29440 lines
819 KiB
JSON
29440 lines
819 KiB
JSON
{
|
|
"best_global_step": null,
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 7.0,
|
|
"eval_steps": 500,
|
|
"global_step": 13363,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"epoch": 0.0026191723415400735,
|
|
"grad_norm": 14.360649611601202,
|
|
"learning_rate": 1.1967090501121915e-07,
|
|
"loss": 0.8127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.34690144658088684,
|
|
"step": 5,
|
|
"valid_targets_mean": 1236.5,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 0.005238344683080147,
|
|
"grad_norm": 15.365253914310367,
|
|
"learning_rate": 2.692595362752431e-07,
|
|
"loss": 0.8159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.4388588070869446,
|
|
"step": 10,
|
|
"valid_targets_mean": 1550.8,
|
|
"valid_targets_min": 612
|
|
},
|
|
{
|
|
"epoch": 0.00785751702462022,
|
|
"grad_norm": 14.009337806771104,
|
|
"learning_rate": 4.1884816753926706e-07,
|
|
"loss": 0.7913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.37569183111190796,
|
|
"step": 15,
|
|
"valid_targets_mean": 1455.8,
|
|
"valid_targets_min": 904
|
|
},
|
|
{
|
|
"epoch": 0.010476689366160294,
|
|
"grad_norm": 13.648701274661068,
|
|
"learning_rate": 5.68436798803291e-07,
|
|
"loss": 0.7905,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.4232363700866699,
|
|
"step": 20,
|
|
"valid_targets_mean": 1514.4,
|
|
"valid_targets_min": 967
|
|
},
|
|
{
|
|
"epoch": 0.013095861707700367,
|
|
"grad_norm": 12.665998950030927,
|
|
"learning_rate": 7.18025430067315e-07,
|
|
"loss": 0.7813,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.36753425002098083,
|
|
"step": 25,
|
|
"valid_targets_mean": 1500.2,
|
|
"valid_targets_min": 718
|
|
},
|
|
{
|
|
"epoch": 0.01571503404924044,
|
|
"grad_norm": 10.58346863646541,
|
|
"learning_rate": 8.676140613313389e-07,
|
|
"loss": 0.7649,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.41818854212760925,
|
|
"step": 30,
|
|
"valid_targets_mean": 1689.9,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 0.018334206390780514,
|
|
"grad_norm": 10.066660675934825,
|
|
"learning_rate": 1.0172026925953627e-06,
|
|
"loss": 0.7398,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.37240898609161377,
|
|
"step": 35,
|
|
"valid_targets_mean": 1543.9,
|
|
"valid_targets_min": 725
|
|
},
|
|
{
|
|
"epoch": 0.020953378732320588,
|
|
"grad_norm": 8.335114097477899,
|
|
"learning_rate": 1.1667913238593869e-06,
|
|
"loss": 0.7298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.3356940448284149,
|
|
"step": 40,
|
|
"valid_targets_mean": 1290.1,
|
|
"valid_targets_min": 610
|
|
},
|
|
{
|
|
"epoch": 0.02357255107386066,
|
|
"grad_norm": 7.3101895788801965,
|
|
"learning_rate": 1.3163799551234108e-06,
|
|
"loss": 0.6865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.3337023854255676,
|
|
"step": 45,
|
|
"valid_targets_mean": 1588.6,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 0.026191723415400735,
|
|
"grad_norm": 6.5252803076428245,
|
|
"learning_rate": 1.4659685863874346e-06,
|
|
"loss": 0.6516,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.37255507707595825,
|
|
"step": 50,
|
|
"valid_targets_mean": 1708.5,
|
|
"valid_targets_min": 865
|
|
},
|
|
{
|
|
"epoch": 0.028810895756940808,
|
|
"grad_norm": 4.800098717068578,
|
|
"learning_rate": 1.6155572176514587e-06,
|
|
"loss": 0.61,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.32511481642723083,
|
|
"step": 55,
|
|
"valid_targets_mean": 1502.1,
|
|
"valid_targets_min": 593
|
|
},
|
|
{
|
|
"epoch": 0.03143006809848088,
|
|
"grad_norm": 4.157839193676562,
|
|
"learning_rate": 1.7651458489154827e-06,
|
|
"loss": 0.5816,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.30514219403266907,
|
|
"step": 60,
|
|
"valid_targets_mean": 1667.9,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 0.03404924044002095,
|
|
"grad_norm": 4.1100198852951975,
|
|
"learning_rate": 1.9147344801795064e-06,
|
|
"loss": 0.5512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22679626941680908,
|
|
"step": 65,
|
|
"valid_targets_mean": 1233.4,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 0.03666841278156103,
|
|
"grad_norm": 2.7793207953264036,
|
|
"learning_rate": 2.0643231114435305e-06,
|
|
"loss": 0.5122,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2324710190296173,
|
|
"step": 70,
|
|
"valid_targets_mean": 1556.0,
|
|
"valid_targets_min": 968
|
|
},
|
|
{
|
|
"epoch": 0.0392875851231011,
|
|
"grad_norm": 2.1990435383084934,
|
|
"learning_rate": 2.2139117427075543e-06,
|
|
"loss": 0.4888,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2724093794822693,
|
|
"step": 75,
|
|
"valid_targets_mean": 1773.1,
|
|
"valid_targets_min": 936
|
|
},
|
|
{
|
|
"epoch": 0.041906757464641176,
|
|
"grad_norm": 1.9295286252565726,
|
|
"learning_rate": 2.3635003739715784e-06,
|
|
"loss": 0.4829,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21531279385089874,
|
|
"step": 80,
|
|
"valid_targets_mean": 1383.6,
|
|
"valid_targets_min": 728
|
|
},
|
|
{
|
|
"epoch": 0.044525929806181246,
|
|
"grad_norm": 1.4255276793624676,
|
|
"learning_rate": 2.513089005235602e-06,
|
|
"loss": 0.4682,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.25964421033859253,
|
|
"step": 85,
|
|
"valid_targets_mean": 2023.5,
|
|
"valid_targets_min": 1143
|
|
},
|
|
{
|
|
"epoch": 0.04714510214772132,
|
|
"grad_norm": 1.4715916655150219,
|
|
"learning_rate": 2.6626776364996263e-06,
|
|
"loss": 0.4534,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.25838708877563477,
|
|
"step": 90,
|
|
"valid_targets_mean": 1411.9,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 0.04976427448926139,
|
|
"grad_norm": 1.5444005353673762,
|
|
"learning_rate": 2.8122662677636505e-06,
|
|
"loss": 0.4287,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21245674788951874,
|
|
"step": 95,
|
|
"valid_targets_mean": 1393.2,
|
|
"valid_targets_min": 625
|
|
},
|
|
{
|
|
"epoch": 0.05238344683080147,
|
|
"grad_norm": 1.1994420345215826,
|
|
"learning_rate": 2.961854899027674e-06,
|
|
"loss": 0.4186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22726021707057953,
|
|
"step": 100,
|
|
"valid_targets_mean": 1949.8,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 0.05500261917234154,
|
|
"grad_norm": 1.3912983718478529,
|
|
"learning_rate": 3.111443530291698e-06,
|
|
"loss": 0.4149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2639816403388977,
|
|
"step": 105,
|
|
"valid_targets_mean": 2193.8,
|
|
"valid_targets_min": 891
|
|
},
|
|
{
|
|
"epoch": 0.057621791513881616,
|
|
"grad_norm": 1.1605250886514002,
|
|
"learning_rate": 3.261032161555722e-06,
|
|
"loss": 0.4063,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1941235363483429,
|
|
"step": 110,
|
|
"valid_targets_mean": 1569.0,
|
|
"valid_targets_min": 850
|
|
},
|
|
{
|
|
"epoch": 0.060240963855421686,
|
|
"grad_norm": 1.1823905718225025,
|
|
"learning_rate": 3.410620792819746e-06,
|
|
"loss": 0.3927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17495793104171753,
|
|
"step": 115,
|
|
"valid_targets_mean": 1384.2,
|
|
"valid_targets_min": 974
|
|
},
|
|
{
|
|
"epoch": 0.06286013619696176,
|
|
"grad_norm": 1.1442967844329546,
|
|
"learning_rate": 3.56020942408377e-06,
|
|
"loss": 0.3807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22913743555545807,
|
|
"step": 120,
|
|
"valid_targets_mean": 1502.9,
|
|
"valid_targets_min": 998
|
|
},
|
|
{
|
|
"epoch": 0.06547930853850184,
|
|
"grad_norm": 1.068888565570554,
|
|
"learning_rate": 3.709798055347794e-06,
|
|
"loss": 0.3723,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20114624500274658,
|
|
"step": 125,
|
|
"valid_targets_mean": 1675.6,
|
|
"valid_targets_min": 1047
|
|
},
|
|
{
|
|
"epoch": 0.0680984808800419,
|
|
"grad_norm": 1.1005607515131652,
|
|
"learning_rate": 3.859386686611818e-06,
|
|
"loss": 0.3641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18255731463432312,
|
|
"step": 130,
|
|
"valid_targets_mean": 1577.6,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 0.07071765322158198,
|
|
"grad_norm": 0.9822788022537852,
|
|
"learning_rate": 4.008975317875842e-06,
|
|
"loss": 0.3579,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19359880685806274,
|
|
"step": 135,
|
|
"valid_targets_mean": 1967.9,
|
|
"valid_targets_min": 1075
|
|
},
|
|
{
|
|
"epoch": 0.07333682556312206,
|
|
"grad_norm": 1.035772431448236,
|
|
"learning_rate": 4.158563949139865e-06,
|
|
"loss": 0.3573,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20526465773582458,
|
|
"step": 140,
|
|
"valid_targets_mean": 1872.6,
|
|
"valid_targets_min": 1317
|
|
},
|
|
{
|
|
"epoch": 0.07595599790466212,
|
|
"grad_norm": 1.0001037815145,
|
|
"learning_rate": 4.30815258040389e-06,
|
|
"loss": 0.3446,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18035730719566345,
|
|
"step": 145,
|
|
"valid_targets_mean": 1697.6,
|
|
"valid_targets_min": 957
|
|
},
|
|
{
|
|
"epoch": 0.0785751702462022,
|
|
"grad_norm": 0.9891303511553036,
|
|
"learning_rate": 4.457741211667914e-06,
|
|
"loss": 0.3512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14854729175567627,
|
|
"step": 150,
|
|
"valid_targets_mean": 1323.0,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 0.08119434258774227,
|
|
"grad_norm": 0.9882196378696704,
|
|
"learning_rate": 4.607329842931937e-06,
|
|
"loss": 0.3444,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14030331373214722,
|
|
"step": 155,
|
|
"valid_targets_mean": 1488.8,
|
|
"valid_targets_min": 1111
|
|
},
|
|
{
|
|
"epoch": 0.08381351492928235,
|
|
"grad_norm": 1.0048944644271929,
|
|
"learning_rate": 4.756918474195962e-06,
|
|
"loss": 0.3347,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16978579759597778,
|
|
"step": 160,
|
|
"valid_targets_mean": 1649.2,
|
|
"valid_targets_min": 1147
|
|
},
|
|
{
|
|
"epoch": 0.08643268727082241,
|
|
"grad_norm": 1.0229637320628204,
|
|
"learning_rate": 4.906507105459986e-06,
|
|
"loss": 0.3281,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16098302602767944,
|
|
"step": 165,
|
|
"valid_targets_mean": 1526.2,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 0.08905185961236249,
|
|
"grad_norm": 1.1203487718427658,
|
|
"learning_rate": 5.056095736724009e-06,
|
|
"loss": 0.3156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13853532075881958,
|
|
"step": 170,
|
|
"valid_targets_mean": 1219.6,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 0.09167103195390257,
|
|
"grad_norm": 1.0656343831689796,
|
|
"learning_rate": 5.205684367988033e-06,
|
|
"loss": 0.3199,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14839622378349304,
|
|
"step": 175,
|
|
"valid_targets_mean": 1574.5,
|
|
"valid_targets_min": 1192
|
|
},
|
|
{
|
|
"epoch": 0.09429020429544265,
|
|
"grad_norm": 1.0462503574281339,
|
|
"learning_rate": 5.355272999252057e-06,
|
|
"loss": 0.3214,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15112197399139404,
|
|
"step": 180,
|
|
"valid_targets_mean": 1496.1,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 0.09690937663698271,
|
|
"grad_norm": 1.0628653049567738,
|
|
"learning_rate": 5.504861630516081e-06,
