
"step": 7629
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 7.519683917256437e-10,
|
|
"loss": 0.5122,
|
|
"step": 7630
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 7.010032341214778e-10,
|
|
"loss": 0.4161,
|
|
"step": 7631
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 6.518262455801339e-10,
|
|
"loss": 0.4443,
|
|
"step": 7632
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 6.044374348967986e-10,
|
|
"loss": 0.4288,
|
|
"step": 7633
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 5.588368105446939e-10,
|
|
"loss": 0.4408,
|
|
"step": 7634
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 5.150243806795186e-10,
|
|
"loss": 0.3932,
|
|
"step": 7635
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 4.730001531361161e-10,
|
|
"loss": 0.4281,
|
|
"step": 7636
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 4.3276413543069664e-10,
|
|
"loss": 0.4107,
|
|
"step": 7637
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 3.9431633475639495e-10,
|
|
"loss": 0.4446,
|
|
"step": 7638
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 3.576567579899326e-10,
|
|
"loss": 0.4666,
|
|
"step": 7639
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 3.2278541168717647e-10,
|
|
"loss": 0.4515,
|
|
"step": 7640
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 2.8970230208535955e-10,
|
|
"loss": 0.4408,
|
|
"step": 7641
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 2.584074350986399e-10,
|
|
"loss": 0.4324,
|
|
"step": 7642
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 2.289008163236517e-10,
|
|
"loss": 0.5067,
|
|
"step": 7643
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 2.011824510383953e-10,
|
|
"loss": 0.4065,
|
|
"step": 7644
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 1.7525234420001646e-10,
|
|
"loss": 0.4269,
|
|
"step": 7645
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 1.5111050044369636e-10,
|
|
"loss": 0.4473,
|
|
"step": 7646
|
|
},
|
|
{
|
|
"epoch": 2.99,
|
|
"learning_rate": 1.2875692408709228e-10,
|
|
"loss": 0.4106,
|
|
"step": 7647
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 1.081916191292276e-10,
|
|
"loss": 0.5294,
|
|
"step": 7648
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 8.941458924605073e-11,
|
|
"loss": 0.458,
|
|
"step": 7649
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 7.242583779598634e-11,
|
|
"loss": 0.492,
|
|
"step": 7650
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 5.7225367817714816e-11,
|
|
"loss": 0.4503,
|
|
"step": 7651
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 4.381318203017237e-11,
|
|
"loss": 0.4085,
|
|
"step": 7652
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 3.218928282922029e-11,
|
|
"loss": 0.439,
|
|
"step": 7653
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 2.2353672295416517e-11,
|
|
"loss": 0.3916,
|
|
"step": 7654
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 1.4306352188464546e-11,
|
|
"loss": 0.4465,
|
|
"step": 7655
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 8.047323944992968e-12,
|
|
"loss": 0.4825,
|
|
"step": 7656
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 3.576588686327043e-12,
|
|
"loss": 0.4082,
|
|
"step": 7657
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 8.941472118273453e-13,
|
|
"loss": 0.4744,
|
|
"step": 7658
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"learning_rate": 0.0,
|
|
"loss": 0.4382,
|
|
"step": 7659
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"step": 7659,
|
|
"total_flos": 9.549714459686076e+18,
|
|
"train_loss": 0.8274361214075563,
|
|
"train_runtime": 96259.3361,
|
|
"train_samples_per_second": 20.369,
|
|
"train_steps_per_second": 0.08
|
|
}
|
|
],
|
|
"max_steps": 7659,
|
|
"num_train_epochs": 3,
|
|
"total_flos": 9.549714459686076e+18,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|