|
|
"loss": 0.3216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1642725020647049,
|
|
"step": 185,
|
|
"valid_targets_mean": 1402.0,
|
|
"valid_targets_min": 657
|
|
},
|
|
{
|
|
"epoch": 0.09952854897852279,
|
|
"grad_norm": 1.1546554901174115,
|
|
"learning_rate": 5.654450261780105e-06,
|
|
"loss": 0.319,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1774584949016571,
|
|
"step": 190,
|
|
"valid_targets_mean": 1471.2,
|
|
"valid_targets_min": 880
|
|
},
|
|
{
|
|
"epoch": 0.10214772132006286,
|
|
"grad_norm": 1.4304632238319623,
|
|
"learning_rate": 5.80403889304413e-06,
|
|
"loss": 0.321,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1903408318758011,
|
|
"step": 195,
|
|
"valid_targets_mean": 1567.6,
|
|
"valid_targets_min": 1163
|
|
},
|
|
{
|
|
"epoch": 0.10476689366160294,
|
|
"grad_norm": 1.1266051731672357,
|
|
"learning_rate": 5.9536275243081535e-06,
|
|
"loss": 0.3091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15488669276237488,
|
|
"step": 200,
|
|
"valid_targets_mean": 1493.1,
|
|
"valid_targets_min": 998
|
|
},
|
|
{
|
|
"epoch": 0.107386066003143,
|
|
"grad_norm": 1.0963428815981444,
|
|
"learning_rate": 6.103216155572177e-06,
|
|
"loss": 0.3028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12211956083774567,
|
|
"step": 205,
|
|
"valid_targets_mean": 1214.1,
|
|
"valid_targets_min": 703
|
|
},
|
|
{
|
|
"epoch": 0.11000523834468308,
|
|
"grad_norm": 1.5598102888973384,
|
|
"learning_rate": 6.252804786836201e-06,
|
|
"loss": 0.3236,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15281397104263306,
|
|
"step": 210,
|
|
"valid_targets_mean": 1459.2,
|
|
"valid_targets_min": 934
|
|
},
|
|
{
|
|
"epoch": 0.11262441068622316,
|
|
"grad_norm": 0.9578105634940497,
|
|
"learning_rate": 6.402393418100225e-06,
|
|
"loss": 0.3268,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20284399390220642,
|
|
"step": 215,
|
|
"valid_targets_mean": 1754.4,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 0.11524358302776323,
|
|
"grad_norm": 1.0061815691692737,
|
|
"learning_rate": 6.5519820493642484e-06,
|
|
"loss": 0.3012,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1887899935245514,
|
|
"step": 220,
|
|
"valid_targets_mean": 1864.2,
|
|
"valid_targets_min": 1270
|
|
},
|
|
{
|
|
"epoch": 0.1178627553693033,
|
|
"grad_norm": 0.9847642087093494,
|
|
"learning_rate": 6.701570680628273e-06,
|
|
"loss": 0.3044,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.155894473195076,
|
|
"step": 225,
|
|
"valid_targets_mean": 1765.5,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 0.12048192771084337,
|
|
"grad_norm": 1.0572078354661676,
|
|
"learning_rate": 6.851159311892297e-06,
|
|
"loss": 0.3094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18833455443382263,
|
|
"step": 230,
|
|
"valid_targets_mean": 1876.4,
|
|
"valid_targets_min": 1172
|
|
},
|
|
{
|
|
"epoch": 0.12310110005238345,
|
|
"grad_norm": 1.0498469428136838,
|
|
"learning_rate": 7.0007479431563205e-06,
|
|
"loss": 0.2971,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15638378262519836,
|
|
"step": 235,
|
|
"valid_targets_mean": 1642.0,
|
|
"valid_targets_min": 1033
|
|
},
|
|
{
|
|
"epoch": 0.12572027239392353,
|
|
"grad_norm": 0.998808766771215,
|
|
"learning_rate": 7.150336574420344e-06,
|
|
"loss": 0.2892,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16653478145599365,
|
|
"step": 240,
|
|
"valid_targets_mean": 1689.0,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 0.1283394447354636,
|
|
"grad_norm": 1.2353667197688452,
|
|
"learning_rate": 7.299925205684369e-06,
|
|
"loss": 0.2891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14583265781402588,
|
|
"step": 245,
|
|
"valid_targets_mean": 1164.2,
|
|
"valid_targets_min": 737
|
|
},
|
|
{
|
|
"epoch": 0.13095861707700368,
|
|
"grad_norm": 1.062022381972708,
|
|
"learning_rate": 7.4495138369483925e-06,
|
|
"loss": 0.3065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16454707086086273,
|
|
"step": 250,
|
|
"valid_targets_mean": 1427.9,
|
|
"valid_targets_min": 791
|
|
},
|
|
{
|
|
"epoch": 0.13357778941854373,
|
|
"grad_norm": 0.8910358787837693,
|
|
"learning_rate": 7.599102468212416e-06,
|
|
"loss": 0.295,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.128499835729599,
|
|
"step": 255,
|
|
"valid_targets_mean": 1559.5,
|
|
"valid_targets_min": 786
|
|
},
|
|
{
|
|
"epoch": 0.1361969617600838,
|
|
"grad_norm": 1.159401017442788,
|
|
"learning_rate": 7.748691099476442e-06,
|
|
"loss": 0.2876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12907026708126068,
|
|
"step": 260,
|
|
"valid_targets_mean": 1411.5,
|
|
"valid_targets_min": 944
|
|
},
|
|
{
|
|
"epoch": 0.13881613410162388,
|
|
"grad_norm": 1.244895408308553,
|
|
"learning_rate": 7.898279730740465e-06,
|
|
"loss": 0.2934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1402992308139801,
|
|
"step": 265,
|
|
"valid_targets_mean": 1454.9,
|
|
"valid_targets_min": 1227
|
|
},
|
|
{
|
|
"epoch": 0.14143530644316396,
|
|
"grad_norm": 1.1306976178593087,
|
|
"learning_rate": 8.047868362004489e-06,
|
|
"loss": 0.2877,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1254289448261261,
|
|
"step": 270,
|
|
"valid_targets_mean": 1446.9,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 0.14405447878470404,
|
|
"grad_norm": 0.921550922645942,
|
|
"learning_rate": 8.197456993268512e-06,
|
|
"loss": 0.2892,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16916373372077942,
|
|
"step": 275,
|
|
"valid_targets_mean": 2122.9,
|
|
"valid_targets_min": 1227
|
|
},
|
|
{
|
|
"epoch": 0.14667365112624411,
|
|
"grad_norm": 0.9894093776769832,
|
|
"learning_rate": 8.347045624532537e-06,
|
|
"loss": 0.2832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12109673023223877,
|
|
"step": 280,
|
|
"valid_targets_mean": 1256.0,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 0.1492928234677842,
|
|
"grad_norm": 1.0629135926381452,
|
|
"learning_rate": 8.49663425579656e-06,
|
|
"loss": 0.2795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09688296169042587,
|
|
"step": 285,
|
|
"valid_targets_mean": 1168.8,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 0.15191199580932424,
|
|
"grad_norm": 1.0185587649609436,
|
|
"learning_rate": 8.646222887060584e-06,
|
|
"loss": 0.2695,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13260769844055176,
|
|
"step": 290,
|
|
"valid_targets_mean": 1275.9,
|
|
"valid_targets_min": 911
|
|
},
|
|
{
|
|
"epoch": 0.15453116815086432,
|
|
"grad_norm": 0.9895587940966053,
|
|
"learning_rate": 8.795811518324609e-06,
|
|
"loss": 0.2781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12507134675979614,
|
|
"step": 295,
|
|
"valid_targets_mean": 1431.9,
|
|
"valid_targets_min": 661
|
|
},
|
|
{
|
|
"epoch": 0.1571503404924044,
|
|
"grad_norm": 1.266883417746457,
|
|
"learning_rate": 8.945400149588632e-06,
|
|
"loss": 0.2794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13883736729621887,
|
|
"step": 300,
|
|
"valid_targets_mean": 1572.1,
|
|
"valid_targets_min": 1164
|
|
},
|
|
{
|
|
"epoch": 0.15976951283394447,
|
|
"grad_norm": 1.1372606163472676,
|
|
"learning_rate": 9.094988780852656e-06,
|
|
"loss": 0.2734,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12079863250255585,
|
|
"step": 305,
|
|
"valid_targets_mean": 1396.6,
|
|
"valid_targets_min": 692
|
|
},
|
|
{
|
|
"epoch": 0.16238868517548455,
|
|
"grad_norm": 1.080445548289665,
|
|
"learning_rate": 9.244577412116679e-06,
|
|
"loss": 0.2777,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12431149184703827,
|
|
"step": 310,
|
|
"valid_targets_mean": 1405.8,
|
|
"valid_targets_min": 923
|
|
},
|
|
{
|
|
"epoch": 0.16500785751702463,
|
|
"grad_norm": 1.0979789118461285,
|
|
"learning_rate": 9.394166043380704e-06,
|
|
"loss": 0.2808,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1450071632862091,
|
|
"step": 315,
|
|
"valid_targets_mean": 1375.1,
|
|
"valid_targets_min": 922
|
|
},
|
|
{
|
|
"epoch": 0.1676270298585647,
|
|
"grad_norm": 0.9732855691376677,
|
|
"learning_rate": 9.543754674644728e-06,
|
|
"loss": 0.2692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14313220977783203,
|
|
"step": 320,
|
|
"valid_targets_mean": 1806.0,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 0.17024620220010478,
|
|
"grad_norm": 1.0693553307142176,
|
|
"learning_rate": 9.693343305908753e-06,
|
|
"loss": 0.2795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1343768686056137,
|
|
"step": 325,
|
|
"valid_targets_mean": 1342.9,
|
|
"valid_targets_min": 836
|
|
},
|
|
{
|
|
"epoch": 0.17286537454164483,
|
|
"grad_norm": 1.2383461126696775,
|
|
"learning_rate": 9.842931937172776e-06,
|
|
"loss": 0.2761,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15635260939598083,
|
|
"step": 330,
|
|
"valid_targets_mean": 1504.5,
|
|
"valid_targets_min": 1013
|
|
},
|
|
{
|
|
"epoch": 0.1754845468831849,
|
|
"grad_norm": 1.0278678326366064,
|
|
"learning_rate": 9.9925205684368e-06,
|
|
"loss": 0.2889,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1264498382806778,
|
|
"step": 335,
|
|
"valid_targets_mean": 1351.6,
|
|
"valid_targets_min": 978
|
|
},
|
|
{
|
|
"epoch": 0.17810371922472498,
|
|
"grad_norm": 0.9446943807345949,
|
|
"learning_rate": 1.0142109199700823e-05,
|
|
"loss": 0.2776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15977764129638672,
|
|
"step": 340,
|
|
"valid_targets_mean": 1812.5,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 0.18072289156626506,
|
|
"grad_norm": 1.0152657984078985,
|
|
"learning_rate": 1.029169783096485e-05,
|
|
"loss": 0.2904,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12069070339202881,
|
|
"step": 345,
|
|
"valid_targets_mean": 1657.9,
|
|
"valid_targets_min": 1219
|
|
},
|
|
{
|
|
"epoch": 0.18334206390780514,
|
|
"grad_norm": 1.1298475109493238,
|
|
"learning_rate": 1.0441286462228872e-05,
|
|
"loss": 0.2796,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17380760610103607,
|
|
"step": 350,
|
|
"valid_targets_mean": 1814.4,
|
|
"valid_targets_min": 1002
|
|
},
|
|
{
|
|
"epoch": 0.1859612362493452,
|
|
"grad_norm": 1.0284732532049468,
|
|
"learning_rate": 1.0590875093492897e-05,
|
|
"loss": 0.2747,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1345786154270172,
|
|
"step": 355,
|
|
"valid_targets_mean": 1390.4,
|
|
"valid_targets_min": 646
|
|
},
|
|
{
|
|
"epoch": 0.1885804085908853,
|
|
"grad_norm": 1.2878881994741163,
|
|
"learning_rate": 1.074046372475692e-05,
|
|
"loss": 0.2733,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15683802962303162,
|
|
"step": 360,
|
|
"valid_targets_mean": 1307.4,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 0.19119958093242537,
|
|
"grad_norm": 0.9809027495356322,
|
|
"learning_rate": 1.0890052356020944e-05,
|
|
"loss": 0.2663,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14160452783107758,
|
|
"step": 365,
|
|
"valid_targets_mean": 1574.8,
|
|
"valid_targets_min": 808
|
|
},
|
|
{
|
|
"epoch": 0.19381875327396542,
|
|
"grad_norm": 1.1326103312750746,
|
|
"learning_rate": 1.1039640987284967e-05,
|
|
"loss": 0.2802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1289289891719818,
|
|
"step": 370,
|
|
"valid_targets_mean": 1173.9,
|
|
"valid_targets_min": 913
|
|
},
|
|
{
|
|
"epoch": 0.1964379256155055,
|
|
"grad_norm": 1.1510183408232784,
|
|
"learning_rate": 1.1189229618548992e-05,
|
|
"loss": 0.2682,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1337447464466095,
|
|
"step": 375,
|
|
"valid_targets_mean": 1354.2,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 0.19905709795704557,
|
|
"grad_norm": 1.0246072291580883,
|
|
"learning_rate": 1.1338818249813015e-05,
|
|
"loss": 0.2615,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12531915307044983,
|
|
"step": 380,
|
|
"valid_targets_mean": 1508.4,
|
|
"valid_targets_min": 1013
|
|
},
|
|
{
|
|
"epoch": 0.20167627029858565,
|
|
"grad_norm": 1.0553388175979033,
|
|
"learning_rate": 1.148840688107704e-05,
|
|
"loss": 0.2544,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12561854720115662,
|
|
"step": 385,
|
|
"valid_targets_mean": 1517.8,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 0.20429544264012572,
|
|
"grad_norm": 1.1033328804149476,
|
|
"learning_rate": 1.1637995512341062e-05,
|
|
"loss": 0.2812,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11696565896272659,
|
|
"step": 390,
|
|
"valid_targets_mean": 1243.6,
|
|
"valid_targets_min": 548
|
|
},
|
|
{
|
|
"epoch": 0.2069146149816658,
|
|
"grad_norm": 1.4355990728777048,
|
|
"learning_rate": 1.1787584143605087e-05,
|
|
"loss": 0.2564,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12579873204231262,
|
|
"step": 395,
|
|
"valid_targets_mean": 1269.0,
|
|
"valid_targets_min": 694
|
|
},
|
|
{
|
|
"epoch": 0.20953378732320588,
|
|
"grad_norm": 0.9410901318700186,
|
|
"learning_rate": 1.193717277486911e-05,
|
|
"loss": 0.2662,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15521228313446045,
|
|
"step": 400,
|
|
"valid_targets_mean": 2141.0,
|
|
"valid_targets_min": 1030
|
|
},
|
|
{
|
|
"epoch": 0.21215295966474593,
|
|
"grad_norm": 0.8804587710612956,
|
|
"learning_rate": 1.2086761406133134e-05,
|
|
"loss": 0.2429,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11668536067008972,
|
|
"step": 405,
|
|
"valid_targets_mean": 1738.6,
|
|
"valid_targets_min": 682
|
|
},
|
|
{
|
|
"epoch": 0.214772132006286,
|
|
"grad_norm": 1.124066683984815,
|
|
"learning_rate": 1.2236350037397157e-05,
|
|
"loss": 0.2532,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13496944308280945,
|
|
"step": 410,
|
|
"valid_targets_mean": 1407.2,
|
|
"valid_targets_min": 990
|
|
},
|
|
{
|
|
"epoch": 0.21739130434782608,
|
|
"grad_norm": 0.9627302938369182,
|
|
"learning_rate": 1.2385938668661183e-05,
|
|
"loss": 0.2605,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11118908226490021,
|
|
"step": 415,
|
|
"valid_targets_mean": 1551.6,
|
|
"valid_targets_min": 639
|
|
},
|
|
{
|
|
"epoch": 0.22001047668936616,
|
|
"grad_norm": 1.0027093220955372,
|
|
"learning_rate": 1.2535527299925208e-05,
|
|
"loss": 0.265,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11878924071788788,
|
|
"step": 420,
|
|
"valid_targets_mean": 1442.6,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 0.22262964903090623,
|
|
"grad_norm": 1.093327589386835,
|
|
"learning_rate": 1.268511593118923e-05,
|
|
"loss": 0.2583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13007187843322754,
|
|
"step": 425,
|
|
"valid_targets_mean": 1426.5,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 0.2252488213724463,
|
|
"grad_norm": 1.056589541047298,
|
|
"learning_rate": 1.2834704562453255e-05,
|
|
"loss": 0.2627,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12732923030853271,
|
|
"step": 430,
|
|
"valid_targets_mean": 1298.2,
|
|
"valid_targets_min": 557
|
|
},
|
|
{
|
|
"epoch": 0.2278679937139864,
|
|
"grad_norm": 1.12863022898657,
|
|
"learning_rate": 1.2984293193717278e-05,
|
|
"loss": 0.2672,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1605558842420578,
|
|
"step": 435,
|
|
"valid_targets_mean": 1472.2,
|
|
"valid_targets_min": 713
|
|
},
|
|
{
|
|
"epoch": 0.23048716605552647,
|
|
"grad_norm": 0.9272349364850403,
|
|
"learning_rate": 1.3133881824981303e-05,
|
|
"loss": 0.2744,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20511457324028015,
|
|
"step": 440,
|
|
"valid_targets_mean": 3139.9,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 0.23310633839706651,
|
|
"grad_norm": 0.6592303408303113,
|
|
"learning_rate": 1.3283470456245326e-05,
|
|
"loss": 0.3305,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16108927130699158,
|
|
"step": 445,
|
|
"valid_targets_mean": 7612.6,
|
|
"valid_targets_min": 5111
|
|
},
|
|
{
|
|
"epoch": 0.2357255107386066,
|
|
"grad_norm": 0.6070904717684076,
|
|
"learning_rate": 1.343305908750935e-05,
|
|
"loss": 0.309,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15576837956905365,
|
|
"step": 450,
|
|
"valid_targets_mean": 6669.0,
|
|
"valid_targets_min": 4480
|
|
},
|
|
{
|
|
"epoch": 0.23834468308014667,
|
|
"grad_norm": 0.5605588322150622,
|
|
"learning_rate": 1.3582647718773375e-05,
|
|
"loss": 0.3049,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14649811387062073,
|
|
"step": 455,
|
|
"valid_targets_mean": 6538.6,
|
|
"valid_targets_min": 4620
|
|
},
|
|
{
|
|
"epoch": 0.24096385542168675,
|
|
"grad_norm": 0.5393212034322579,
|
|
"learning_rate": 1.3732236350037398e-05,
|
|
"loss": 0.2975,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15623930096626282,
|
|
"step": 460,
|
|
"valid_targets_mean": 6802.6,
|
|
"valid_targets_min": 4873
|
|
},
|
|
{
|
|
"epoch": 0.24358302776322682,
|
|
"grad_norm": 0.5656537757646136,
|
|
"learning_rate": 1.3881824981301422e-05,
|
|
"loss": 0.3118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.180804044008255,
|
|
"step": 465,
|
|
"valid_targets_mean": 7563.2,
|
|
"valid_targets_min": 4229
|
|
},
|
|
{
|
|
"epoch": 0.2462022001047669,
|
|
"grad_norm": 0.6043634022436999,
|
|
"learning_rate": 1.4031413612565445e-05,
|
|
"loss": 0.2989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16554348170757294,
|
|
"step": 470,
|
|
"valid_targets_mean": 6109.0,
|
|
"valid_targets_min": 4279
|
|
},
|
|
{
|
|
"epoch": 0.24882137244630698,
|
|
"grad_norm": 0.5674853412227062,
|
|
"learning_rate": 1.418100224382947e-05,
|
|
"loss": 0.2798,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11345628648996353,
|
|
"step": 475,
|
|
"valid_targets_mean": 5549.6,
|
|
"valid_targets_min": 4542
|
|
},
|
|
{
|
|
"epoch": 0.25144054478784705,
|
|
"grad_norm": 0.7434256656586965,
|
|
"learning_rate": 1.4330590875093493e-05,
|
|
"loss": 0.3047,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12726938724517822,
|
|
"step": 480,
|
|
"valid_targets_mean": 2525.8,
|
|
"valid_targets_min": 1497
|
|
},
|
|
{
|
|
"epoch": 0.25405971712938713,
|
|
"grad_norm": 0.5220549071443945,
|
|
"learning_rate": 1.4480179506357519e-05,
|
|
"loss": 0.2726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14162325859069824,
|
|
"step": 485,
|
|
"valid_targets_mean": 6166.2,
|
|
"valid_targets_min": 5332
|
|
},
|
|
{
|
|
"epoch": 0.2566788894709272,
|
|
"grad_norm": 0.5352536558586826,
|
|
"learning_rate": 1.4629768137621543e-05,
|
|
"loss": 0.2711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1418171525001526,
|
|
"step": 490,
|
|
"valid_targets_mean": 5823.5,
|
|
"valid_targets_min": 4612
|
|
},
|
|
{
|
|
"epoch": 0.2592980618124673,
|
|
"grad_norm": 0.49634651905061267,
|
|
"learning_rate": 1.4779356768885566e-05,
|
|
"loss": 0.2801,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1447383463382721,
|
|
"step": 495,
|
|
"valid_targets_mean": 6896.5,
|
|
"valid_targets_min": 4491
|
|
},
|
|
{
|
|
"epoch": 0.26191723415400736,
|
|
"grad_norm": 0.595603945724847,
|
|
"learning_rate": 1.4928945400149591e-05,
|
|
"loss": 0.2682,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13166537880897522,
|
|
"step": 500,
|
|
"valid_targets_mean": 6611.1,
|
|
"valid_targets_min": 4754
|
|
},
|
|
{
|
|
"epoch": 0.2645364064955474,
|
|
"grad_norm": 0.5393190418295071,
|
|
"learning_rate": 1.5078534031413614e-05,
|
|
"loss": 0.2597,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12661321461200714,
|
|
"step": 505,
|
|
"valid_targets_mean": 5907.1,
|
|
"valid_targets_min": 4787
|
|
},
|
|
{
|
|
"epoch": 0.26715557883708746,
|
|
"grad_norm": 0.4805967233196434,
|
|
"learning_rate": 1.5228122662677638e-05,
|
|
"loss": 0.2501,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12104909867048264,
|
|
"step": 510,
|
|
"valid_targets_mean": 6055.5,
|
|
"valid_targets_min": 5052
|
|
},
|
|
{
|
|
"epoch": 0.26977475117862754,
|
|
"grad_norm": 0.523906305492081,
|
|
"learning_rate": 1.5377711293941663e-05,
|
|
"loss": 0.2486,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11703436076641083,
|
|
"step": 515,
|
|
"valid_targets_mean": 6365.0,
|
|
"valid_targets_min": 4664
|
|
},
|
|
{
|
|
"epoch": 0.2723939235201676,
|
|
"grad_norm": 0.5224516826211744,
|
|
"learning_rate": 1.5527299925205686e-05,
|
|
"loss": 0.2656,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12053846567869186,
|
|
"step": 520,
|
|
"valid_targets_mean": 5928.2,
|
|
"valid_targets_min": 4211
|
|
},
|
|
{
|
|
"epoch": 0.2750130958617077,
|
|
"grad_norm": 0.48612267922315106,
|
|
"learning_rate": 1.567688855646971e-05,
|
|
"loss": 0.2612,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13059380650520325,
|
|
"step": 525,
|
|
"valid_targets_mean": 6415.9,
|
|
"valid_targets_min": 4582
|
|
},
|
|
{
|
|
"epoch": 0.27763226820324777,
|
|
"grad_norm": 0.5147679122113759,
|
|
"learning_rate": 1.582647718773373e-05,
|
|
"loss": 0.241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10836943238973618,
|
|
"step": 530,
|
|
"valid_targets_mean": 6044.8,
|
|
"valid_targets_min": 4800
|
|
},
|
|
{
|
|
"epoch": 0.28025144054478784,
|
|
"grad_norm": 0.4734791716055116,
|
|
"learning_rate": 1.5976065818997758e-05,
|
|
"loss": 0.2353,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11990270763635635,
|
|
"step": 535,
|
|
"valid_targets_mean": 8009.0,
|
|
"valid_targets_min": 6651
|
|
},
|
|
{
|
|
"epoch": 0.2828706128863279,
|
|
"grad_norm": 1.140668751349671,
|
|
"learning_rate": 1.612565445026178e-05,
|
|
"loss": 0.2596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1416817456483841,
|
|
"step": 540,
|
|
"valid_targets_mean": 5887.5,
|
|
"valid_targets_min": 3477
|
|
},
|
|
{
|
|
"epoch": 0.285489785227868,
|
|
"grad_norm": 0.5997001161199454,
|
|
"learning_rate": 1.6275243081525804e-05,
|
|
"loss": 0.2832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1452217996120453,
|
|
"step": 545,
|
|
"valid_targets_mean": 5471.1,
|
|
"valid_targets_min": 965
|
|
},
|
|
{
|
|
"epoch": 0.2881089575694081,
|
|
"grad_norm": 0.5265818655921063,
|
|
"learning_rate": 1.6424831712789827e-05,
|
|
"loss": 0.2413,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1338367462158203,
|
|
"step": 550,
|
|
"valid_targets_mean": 6654.1,
|
|
"valid_targets_min": 5226
|
|
},
|
|
{
|
|
"epoch": 0.29072812991094815,
|
|
"grad_norm": 0.5748870112445678,
|
|
"learning_rate": 1.6574420344053853e-05,
|
|
"loss": 0.2518,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10587285459041595,
|
|
"step": 555,
|
|
"valid_targets_mean": 6363.4,
|
|
"valid_targets_min": 4213
|
|
},
|
|
{
|
|
"epoch": 0.29334730225248823,
|
|
"grad_norm": 0.5572025930451594,
|
|
"learning_rate": 1.672400897531788e-05,
|
|
"loss": 0.27,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13715760409832,
|
|
"step": 560,
|
|
"valid_targets_mean": 5469.9,
|
|
"valid_targets_min": 3761
|
|
},
|
|
{
|
|
"epoch": 0.2959664745940283,
|
|
"grad_norm": 0.4978200163320225,
|
|
"learning_rate": 1.6873597606581902e-05,
|
|
"loss": 0.2871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16060104966163635,
|
|
"step": 565,
|
|
"valid_targets_mean": 7206.2,
|
|
"valid_targets_min": 4993
|
|
},
|
|
{
|
|
"epoch": 0.2985856469355684,
|
|
"grad_norm": 0.5876389734069661,
|
|
"learning_rate": 1.7023186237845925e-05,
|
|
"loss": 0.2815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1361798644065857,
|
|
"step": 570,
|
|
"valid_targets_mean": 6402.2,
|
|
"valid_targets_min": 5199
|
|
},
|
|
{
|
|
"epoch": 0.30120481927710846,
|
|
"grad_norm": 0.49390495353916386,
|
|
"learning_rate": 1.7172774869109948e-05,
|
|
"loss": 0.257,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1463567018508911,
|
|
"step": 575,
|
|
"valid_targets_mean": 7578.1,
|
|
"valid_targets_min": 4534
|
|
},
|
|
{
|
|
"epoch": 0.3038239916186485,
|
|
"grad_norm": 0.8190322032239803,
|
|
"learning_rate": 1.7322363500373974e-05,
|
|
"loss": 0.2559,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11292055994272232,
|
|
"step": 580,
|
|
"valid_targets_mean": 6264.4,
|
|
"valid_targets_min": 4751
|
|
},
|
|
{
|
|
"epoch": 0.30644316396018856,
|
|
"grad_norm": 0.5825495479534828,
|
|
"learning_rate": 1.7471952131637997e-05,
|
|
"loss": 0.2546,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11191920191049576,
|
|
"step": 585,
|
|
"valid_targets_mean": 5604.1,
|
|
"valid_targets_min": 3888
|
|
},
|
|
{
|
|
"epoch": 0.30906233630172864,
|
|
"grad_norm": 0.6079995425568043,
|
|
"learning_rate": 1.762154076290202e-05,
|
|
"loss": 0.2632,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14392173290252686,
|
|
"step": 590,
|
|
"valid_targets_mean": 7001.6,
|
|
"valid_targets_min": 4455
|
|
},
|
|
{
|
|
"epoch": 0.3116815086432687,
|
|
"grad_norm": 0.5099427304498049,
|
|
"learning_rate": 1.7771129394166046e-05,
|
|
"loss": 0.252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14200888574123383,
|
|
"step": 595,
|
|
"valid_targets_mean": 6955.5,
|
|
"valid_targets_min": 5230
|
|
},
|
|
{
|
|
"epoch": 0.3143006809848088,
|
|
"grad_norm": 0.5221532590692329,
|
|
"learning_rate": 1.792071802543007e-05,
|
|
"loss": 0.2576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11283424496650696,
|
|
"step": 600,
|
|
"valid_targets_mean": 7295.8,
|
|
"valid_targets_min": 5668
|
|
},
|
|
{
|
|
"epoch": 0.31691985332634887,
|
|
"grad_norm": 0.6000692376731359,
|
|
"learning_rate": 1.8070306656694092e-05,
|
|
"loss": 0.2546,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13104653358459473,
|
|
"step": 605,
|
|
"valid_targets_mean": 6350.4,
|
|
"valid_targets_min": 4871
|
|
},
|
|
{
|
|
"epoch": 0.31953902566788894,
|
|
"grad_norm": 0.5108942792231622,
|
|
"learning_rate": 1.8219895287958115e-05,
|
|
"loss": 0.2418,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1462014615535736,
|
|
"step": 610,
|
|
"valid_targets_mean": 8015.2,
|
|
"valid_targets_min": 6018
|
|
},
|
|
{
|
|
"epoch": 0.322158198009429,
|
|
"grad_norm": 0.4384117592659271,
|
|
"learning_rate": 1.836948391922214e-05,
|
|
"loss": 0.2207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11725406348705292,
|
|
"step": 615,
|
|
"valid_targets_mean": 7110.1,
|
|
"valid_targets_min": 3533
|
|
},
|
|
{
|
|
"epoch": 0.3247773703509691,
|
|
"grad_norm": 0.5213116308490782,
|
|
"learning_rate": 1.8519072550486164e-05,
|
|
"loss": 0.2543,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13383880257606506,
|
|
"step": 620,
|
|
"valid_targets_mean": 6875.0,
|
|
"valid_targets_min": 5199
|
|
},
|
|
{
|
|
"epoch": 0.3273965426925092,
|
|
"grad_norm": 0.5834532687839684,
|
|
"learning_rate": 1.866866118175019e-05,
|
|
"loss": 0.2551,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12733125686645508,
|
|
"step": 625,
|
|
"valid_targets_mean": 5568.4,
|
|
"valid_targets_min": 4923
|
|
},
|
|
{
|
|
"epoch": 0.33001571503404925,
|
|
"grad_norm": 0.519113028355725,
|
|
"learning_rate": 1.8818249813014213e-05,
|
|
"loss": 0.2494,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0986289530992508,
|
|
"step": 630,
|
|
"valid_targets_mean": 6421.9,
|
|
"valid_targets_min": 4527
|
|
},
|
|
{
|
|
"epoch": 0.3326348873755893,
|
|
"grad_norm": 0.8163787743694745,
|
|
"learning_rate": 1.8967838444278236e-05,
|
|
"loss": 0.2461,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1518421769142151,
|
|
"step": 635,
|
|
"valid_targets_mean": 7147.9,
|
|
"valid_targets_min": 3082
|
|
},
|
|
{
|
|
"epoch": 0.3352540597171294,
|
|
"grad_norm": 0.5104174282267507,
|
|
"learning_rate": 1.9117427075542262e-05,
|
|
"loss": 0.2262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11866091936826706,
|
|
"step": 640,
|
|
"valid_targets_mean": 6880.6,
|
|
"valid_targets_min": 4685
|
|
},
|
|
{
|
|
"epoch": 0.3378732320586695,
|
|
"grad_norm": 0.570898453985933,
|
|
"learning_rate": 1.9267015706806285e-05,
|
|
"loss": 0.2479,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11430183053016663,
|
|
"step": 645,
|
|
"valid_targets_mean": 5915.5,
|
|
"valid_targets_min": 4875
|
|
},
|
|
{
|
|
"epoch": 0.34049240440020956,
|
|
"grad_norm": 0.5480651974305227,
|
|
"learning_rate": 1.9416604338070308e-05,
|
|
"loss": 0.247,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1123688668012619,
|
|
"step": 650,
|
|
"valid_targets_mean": 6234.5,
|
|
"valid_targets_min": 4433
|
|
},
|
|
{
|
|
"epoch": 0.3431115767417496,
|
|
"grad_norm": 0.5442230880157303,
|
|
"learning_rate": 1.956619296933433e-05,
|
|
"loss": 0.2364,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10760192573070526,
|
|
"step": 655,
|
|
"valid_targets_mean": 5684.0,
|
|
"valid_targets_min": 4540
|
|
},
|
|
{
|
|
"epoch": 0.34573074908328966,
|
|
"grad_norm": 0.5346676203693183,
|
|
"learning_rate": 1.9715781600598357e-05,
|
|
"loss": 0.2514,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15074855089187622,
|
|
"step": 660,
|
|
"valid_targets_mean": 7353.4,
|
|
"valid_targets_min": 5101
|
|
},
|
|
{
|
|
"epoch": 0.34834992142482973,
|
|
"grad_norm": 0.5824277658679824,
|
|
"learning_rate": 1.986537023186238e-05,
|
|
"loss": 0.2413,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10114076733589172,
|
|
"step": 665,
|
|
"valid_targets_mean": 5300.5,
|
|
"valid_targets_min": 4442
|
|
},
|
|
{
|
|
"epoch": 0.3509690937663698,
|
|
"grad_norm": 0.4835835842335859,
|
|
"learning_rate": 2.0014958863126406e-05,
|
|
"loss": 0.2111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09848757833242416,
|
|
"step": 670,
|
|
"valid_targets_mean": 6084.1,
|
|
"valid_targets_min": 5268
|
|
},
|
|
{
|
|
"epoch": 0.3535882661079099,
|
|
"grad_norm": 0.48641986127969844,
|
|
"learning_rate": 2.016454749439043e-05,
|
|
"loss": 0.2142,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11454112082719803,
|
|
"step": 675,
|
|
"valid_targets_mean": 6446.6,
|
|
"valid_targets_min": 5074
|
|
},
|
|
{
|
|
"epoch": 0.35620743844944996,
|
|
"grad_norm": 0.4862581142799685,
|
|
"learning_rate": 2.0314136125654452e-05,
|
|
"loss": 0.2243,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10703176259994507,
|
|
"step": 680,
|
|
"valid_targets_mean": 6195.0,
|
|
"valid_targets_min": 5221
|
|
},
|
|
{
|
|
"epoch": 0.35882661079099004,
|
|
"grad_norm": 0.5438834169467284,
|
|
"learning_rate": 2.046372475691848e-05,
|
|
"loss": 0.2368,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11888057738542557,
|
|
"step": 685,
|
|
"valid_targets_mean": 5689.1,
|
|
"valid_targets_min": 4642
|
|
},
|
|
{
|
|
"epoch": 0.3614457831325301,
|
|
"grad_norm": 0.6833967212154707,
|
|
"learning_rate": 2.06133133881825e-05,
|
|
"loss": 0.2662,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13145872950553894,
|
|
"step": 690,
|
|
"valid_targets_mean": 4461.4,
|
|
"valid_targets_min": 3222
|
|
},
|
|
{
|
|
"epoch": 0.3640649554740702,
|
|
"grad_norm": 0.4757801544127953,
|
|
"learning_rate": 2.0762902019446524e-05,
|
|
"loss": 0.2772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12056447565555573,
|
|
"step": 695,
|
|
"valid_targets_mean": 6822.2,
|
|
"valid_targets_min": 4371
|
|
},
|
|
{
|
|
"epoch": 0.3666841278156103,
|
|
"grad_norm": 0.6379502088686836,
|
|
"learning_rate": 2.0912490650710547e-05,
|
|
"loss": 0.2297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12404558807611465,
|
|
"step": 700,
|
|
"valid_targets_mean": 5632.6,
|
|
"valid_targets_min": 3689
|
|
},
|
|
{
|
|
"epoch": 0.36930330015715035,
|
|
"grad_norm": 0.5476812114299602,
|
|
"learning_rate": 2.1062079281974573e-05,
|
|
"loss": 0.2392,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12419216334819794,
|
|
"step": 705,
|
|
"valid_targets_mean": 6364.1,
|
|
"valid_targets_min": 4281
|
|
},
|
|
{
|
|
"epoch": 0.3719224724986904,
|
|
"grad_norm": 0.5104513503681111,
|
|
"learning_rate": 2.1211667913238596e-05,
|
|
"loss": 0.2406,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1376412808895111,
|
|
"step": 710,
|
|
"valid_targets_mean": 6337.8,
|
|
"valid_targets_min": 4844
|
|
},
|
|
{
|
|
"epoch": 0.3745416448402305,
|
|
"grad_norm": 0.5204775395935077,
|
|
"learning_rate": 2.136125654450262e-05,
|
|
"loss": 0.2567,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10598008334636688,
|
|
"step": 715,
|
|
"valid_targets_mean": 5507.5,
|
|
"valid_targets_min": 4549
|
|
},
|
|
{
|
|
"epoch": 0.3771608171817706,
|
|
"grad_norm": 0.47648945431326273,
|
|
"learning_rate": 2.1510845175766645e-05,
|
|
"loss": 0.235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1064852625131607,
|
|
"step": 720,
|
|
"valid_targets_mean": 5274.8,
|
|
"valid_targets_min": 2461
|
|
},
|
|
{
|
|
"epoch": 0.37977998952331066,
|
|
"grad_norm": 0.6232540892767168,
|
|
"learning_rate": 2.1660433807030668e-05,
|
|
"loss": 0.245,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13343873620033264,
|
|
"step": 725,
|
|
"valid_targets_mean": 6465.2,
|
|
"valid_targets_min": 2815
|
|
},
|
|
{
|
|
"epoch": 0.38239916186485073,
|
|
"grad_norm": 0.4884239403527553,
|
|
"learning_rate": 2.181002243829469e-05,
|
|
"loss": 0.2363,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10519768297672272,
|
|
"step": 730,
|
|
"valid_targets_mean": 6715.5,
|
|
"valid_targets_min": 5192
|
|
},
|
|
{
|
|
"epoch": 0.38501833420639076,
|
|
"grad_norm": 0.48078334521709176,
|
|
"learning_rate": 2.1959611069558714e-05,
|
|
"loss": 0.2158,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10113941133022308,
|
|
"step": 735,
|
|
"valid_targets_mean": 5737.2,
|
|
"valid_targets_min": 5186
|
|
},
|
|
{
|
|
"epoch": 0.38763750654793083,
|
|
"grad_norm": 0.4914084122068677,
|
|
"learning_rate": 2.210919970082274e-05,
|
|
"loss": 0.2324,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09212394058704376,
|
|
"step": 740,
|
|
"valid_targets_mean": 5329.5,
|
|
"valid_targets_min": 3786
|
|
},
|
|
{
|
|
"epoch": 0.3902566788894709,
|
|
"grad_norm": 0.556681897033246,
|
|
"learning_rate": 2.2258788332086763e-05,
|
|
"loss": 0.2831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13327279686927795,
|
|
"step": 745,
|
|
"valid_targets_mean": 6233.8,
|
|
"valid_targets_min": 5147
|
|
},
|
|
{
|
|
"epoch": 0.392875851231011,
|
|
"grad_norm": 0.518177762534259,
|
|
"learning_rate": 2.2408376963350786e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13125228881835938,
|
|
"step": 750,
|
|
"valid_targets_mean": 5807.6,
|
|
"valid_targets_min": 3952
|
|
},
|
|
{
|
|
"epoch": 0.39549502357255106,
|
|
"grad_norm": 0.5591532007746006,
|
|
"learning_rate": 2.255796559461481e-05,
|
|
"loss": 0.2392,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13699626922607422,
|
|
"step": 755,
|
|
"valid_targets_mean": 6452.8,
|
|
"valid_targets_min": 5079
|
|
},
|
|
{
|
|
"epoch": 0.39811419591409114,
|
|
"grad_norm": 0.5270316309713508,
|
|
"learning_rate": 2.2707554225878835e-05,
|
|
"loss": 0.2452,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12873396277427673,
|
|
"step": 760,
|
|
"valid_targets_mean": 6432.9,
|
|
"valid_targets_min": 3762
|
|
},
|
|
{
|
|
"epoch": 0.4007333682556312,
|
|
"grad_norm": 0.46395489991397554,
|
|
"learning_rate": 2.2857142857142858e-05,
|
|
"loss": 0.2165,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1013445183634758,
|
|
"step": 765,
|
|
"valid_targets_mean": 5954.1,
|
|
"valid_targets_min": 4726
|
|
},
|
|
{
|
|
"epoch": 0.4033525405971713,
|
|
"grad_norm": 0.5726114581033401,
|
|
"learning_rate": 2.300673148840688e-05,
|
|
"loss": 0.2236,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12622445821762085,
|
|
"step": 770,
|
|
"valid_targets_mean": 5874.6,
|
|
"valid_targets_min": 3724
|
|
},
|
|
{
|
|
"epoch": 0.40597171293871137,
|
|
"grad_norm": 0.539701890419442,
|
|
"learning_rate": 2.3156320119670904e-05,
|
|
"loss": 0.2083,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10788456350564957,
|
|
"step": 775,
|
|
"valid_targets_mean": 6436.9,
|
|
"valid_targets_min": 5578
|
|
},
|
|
{
|
|
"epoch": 0.40859088528025145,
|
|
"grad_norm": 0.50097922607963,
|
|
"learning_rate": 2.330590875093493e-05,
|
|
"loss": 0.2296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10845942050218582,
|
|
"step": 780,
|
|
"valid_targets_mean": 6727.8,
|
|
"valid_targets_min": 5795
|
|
},
|
|
{
|
|
"epoch": 0.4112100576217915,
|
|
"grad_norm": 0.4773488140437983,
|
|
"learning_rate": 2.3455497382198953e-05,
|
|
"loss": 0.2207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09601639211177826,
|
|
"step": 785,
|
|
"valid_targets_mean": 6116.2,
|
|
"valid_targets_min": 5268
|
|
},
|
|
{
|
|
"epoch": 0.4138292299633316,
|
|
"grad_norm": 0.4195165425204695,
|
|
"learning_rate": 2.3605086013462976e-05,
|
|
"loss": 0.2224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10880352556705475,
|
|
"step": 790,
|
|
"valid_targets_mean": 7379.5,
|
|
"valid_targets_min": 4920
|
|
},
|
|
{
|
|
"epoch": 0.4164484023048717,
|
|
"grad_norm": 0.4306956855558924,
|
|
"learning_rate": 2.3754674644727006e-05,
|
|
"loss": 0.2288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08985568583011627,
|
|
"step": 795,
|
|
"valid_targets_mean": 7048.4,
|
|
"valid_targets_min": 5243
|
|
},
|
|
{
|
|
"epoch": 0.41906757464641176,
|
|
"grad_norm": 0.5369718687993797,
|
|
"learning_rate": 2.390426327599103e-05,
|
|
"loss": 0.2172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12296460568904877,
|
|
"step": 800,
|
|
"valid_targets_mean": 6385.8,
|
|
"valid_targets_min": 4794
|
|
},
|
|
{
|
|
"epoch": 0.42168674698795183,
|
|
"grad_norm": 0.4941586673987478,
|
|
"learning_rate": 2.405385190725505e-05,
|
|
"loss": 0.2392,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10740514099597931,
|
|
"step": 805,
|
|
"valid_targets_mean": 6215.4,
|
|
"valid_targets_min": 5168
|
|
},
|
|
{
|
|
"epoch": 0.42430591932949185,
|
|
"grad_norm": 0.582899412841253,
|
|
"learning_rate": 2.4203440538519078e-05,
|
|
"loss": 0.2465,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.138446643948555,
|
|
"step": 810,
|
|
"valid_targets_mean": 5333.2,
|
|
"valid_targets_min": 3130
|
|
},
|
|
{
|
|
"epoch": 0.42692509167103193,
|
|
"grad_norm": 0.7098374759246038,
|
|
"learning_rate": 2.43530291697831e-05,
|
|
"loss": 0.3222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17193308472633362,
|
|
"step": 815,
|
|
"valid_targets_mean": 5546.2,
|
|
"valid_targets_min": 1971
|
|
},
|
|
{
|
|
"epoch": 0.429544264012572,
|
|
"grad_norm": 0.7132928007433752,
|
|
"learning_rate": 2.4502617801047123e-05,
|
|
"loss": 0.3978,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2480137199163437,
|
|
"step": 820,
|
|
"valid_targets_mean": 5648.9,
|
|
"valid_targets_min": 1631
|
|
},
|
|
{
|
|
"epoch": 0.4321634363541121,
|
|
"grad_norm": 0.6728986690741761,
|
|
"learning_rate": 2.4652206432311146e-05,
|
|
"loss": 0.3876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22709880769252777,
|
|
"step": 825,
|
|
"valid_targets_mean": 5411.4,
|
|
"valid_targets_min": 1509
|
|
},
|
|
{
|
|
"epoch": 0.43478260869565216,
|
|
"grad_norm": 0.644504704154439,
|
|
"learning_rate": 2.4801795063575172e-05,
|
|
"loss": 0.3675,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1710299700498581,
|
|
"step": 830,
|
|
"valid_targets_mean": 4960.2,
|
|
"valid_targets_min": 2736
|
|
},
|
|
{
|
|
"epoch": 0.43740178103719224,
|
|
"grad_norm": 0.603144229645836,
|
|
"learning_rate": 2.4951383694839195e-05,
|
|
"loss": 0.3715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15948154032230377,
|
|
"step": 835,
|
|
"valid_targets_mean": 3726.8,
|
|
"valid_targets_min": 1439
|
|
},
|
|
{
|
|
"epoch": 0.4400209533787323,
|
|
"grad_norm": 0.6048927321251332,
|
|
"learning_rate": 2.5100972326103218e-05,
|
|
"loss": 0.3826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16860568523406982,
|
|
"step": 840,
|
|
"valid_targets_mean": 4269.4,
|
|
"valid_targets_min": 1838
|
|
},
|
|
{
|
|
"epoch": 0.4426401257202724,
|
|
"grad_norm": 0.5645343149320301,
|
|
"learning_rate": 2.525056095736724e-05,
|
|
"loss": 0.3535,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22065946459770203,
|
|
"step": 845,
|
|
"valid_targets_mean": 5973.6,
|
|
"valid_targets_min": 2762
|
|
},
|
|
{
|
|
"epoch": 0.44525929806181247,
|
|
"grad_norm": 0.5696632258416167,
|
|
"learning_rate": 2.5400149588631267e-05,
|
|
"loss": 0.3636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16765108704566956,
|
|
"step": 850,
|
|
"valid_targets_mean": 3987.1,
|
|
"valid_targets_min": 1654
|
|
},
|
|
{
|
|
"epoch": 0.44787847040335255,
|
|
"grad_norm": 0.5461627337763696,
|
|
"learning_rate": 2.554973821989529e-05,
|
|
"loss": 0.3634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1650678813457489,
|
|
"step": 855,
|
|
"valid_targets_mean": 5043.6,
|
|
"valid_targets_min": 2493
|
|
},
|
|
{
|
|
"epoch": 0.4504976427448926,
|
|
"grad_norm": 0.6356464309990801,
|
|
"learning_rate": 2.5699326851159313e-05,
|
|
"loss": 0.3755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17353945970535278,
|
|
"step": 860,
|
|
"valid_targets_mean": 4468.9,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 0.4531168150864327,
|
|
"grad_norm": 0.5446940060410611,
|
|
"learning_rate": 2.584891548242334e-05,
|
|
"loss": 0.3598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16391149163246155,
|
|
"step": 865,
|
|
"valid_targets_mean": 5241.0,
|
|
"valid_targets_min": 953
|
|
},
|
|
{
|
|
"epoch": 0.4557359874279728,
|
|
"grad_norm": 0.6074855741100244,
|
|
"learning_rate": 2.5998504113687362e-05,
|
|
"loss": 0.3662,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22667014598846436,
|
|
"step": 870,
|
|
"valid_targets_mean": 5277.0,
|
|
"valid_targets_min": 1265
|
|
},
|
|
{
|
|
"epoch": 0.45835515976951285,
|
|
"grad_norm": 0.5397370112483811,
|
|
"learning_rate": 2.6148092744951385e-05,
|
|
"loss": 0.357,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.189674511551857,
|
|
"step": 875,
|
|
"valid_targets_mean": 6127.8,
|
|
"valid_targets_min": 2566
|
|
},
|
|
{
|
|
"epoch": 0.46097433211105293,
|
|
"grad_norm": 0.6193870279812388,
|
|
"learning_rate": 2.6297681376215408e-05,
|
|
"loss": 0.3642,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1588752567768097,
|
|
"step": 880,
|
|
"valid_targets_mean": 3446.9,
|
|
"valid_targets_min": 1759
|
|
},
|
|
{
|
|
"epoch": 0.463593504452593,
|
|
"grad_norm": 0.7161305662391659,
|
|
"learning_rate": 2.6447270007479434e-05,
|
|
"loss": 0.3692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18031099438667297,
|
|
"step": 885,
|
|
"valid_targets_mean": 3659.4,
|
|
"valid_targets_min": 2179
|
|
},
|
|
{
|
|
"epoch": 0.46621267679413303,
|
|
"grad_norm": 0.6833309594222867,
|
|
"learning_rate": 2.6596858638743457e-05,
|
|
"loss": 0.3705,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19416679441928864,
|
|
"step": 890,
|
|
"valid_targets_mean": 4161.5,
|
|
"valid_targets_min": 1785
|
|
},
|
|
{
|
|
"epoch": 0.4688318491356731,
|
|
"grad_norm": 0.6686531726084138,
|
|
"learning_rate": 2.674644727000748e-05,
|
|
"loss": 0.3619,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21306106448173523,
|
|
"step": 895,
|
|
"valid_targets_mean": 4467.9,
|
|
"valid_targets_min": 2345
|
|
},
|
|
{
|
|
"epoch": 0.4714510214772132,
|
|
"grad_norm": 0.6514316529464631,
|
|
"learning_rate": 2.6896035901271503e-05,
|
|
"loss": 0.3779,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2033441960811615,
|
|
"step": 900,
|
|
"valid_targets_mean": 4057.2,
|
|
"valid_targets_min": 2767
|
|
},
|
|
{
|
|
"epoch": 0.47407019381875326,
|
|
"grad_norm": 0.7367261083208211,
|
|
"learning_rate": 2.704562453253553e-05,
|
|
"loss": 0.3782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19121892750263214,
|
|
"step": 905,
|
|
"valid_targets_mean": 3450.2,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 0.47668936616029334,
|
|
"grad_norm": 0.7078706067989683,
|
|
"learning_rate": 2.7195213163799552e-05,
|
|
"loss": 0.3673,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1769869178533554,
|
|
"step": 910,
|
|
"valid_targets_mean": 3777.1,
|
|
"valid_targets_min": 2004
|
|
},
|
|
{
|
|
"epoch": 0.4793085385018334,
|
|
"grad_norm": 0.6092132151273473,
|
|
"learning_rate": 2.7344801795063575e-05,
|
|
"loss": 0.3723,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1985096037387848,
|
|
"step": 915,
|
|
"valid_targets_mean": 4273.0,
|
|
"valid_targets_min": 1469
|
|
},
|
|
{
|
|
"epoch": 0.4819277108433735,
|
|
"grad_norm": 0.7040547478222255,
|
|
"learning_rate": 2.7494390426327598e-05,
|
|
"loss": 0.3514,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2201271951198578,
|
|
"step": 920,
|
|
"valid_targets_mean": 4196.8,
|
|
"valid_targets_min": 1889
|
|
},
|
|
{
|
|
"epoch": 0.48454688318491357,
|
|
"grad_norm": 0.6604000483479482,
|
|
"learning_rate": 2.7643979057591624e-05,
|
|
"loss": 0.3455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18622466921806335,
|
|
"step": 925,
|
|
"valid_targets_mean": 4319.2,
|
|
"valid_targets_min": 2046
|
|
},
|
|
{
|
|
"epoch": 0.48716605552645365,
|
|
"grad_norm": 0.6969880777230246,
|
|
"learning_rate": 2.7793567688855647e-05,
|
|
"loss": 0.3496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21243366599082947,
|
|
"step": 930,
|
|
"valid_targets_mean": 4580.1,
|
|
"valid_targets_min": 3373
|
|
},
|
|
{
|
|
"epoch": 0.4897852278679937,
|
|
"grad_norm": 0.7477837127123838,
|
|
"learning_rate": 2.7943156320119677e-05,
|
|
"loss": 0.3645,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18580077588558197,
|
|
"step": 935,
|
|
"valid_targets_mean": 3786.0,
|
|
"valid_targets_min": 1747
|
|
},
|
|
{
|
|
"epoch": 0.4924044002095338,
|
|
"grad_norm": 0.66746084891723,
|
|
"learning_rate": 2.80927449513837e-05,
|
|
"loss": 0.3639,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1607457548379898,
|
|
"step": 940,
|
|
"valid_targets_mean": 3339.9,
|
|
"valid_targets_min": 1166
|
|
},
|
|
{
|
|
"epoch": 0.4950235725510739,
|
|
"grad_norm": 0.6087369419354419,
|
|
"learning_rate": 2.8242333582647723e-05,
|
|
"loss": 0.3452,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17224672436714172,
|
|
"step": 945,
|
|
"valid_targets_mean": 4114.1,
|
|
"valid_targets_min": 1537
|
|
},
|
|
{
|
|
"epoch": 0.49764274489261395,
|
|
"grad_norm": 0.5999130824203768,
|
|
"learning_rate": 2.8391922213911745e-05,
|
|
"loss": 0.3457,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15660347044467926,
|
|
"step": 950,
|
|
"valid_targets_mean": 4550.4,
|
|
"valid_targets_min": 3119
|
|
},
|
|
{
|
|
"epoch": 0.500261917234154,
|
|
"grad_norm": 0.6722215026458347,
|
|
"learning_rate": 2.8541510845175772e-05,
|
|
"loss": 0.3584,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1680639684200287,
|
|
"step": 955,
|
|
"valid_targets_mean": 3810.8,
|
|
"valid_targets_min": 2562
|
|
},
|
|
{
|
|
"epoch": 0.5028810895756941,
|
|
"grad_norm": 0.666118761172745,
|
|
"learning_rate": 2.8691099476439795e-05,
|
|
"loss": 0.3535,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17258156836032867,
|
|
"step": 960,
|
|
"valid_targets_mean": 4045.5,
|
|
"valid_targets_min": 2659
|
|
},
|
|
{
|
|
"epoch": 0.5055002619172342,
|
|
"grad_norm": 0.7024796510442974,
|
|
"learning_rate": 2.8840688107703817e-05,
|
|
"loss": 0.342,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15005125105381012,
|
|
"step": 965,
|
|
"valid_targets_mean": 4064.8,
|
|
"valid_targets_min": 1349
|
|
},
|
|
{
|
|
"epoch": 0.5081194342587743,
|
|
"grad_norm": 0.7539571461509252,
|
|
"learning_rate": 2.899027673896784e-05,
|
|
"loss": 0.3495,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16101238131523132,
|
|
"step": 970,
|
|
"valid_targets_mean": 3371.4,
|
|
"valid_targets_min": 2474
|
|
},
|
|
{
|
|
"epoch": 0.5107386066003143,
|
|
"grad_norm": 0.7517420716020826,
|
|
"learning_rate": 2.9139865370231867e-05,
|
|
"loss": 0.3601,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14872035384178162,
|
|
"step": 975,
|
|
"valid_targets_mean": 2590.5,
|
|
"valid_targets_min": 738
|
|
},
|
|
{
|
|
"epoch": 0.5133577789418544,
|
|
"grad_norm": 0.6639925547944596,
|
|
"learning_rate": 2.928945400149589e-05,
|
|
"loss": 0.3364,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16390451788902283,
|
|
"step": 980,
|
|
"valid_targets_mean": 3570.5,
|
|
"valid_targets_min": 2033
|
|
},
|
|
{
|
|
"epoch": 0.5159769512833945,
|
|
"grad_norm": 0.7875063369111589,
|
|
"learning_rate": 2.9439042632759912e-05,
|
|
"loss": 0.3639,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1987815797328949,
|
|
"step": 985,
|
|
"valid_targets_mean": 2933.1,
|
|
"valid_targets_min": 1346
|
|
},
|
|
{
|
|
"epoch": 0.5185961236249346,
|
|
"grad_norm": 0.8229821761970155,
|
|
"learning_rate": 2.958863126402394e-05,
|
|
"loss": 0.356,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19160673022270203,
|
|
"step": 990,
|
|
"valid_targets_mean": 2991.1,
|
|
"valid_targets_min": 1821
|
|
},
|
|
{
|
|
"epoch": 0.5212152959664746,
|
|
"grad_norm": 0.6966642598769553,
|
|
"learning_rate": 2.973821989528796e-05,
|
|
"loss": 0.348,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16364285349845886,
|
|
"step": 995,
|
|
"valid_targets_mean": 3400.1,
|
|
"valid_targets_min": 2535
|
|
},
|
|
{
|
|
"epoch": 0.5238344683080147,
|
|
"grad_norm": 0.6969200634499667,
|
|
"learning_rate": 2.9887808526551984e-05,
|
|
"loss": 0.3399,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16623015701770782,
|
|
"step": 1000,
|
|
"valid_targets_mean": 3411.5,
|
|
"valid_targets_min": 2352
|
|
},
|
|
{
|
|
"epoch": 0.5264536406495547,
|
|
"grad_norm": 0.698223030299044,
|
|
"learning_rate": 3.0037397157816007e-05,
|
|
"loss": 0.3464,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15625500679016113,
|
|
"step": 1005,
|
|
"valid_targets_mean": 3319.8,
|
|
"valid_targets_min": 1153
|
|
},
|
|
{
|
|
"epoch": 0.5290728129910948,
|
|
"grad_norm": 0.66010760037979,
|
|
"learning_rate": 3.0186985789080034e-05,
|
|
"loss": 0.3412,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17209413647651672,
|
|
"step": 1010,
|
|
"valid_targets_mean": 3718.0,
|
|
"valid_targets_min": 2072
|
|
},
|
|
{
|
|
"epoch": 0.5316919853326348,
|
|
"grad_norm": 0.6744391263190171,
|
|
"learning_rate": 3.0336574420344056e-05,
|
|
"loss": 0.3529,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16139277815818787,
|
|
"step": 1015,
|
|
"valid_targets_mean": 3451.0,
|
|
"valid_targets_min": 922
|
|
},
|
|
{
|
|
"epoch": 0.5343111576741749,
|
|
"grad_norm": 0.6079834396763925,
|
|
"learning_rate": 3.048616305160808e-05,
|
|
"loss": 0.3482,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13819198310375214,
|
|
"step": 1020,
|
|
"valid_targets_mean": 3381.6,
|
|
"valid_targets_min": 1417
|
|
},
|
|
{
|
|
"epoch": 0.536930330015715,
|
|
"grad_norm": 0.5887998623425829,
|
|
"learning_rate": 3.06357516828721e-05,
|
|
"loss": 0.3442,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15055982768535614,
|
|
"step": 1025,
|
|
"valid_targets_mean": 4260.6,
|
|
"valid_targets_min": 1844
|
|
},
|
|
{
|
|
"epoch": 0.5395495023572551,
|
|
"grad_norm": 0.72423300552497,
|
|
"learning_rate": 3.078534031413613e-05,
|
|
"loss": 0.3443,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17637065052986145,
|
|
"step": 1030,
|
|
"valid_targets_mean": 3525.2,
|
|
"valid_targets_min": 1261
|
|
},
|
|
{
|
|
"epoch": 0.5421686746987951,
|
|
"grad_norm": 0.6296896185231665,
|
|
"learning_rate": 3.093492894540015e-05,
|
|
"loss": 0.3388,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20019851624965668,
|
|
"step": 1035,
|
|
"valid_targets_mean": 4548.6,
|
|
"valid_targets_min": 2077
|
|
},
|
|
{
|
|
"epoch": 0.5447878470403352,
|
|
"grad_norm": 0.6461874598516931,
|
|
"learning_rate": 3.1084517576664174e-05,
|
|
"loss": 0.345,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18294253945350647,
|
|
"step": 1040,
|
|
"valid_targets_mean": 4140.9,
|
|
"valid_targets_min": 2724
|
|
},
|
|
{
|
|
"epoch": 0.5474070193818753,
|
|
"grad_norm": 0.5734534078703716,
|
|
"learning_rate": 3.12341062079282e-05,
|
|
"loss": 0.333,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21917563676834106,
|
|
"step": 1045,
|
|
"valid_targets_mean": 5291.4,
|
|
"valid_targets_min": 2292
|
|
},
|
|
{
|
|
"epoch": 0.5500261917234154,
|
|
"grad_norm": 0.7655146463989795,
|
|
"learning_rate": 3.138369483919222e-05,
|
|
"loss": 0.3546,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17243260145187378,
|
|
"step": 1050,
|
|
"valid_targets_mean": 2965.1,
|
|
"valid_targets_min": 1377
|
|
},
|
|
{
|
|
"epoch": 0.5526453640649555,
|
|
"grad_norm": 0.6429273155264622,
|
|
"learning_rate": 3.1533283470456246e-05,
|
|
"loss": 0.3555,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15151354670524597,
|
|
"step": 1055,
|
|
"valid_targets_mean": 3099.5,
|
|
"valid_targets_min": 1350
|
|
},
|
|
{
|
|
"epoch": 0.5552645364064955,
|
|
"grad_norm": 0.666416181336385,
|
|
"learning_rate": 3.168287210172027e-05,
|
|
"loss": 0.3419,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16854865849018097,
|
|
"step": 1060,
|
|
"valid_targets_mean": 3031.9,
|
|
"valid_targets_min": 654
|
|
},
|
|
{
|
|
"epoch": 0.5578837087480356,
|
|
"grad_norm": 0.6585380604523907,
|
|
"learning_rate": 3.183246073298429e-05,
|
|
"loss": 0.3412,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13452671468257904,
|
|
"step": 1065,
|
|
"valid_targets_mean": 3463.8,
|
|
"valid_targets_min": 1706
|
|
},
|
|
{
|
|
"epoch": 0.5605028810895757,
|
|
"grad_norm": 0.6303742591532483,
|
|
"learning_rate": 3.198204936424832e-05,
|
|
"loss": 0.3465,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18186776340007782,
|
|
"step": 1070,
|
|
"valid_targets_mean": 4105.5,
|
|
"valid_targets_min": 1739
|
|
},
|
|
{
|
|
"epoch": 0.5631220534311158,
|
|
"grad_norm": 0.7100853348756141,
|
|
"learning_rate": 3.2131637995512345e-05,
|
|
"loss": 0.3423,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1468432992696762,
|
|
"step": 1075,
|
|
"valid_targets_mean": 2944.4,
|
|
"valid_targets_min": 1302
|
|
},
|
|
{
|
|
"epoch": 0.5657412257726558,
|
|
"grad_norm": 0.6102827886643082,
|
|
"learning_rate": 3.228122662677637e-05,
|
|
"loss": 0.3196,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1456199586391449,
|
|
"step": 1080,
|
|
"valid_targets_mean": 3531.5,
|
|
"valid_targets_min": 2501
|
|
},
|
|
{
|
|
"epoch": 0.5683603981141959,
|
|
"grad_norm": 0.5957194997566689,
|
|
"learning_rate": 3.243081525804039e-05,
|
|
"loss": 0.3376,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17194604873657227,
|
|
"step": 1085,
|
|
"valid_targets_mean": 4673.8,
|
|
"valid_targets_min": 3232
|
|
},
|
|
{
|
|
"epoch": 0.570979570455736,
|
|
"grad_norm": 0.6126169513308506,
|
|
"learning_rate": 3.258040388930442e-05,
|
|
"loss": 0.3195,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16547763347625732,
|
|
"step": 1090,
|
|
"valid_targets_mean": 4465.1,
|
|
"valid_targets_min": 1101
|
|
},
|
|
{
|
|
"epoch": 0.5735987427972761,
|
|
"grad_norm": 0.6659288984648957,
|
|
"learning_rate": 3.272999252056844e-05,
|
|
"loss": 0.3541,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1823541522026062,
|
|
"step": 1095,
|
|
"valid_targets_mean": 4113.2,
|
|
"valid_targets_min": 2079
|
|
},
|
|
{
|
|
"epoch": 0.5762179151388162,
|
|
"grad_norm": 0.6253409519694156,
|
|
"learning_rate": 3.287958115183246e-05,
|
|
"loss": 0.3324,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15754787623882294,
|
|
"step": 1100,
|
|
"valid_targets_mean": 3891.4,
|
|
"valid_targets_min": 2138
|
|
},
|
|
{
|
|
"epoch": 0.5788370874803562,
|
|
"grad_norm": 0.6051674142458731,
|
|
"learning_rate": 3.302916978309649e-05,
|
|
"loss": 0.3415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16496360301971436,
|
|
"step": 1105,
|
|
"valid_targets_mean": 3470.2,
|
|
"valid_targets_min": 1234
|
|
},
|
|
{
|
|
"epoch": 0.5814562598218963,
|
|
"grad_norm": 0.6431340211396896,
|
|
"learning_rate": 3.3178758414360515e-05,
|
|
"loss": 0.3432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1701410710811615,
|
|
"step": 1110,
|
|
"valid_targets_mean": 4038.5,
|
|
"valid_targets_min": 1822
|
|
},
|
|
{
|
|
"epoch": 0.5840754321634364,
|
|
"grad_norm": 0.5805445731574228,
|
|
"learning_rate": 3.3328347045624535e-05,
|
|
"loss": 0.35,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17148751020431519,
|
|
"step": 1115,
|
|
"valid_targets_mean": 4214.6,
|
|
"valid_targets_min": 1651
|
|
},
|
|
{
|
|
"epoch": 0.5866946045049765,
|
|
"grad_norm": 0.687079516708165,
|
|
"learning_rate": 3.347793567688856e-05,
|
|
"loss": 0.3395,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19782204926013947,
|
|
"step": 1120,
|
|
"valid_targets_mean": 3447.9,
|
|
"valid_targets_min": 1434
|
|
},
|
|
{
|
|
"epoch": 0.5893137768465165,
|
|
"grad_norm": 0.6211556614620224,
|
|
"learning_rate": 3.362752430815259e-05,
|
|
"loss": 0.328,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18377473950386047,
|
|
"step": 1125,
|
|
"valid_targets_mean": 4753.4,
|
|
"valid_targets_min": 2725
|
|
},
|
|
{
|
|
"epoch": 0.5919329491880566,
|
|
"grad_norm": 0.6201693042858446,
|
|
"learning_rate": 3.3777112939416607e-05,
|
|
"loss": 0.3277,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16360099613666534,
|
|
"step": 1130,
|
|
"valid_targets_mean": 4005.8,
|
|
"valid_targets_min": 1966
|
|
},
|
|
{
|
|
"epoch": 0.5945521215295967,
|
|
"grad_norm": 0.6141969340369716,
|
|
"learning_rate": 3.392670157068063e-05,
|
|
"loss": 0.3301,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1609821319580078,
|
|
"step": 1135,
|
|
"valid_targets_mean": 3669.0,
|
|
"valid_targets_min": 1407
|
|
},
|
|
{
|
|
"epoch": 0.5971712938711368,
|
|
"grad_norm": 0.6142388805501698,
|
|
"learning_rate": 3.407629020194465e-05,
|
|
"loss": 0.3288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17364542186260223,
|
|
"step": 1140,
|
|
"valid_targets_mean": 4330.5,
|
|
"valid_targets_min": 1444
|
|
},
|
|
{
|
|
"epoch": 0.5997904662126768,
|
|
"grad_norm": 0.6084379669335195,
|
|
"learning_rate": 3.422587883320868e-05,
|
|
"loss": 0.3146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1472238004207611,
|
|
"step": 1145,
|
|
"valid_targets_mean": 3077.8,
|
|
"valid_targets_min": 1504
|
|
},
|
|
{
|
|
"epoch": 0.6024096385542169,
|
|
"grad_norm": 0.611062889831383,
|
|
"learning_rate": 3.4375467464472705e-05,
|
|
"loss": 0.3208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13173359632492065,
|
|
"step": 1150,
|
|
"valid_targets_mean": 3051.6,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 0.6050288108957569,
|
|
"grad_norm": 0.6400296995620607,
|
|
"learning_rate": 3.4525056095736724e-05,
|
|
"loss": 0.3359,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1425532102584839,
|
|
"step": 1155,
|
|
"valid_targets_mean": 3231.8,
|
|
"valid_targets_min": 1288
|
|
},
|
|
{
|
|
"epoch": 0.607647983237297,
|
|
"grad_norm": 0.5966035226835973,
|
|
"learning_rate": 3.467464472700075e-05,
|
|
"loss": 0.3566,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16310977935791016,
|
|
"step": 1160,
|
|
"valid_targets_mean": 4053.0,
|
|
"valid_targets_min": 1537
|
|
},
|
|
{
|
|
"epoch": 0.610267155578837,
|
|
"grad_norm": 0.6248556637700886,
|
|
"learning_rate": 3.482423335826478e-05,
|
|
"loss": 0.3294,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16761454939842224,
|
|
"step": 1165,
|
|
"valid_targets_mean": 3917.1,
|
|
"valid_targets_min": 2060
|
|
},
|
|
{
|
|
"epoch": 0.6128863279203771,
|
|
"grad_norm": 0.6548057486516451,
|
|
"learning_rate": 3.4973821989528796e-05,
|
|
"loss": 0.3425,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18218758702278137,
|
|
"step": 1170,
|
|
"valid_targets_mean": 4482.6,
|
|
"valid_targets_min": 2515
|
|
},
|
|
{
|
|
"epoch": 0.6155055002619172,
|
|
"grad_norm": 0.6110367045018402,
|
|
"learning_rate": 3.512341062079282e-05,
|
|
"loss": 0.322,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1644124835729599,
|
|
"step": 1175,
|
|
"valid_targets_mean": 3847.2,
|
|
"valid_targets_min": 1744
|
|
},
|
|
{
|
|
"epoch": 0.6181246726034573,
|
|
"grad_norm": 0.5382561625787428,
|
|
"learning_rate": 3.527299925205684e-05,
|
|
"loss": 0.3334,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15760695934295654,
|
|
"step": 1180,
|
|
"valid_targets_mean": 5009.9,
|
|
"valid_targets_min": 2062
|
|
},
|
|
{
|
|
"epoch": 0.6207438449449973,
|
|
"grad_norm": 0.6120159820107818,
|
|
"learning_rate": 3.542258788332087e-05,
|
|
"loss": 0.3292,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15463533997535706,
|
|
"step": 1185,
|
|
"valid_targets_mean": 3550.4,
|
|
"valid_targets_min": 1181
|
|
},
|
|
{
|
|
"epoch": 0.6233630172865374,
|
|
"grad_norm": 0.6205449479706852,
|
|
"learning_rate": 3.5572176514584895e-05,
|
|
"loss": 0.3136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17844107747077942,
|
|
"step": 1190,
|
|
"valid_targets_mean": 4010.6,
|
|
"valid_targets_min": 1937
|
|
},
|
|
{
|
|
"epoch": 0.6259821896280775,
|
|
"grad_norm": 0.6661879722953408,
|
|
"learning_rate": 3.5721765145848914e-05,
|
|
"loss": 0.3422,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20328474044799805,
|
|
"step": 1195,
|
|
"valid_targets_mean": 3624.9,
|
|
"valid_targets_min": 1649
|
|
},
|
|
{
|
|
"epoch": 0.6286013619696176,
|
|
"grad_norm": 0.7204535756744255,
|
|
"learning_rate": 3.587135377711294e-05,
|
|
"loss": 0.3239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15000876784324646,
|
|
"step": 1200,
|
|
"valid_targets_mean": 2808.1,
|
|
"valid_targets_min": 1197
|
|
},
|
|
{
|
|
"epoch": 0.6312205343111577,
|
|
"grad_norm": 0.7212522617141228,
|
|
"learning_rate": 3.602094240837697e-05,
|
|
"loss": 0.339,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17222219705581665,
|
|
"step": 1205,
|
|
"valid_targets_mean": 3251.4,
|
|
"valid_targets_min": 1824
|
|
},
|
|
{
|
|
"epoch": 0.6338397066526977,
|
|
"grad_norm": 0.9133985994273064,
|
|
"learning_rate": 3.6170531039640986e-05,
|
|
"loss": 0.3435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1720942258834839,
|
|
"step": 1210,
|
|
"valid_targets_mean": 3145.6,
|
|
"valid_targets_min": 538
|
|
},
|
|
{
|
|
"epoch": 0.6364588789942378,
|
|
"grad_norm": 0.607433859360118,
|
|
"learning_rate": 3.632011967090502e-05,
|
|
"loss": 0.3311,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14743439853191376,
|
|
"step": 1215,
|
|
"valid_targets_mean": 3597.1,
|
|
"valid_targets_min": 537
|
|
},
|
|
{
|
|
"epoch": 0.6390780513357779,
|
|
"grad_norm": 0.693537651034399,
|
|
"learning_rate": 3.646970830216904e-05,
|
|
"loss": 0.3632,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16639968752861023,
|
|
"step": 1220,
|
|
"valid_targets_mean": 3386.5,
|
|
"valid_targets_min": 2169
|
|
},
|
|
{
|
|
"epoch": 0.641697223677318,
|
|
"grad_norm": 0.6611300941454468,
|
|
"learning_rate": 3.6619296933433065e-05,
|
|
"loss": 0.3325,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17038226127624512,
|
|
"step": 1225,
|
|
"valid_targets_mean": 3687.6,
|
|
"valid_targets_min": 1625
|
|
},
|
|
{
|
|
"epoch": 0.644316396018858,
|
|
"grad_norm": 0.6004055429958011,
|
|
"learning_rate": 3.6768885564697085e-05,
|
|
"loss": 0.338,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14659348130226135,
|
|
"step": 1230,
|
|
"valid_targets_mean": 3807.0,
|
|
"valid_targets_min": 2642
|
|
},
|
|
{
|
|
"epoch": 0.6469355683603981,
|
|
"grad_norm": 0.6119002554696095,
|
|
"learning_rate": 3.691847419596111e-05,
|
|
"loss": 0.3455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18825103342533112,
|
|
"step": 1235,
|
|
"valid_targets_mean": 4289.9,
|
|
"valid_targets_min": 2935
|
|
},
|
|
{
|
|
"epoch": 0.6495547407019382,
|
|
"grad_norm": 0.5296888652650795,
|
|
"learning_rate": 3.706806282722514e-05,
|
|
"loss": 0.34,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17498472332954407,
|
|
"step": 1240,
|
|
"valid_targets_mean": 5189.9,
|
|
"valid_targets_min": 2652
|
|
},
|
|
{
|
|
"epoch": 0.6521739130434783,
|
|
"grad_norm": 0.6212330116518527,
|
|
"learning_rate": 3.7217651458489157e-05,
|
|
"loss": 0.3409,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.221257746219635,
|
|
"step": 1245,
|
|
"valid_targets_mean": 4117.2,
|
|
"valid_targets_min": 648
|
|
},
|
|
{
|
|
"epoch": 0.6547930853850183,
|
|
"grad_norm": 0.6181123400246369,
|
|
"learning_rate": 3.736724008975318e-05,
|
|
"loss": 0.3405,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14589467644691467,
|
|
"step": 1250,
|
|
"valid_targets_mean": 3492.8,
|
|
"valid_targets_min": 637
|
|
},
|
|
{
|
|
"epoch": 0.6574122577265584,
|
|
"grad_norm": 0.6588912659727094,
|
|
"learning_rate": 3.751682872101721e-05,
|
|
"loss": 0.3393,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17551174759864807,
|
|
"step": 1255,
|
|
"valid_targets_mean": 3412.5,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 0.6600314300680985,
|
|
"grad_norm": 0.5476010398843242,
|
|
"learning_rate": 3.766641735228123e-05,
|
|
"loss": 0.3346,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15114666521549225,
|
|
"step": 1260,
|
|
"valid_targets_mean": 4227.4,
|
|
"valid_targets_min": 2182
|
|
},
|
|
{
|
|
"epoch": 0.6626506024096386,
|
|
"grad_norm": 0.6126328425922989,
|
|
"learning_rate": 3.7816005983545255e-05,
|
|
"loss": 0.3249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15365070104599,
|
|
"step": 1265,
|
|
"valid_targets_mean": 3211.1,
|
|
"valid_targets_min": 1372
|
|
},
|
|
{
|
|
"epoch": 0.6652697747511787,
|
|
"grad_norm": 0.6205042228668959,
|
|
"learning_rate": 3.796559461480928e-05,
|
|
"loss": 0.3172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1903889924287796,
|
|
"step": 1270,
|
|
"valid_targets_mean": 4490.9,
|
|
"valid_targets_min": 2594
|
|
},
|
|
{
|
|
"epoch": 0.6678889470927187,
|
|
"grad_norm": 0.5812199952484004,
|
|
"learning_rate": 3.81151832460733e-05,
|
|
"loss": 0.3303,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14865991473197937,
|
|
"step": 1275,
|
|
"valid_targets_mean": 3054.6,
|
|
"valid_targets_min": 530
|
|
},
|
|
{
|
|
"epoch": 0.6705081194342588,
|
|
"grad_norm": 0.6461823490955528,
|
|
"learning_rate": 3.826477187733733e-05,
|
|
"loss": 0.3316,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21536710858345032,
|
|
"step": 1280,
|
|
"valid_targets_mean": 4110.2,
|
|
"valid_targets_min": 611
|
|
},
|
|
{
|
|
"epoch": 0.6731272917757989,
|
|
"grad_norm": 0.659991493743984,
|
|
"learning_rate": 3.8414360508601346e-05,
|
|
"loss": 0.3289,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1761910766363144,
|
|
"step": 1285,
|
|
"valid_targets_mean": 3804.2,
|
|
"valid_targets_min": 1987
|
|
},
|
|
{
|
|
"epoch": 0.675746464117339,
|
|
"grad_norm": 0.7018636647048128,
|
|
"learning_rate": 3.856394913986537e-05,
|
|
"loss": 0.3347,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16965290904045105,
|
|
"step": 1290,
|
|
"valid_targets_mean": 3203.2,
|
|
"valid_targets_min": 1362
|
|
},
|
|
{
|
|
"epoch": 0.678365636458879,
|
|
"grad_norm": 0.5842545058219607,
|
|
"learning_rate": 3.87135377711294e-05,
|
|
"loss": 0.3254,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1561589390039444,
|
|
"step": 1295,
|
|
"valid_targets_mean": 3245.1,
|
|
"valid_targets_min": 852
|
|
},
|
|
{
|
|
"epoch": 0.6809848088004191,
|
|
"grad_norm": 0.7090994749538716,
|
|
"learning_rate": 3.886312640239342e-05,
|
|
"loss": 0.3361,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13681864738464355,
|
|
"step": 1300,
|
|
"valid_targets_mean": 2537.4,
|
|
"valid_targets_min": 1195
|
|
},
|
|
{
|
|
"epoch": 0.6836039811419592,
|
|
"grad_norm": 0.6173257676382597,
|
|
"learning_rate": 3.9012715033657445e-05,
|
|
"loss": 0.3275,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17246338725090027,
|
|
"step": 1305,
|
|
"valid_targets_mean": 4383.0,
|
|
"valid_targets_min": 1397
|
|
},
|
|
{
|
|
"epoch": 0.6862231534834992,
|
|
"grad_norm": 0.565658382473205,
|
|
"learning_rate": 3.916230366492147e-05,
|
|
"loss": 0.3172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12798964977264404,
|
|
"step": 1310,
|
|
"valid_targets_mean": 3421.0,
|
|
"valid_targets_min": 1975
|
|
},
|
|
{
|
|
"epoch": 0.6888423258250392,
|
|
"grad_norm": 0.6247345137814574,
|
|
"learning_rate": 3.931189229618549e-05,
|
|
"loss": 0.3099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16652032732963562,
|
|
"step": 1315,
|
|
"valid_targets_mean": 4100.6,
|
|
"valid_targets_min": 2564
|
|
},
|
|
{
|
|
"epoch": 0.6914614981665793,
|
|
"grad_norm": 0.5083389735343604,
|
|
"learning_rate": 3.946148092744952e-05,
|
|
"loss": 0.3158,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14584901928901672,
|
|
"step": 1320,
|
|
"valid_targets_mean": 4986.9,
|
|
"valid_targets_min": 2595
|
|
},
|
|
{
|
|
"epoch": 0.6940806705081194,
|
|
"grad_norm": 0.5862521496377336,
|
|
"learning_rate": 3.9611069558713536e-05,
|
|
"loss": 0.3258,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18245020508766174,
|
|
"step": 1325,
|
|
"valid_targets_mean": 3992.9,
|
|
"valid_targets_min": 2845
|
|
},
|
|
{
|
|
"epoch": 0.6966998428496595,
|
|
"grad_norm": 0.72016307931705,
|
|
"learning_rate": 3.976065818997756e-05,
|
|
"loss": 0.3132,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1214650347828865,
|
|
"step": 1330,
|
|
"valid_targets_mean": 2390.1,
|
|
"valid_targets_min": 993
|
|
},
|
|
{
|
|
"epoch": 0.6993190151911995,
|
|
"grad_norm": 0.595404792324118,
|
|
"learning_rate": 3.991024682124159e-05,
|
|
"loss": 0.3129,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1291854977607727,
|
|
"step": 1335,
|
|
"valid_targets_mean": 3729.1,
|
|
"valid_targets_min": 1575
|
|
},
|
|
{
|
|
"epoch": 0.7019381875327396,
|
|
"grad_norm": 0.6414455374657542,
|
|
"learning_rate": 3.999999727028487e-05,
|
|
"loss": 0.3248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17417430877685547,
|
|
"step": 1340,
|
|
"valid_targets_mean": 3469.0,
|
|
"valid_targets_min": 1360
|
|
},
|
|
{
|
|
"epoch": 0.7045573598742797,
|
|
"grad_norm": 0.6355175987771899,
|
|
"learning_rate": 3.999996656099815e-05,
|
|
"loss": 0.3136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15191948413848877,
|
|
"step": 1345,
|
|
"valid_targets_mean": 3555.5,
|
|
"valid_targets_min": 2278
|
|
},
|
|
{
|
|
"epoch": 0.7071765322158198,
|
|
"grad_norm": 0.588717387625224,
|
|
"learning_rate": 3.999990173033336e-05,
|
|
"loss": 0.3435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18690279126167297,
|
|
"step": 1350,
|
|
"valid_targets_mean": 3812.2,
|
|
"valid_targets_min": 1454
|
|
},
|
|
{
|
|
"epoch": 0.7097957045573599,
|
|
"grad_norm": 0.6344109053156491,
|
|
"learning_rate": 3.99998027784011e-05,
|
|
"loss": 0.3304,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17694224417209625,
|
|
"step": 1355,
|
|
"valid_targets_mean": 3796.6,
|
|
"valid_targets_min": 2344
|
|
},
|
|
{
|
|
"epoch": 0.7124148768988999,
|
|
"grad_norm": 0.7385041094604863,
|
|
"learning_rate": 3.999966970537021e-05,
|
|
"loss": 0.3294,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1743699014186859,
|
|
"step": 1360,
|
|
"valid_targets_mean": 3324.1,
|
|
"valid_targets_min": 1278
|
|
},
|
|
{
|
|
"epoch": 0.71503404924044,
|
|
"grad_norm": 0.6289043093306019,
|
|
"learning_rate": 3.999950251146769e-05,
|
|
"loss": 0.3133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17869730293750763,
|
|
"step": 1365,
|
|
"valid_targets_mean": 3381.4,
|
|
"valid_targets_min": 1858
|
|
},
|
|
{
|
|
"epoch": 0.7176532215819801,
|
|
"grad_norm": 0.7124677183882361,
|
|
"learning_rate": 3.9999301196978804e-05,
|
|
"loss": 0.3244,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12725886702537537,
|
|
"step": 1370,
|
|
"valid_targets_mean": 2566.0,
|
|
"valid_targets_min": 420
|
|
},
|
|
{
|
|
"epoch": 0.7202723939235202,
|
|
"grad_norm": 0.6978765070091023,
|
|
"learning_rate": 3.9999065762247e-05,
|
|
"loss": 0.3187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1442769169807434,
|
|
"step": 1375,
|
|
"valid_targets_mean": 2649.4,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 0.7228915662650602,
|
|
"grad_norm": 0.5703472717476405,
|
|
"learning_rate": 3.999879620767397e-05,
|
|
"loss": 0.3318,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1461399644613266,
|
|
"step": 1380,
|
|
"valid_targets_mean": 4475.1,
|
|
"valid_targets_min": 1737
|
|
},
|
|
{
|
|
"epoch": 0.7255107386066003,
|
|
"grad_norm": 0.6082942173366175,
|
|
"learning_rate": 3.999849253371956e-05,
|
|
"loss": 0.324,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1938718557357788,
|
|
"step": 1385,
|
|
"valid_targets_mean": 4150.8,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 0.7281299109481404,
|
|
"grad_norm": 0.5986725851757726,
|
|
"learning_rate": 3.999815474090188e-05,
|
|
"loss": 0.322,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1633826494216919,
|
|
"step": 1390,
|
|
"valid_targets_mean": 3864.1,
|
|
"valid_targets_min": 1780
|
|
},
|
|
{
|
|
"epoch": 0.7307490832896805,
|
|
"grad_norm": 0.5941021429982363,
|
|
"learning_rate": 3.9997782829797225e-05,
|
|
"loss": 0.33,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1542864739894867,
|
|
"step": 1395,
|
|
"valid_targets_mean": 4058.0,
|
|
"valid_targets_min": 2185
|
|
},
|
|
{
|
|
"epoch": 0.7333682556312205,
|
|
"grad_norm": 0.7314816153433888,
|
|
"learning_rate": 3.99973768010401e-05,
|
|
"loss": 0.2847,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09025775641202927,
|
|
"step": 1400,
|
|
"valid_targets_mean": 2736.0,
|
|
"valid_targets_min": 876
|
|
},
|
|
{
|
|
"epoch": 0.7359874279727606,
|
|
"grad_norm": 0.7986663261684929,
|
|
"learning_rate": 3.9996936655323216e-05,
|
|
"loss": 0.1956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0739547535777092,
|
|
"step": 1405,
|
|
"valid_targets_mean": 1472.5,
|
|
"valid_targets_min": 324
|
|
},
|
|
{
|
|
"epoch": 0.7386066003143007,
|
|
"grad_norm": 0.5158462550711758,
|
|
"learning_rate": 3.9996462393397505e-05,
|
|
"loss": 0.1507,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.086447574198246,
|
|
"step": 1410,
|
|
"valid_targets_mean": 2699.9,
|
|
"valid_targets_min": 1064
|
|
},
|
|
{
|
|
"epoch": 0.7412257726558408,
|
|
"grad_norm": 0.5424757814397492,
|
|
"learning_rate": 3.999595401607208e-05,
|
|
"loss": 0.1439,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07466402649879456,
|
|
"step": 1415,
|
|
"valid_targets_mean": 3137.0,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 0.7438449449973809,
|
|
"grad_norm": 0.5353903699192518,
|
|
"learning_rate": 3.9995411524214275e-05,
|
|
"loss": 0.161,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05025868117809296,
|
|
"step": 1420,
|
|
"valid_targets_mean": 3472.9,
|
|
"valid_targets_min": 1223
|
|
},
|
|
{
|
|
"epoch": 0.7464641173389209,
|
|
"grad_norm": 0.5130922985032678,
|
|
"learning_rate": 3.999483491874962e-05,
|
|
"loss": 0.1463,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08507537841796875,
|
|
"step": 1425,
|
|
"valid_targets_mean": 3326.0,
|
|
"valid_targets_min": 1431
|
|
},
|
|
{
|
|
"epoch": 0.749083289680461,
|
|
"grad_norm": 0.8635294205181738,
|
|
"learning_rate": 3.999422420066184e-05,
|
|
"loss": 0.1984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08754102885723114,
|
|
"step": 1430,
|
|
"valid_targets_mean": 1161.2,
|
|
"valid_targets_min": 559
|
|
},
|
|
{
|
|
"epoch": 0.7517024620220011,
|
|
"grad_norm": 0.46394205723150345,
|
|
"learning_rate": 3.9993579370992874e-05,
|
|
"loss": 0.1862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06998488306999207,
|
|
"step": 1435,
|
|
"valid_targets_mean": 3598.2,
|
|
"valid_targets_min": 1623
|
|
},
|
|
{
|
|
"epoch": 0.7543216343635412,
|
|
"grad_norm": 0.49592682911702274,
|
|
"learning_rate": 3.9992900430842836e-05,
|
|
"loss": 0.1576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0781191736459732,
|
|
"step": 1440,
|
|
"valid_targets_mean": 3178.2,
|
|
"valid_targets_min": 846
|
|
},
|
|
{
|
|
"epoch": 0.7569408067050812,
|
|
"grad_norm": 0.6450361626216348,
|
|
"learning_rate": 3.999218738137005e-05,
|
|
"loss": 0.2034,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13390403985977173,
|
|
"step": 1445,
|
|
"valid_targets_mean": 3622.6,
|
|
"valid_targets_min": 1949
|
|
},
|
|
{
|
|
"epoch": 0.7595599790466213,
|
|
"grad_norm": 0.41886270656201374,
|
|
"learning_rate": 3.999144022379103e-05,
|
|
"loss": 0.1169,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06320428848266602,
|
|
"step": 1450,
|
|
"valid_targets_mean": 3344.9,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 0.7621791513881614,
|
|
"grad_norm": 0.5860692709334011,
|
|
"learning_rate": 3.9990658959380485e-05,
|
|
"loss": 0.1393,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0750679075717926,
|
|
"step": 1455,
|
|
"valid_targets_mean": 2391.6,
|
|
"valid_targets_min": 776
|
|
},
|
|
{
|
|
"epoch": 0.7647983237297015,
|
|
"grad_norm": 0.5010127595551124,
|
|
"learning_rate": 3.998984358947131e-05,
|
|
"loss": 0.1386,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07392105460166931,
|
|
"step": 1460,
|
|
"valid_targets_mean": 2818.5,
|
|
"valid_targets_min": 754
|
|
},
|
|
{
|
|
"epoch": 0.7674174960712414,
|
|
"grad_norm": 0.5866841846039216,
|
|
"learning_rate": 3.998899411545457e-05,
|
|
"loss": 0.1669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08332496881484985,
|
|
"step": 1465,
|
|
"valid_targets_mean": 1646.6,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 0.7700366684127815,
|
|
"grad_norm": 0.4642834513510014,
|
|
"learning_rate": 3.998811053877955e-05,
|
|
"loss": 0.1536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07143302261829376,
|
|
"step": 1470,
|
|
"valid_targets_mean": 3374.8,
|
|
"valid_targets_min": 1862
|
|
},
|
|
{
|
|
"epoch": 0.7726558407543216,
|
|
"grad_norm": 0.515280742585236,
|
|
"learning_rate": 3.998719286095368e-05,
|
|
"loss": 0.1404,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08594658970832825,
|
|
"step": 1475,
|
|
"valid_targets_mean": 2837.4,
|
|
"valid_targets_min": 574
|
|
},
|
|
{
|
|
"epoch": 0.7752750130958617,
|
|
"grad_norm": 0.5213394719374975,
|
|
"learning_rate": 3.998624108354258e-05,
|
|
"loss": 0.1332,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06072106584906578,
|
|
"step": 1480,
|
|
"valid_targets_mean": 2618.4,
|
|
"valid_targets_min": 523
|
|
},
|
|
{
|
|
"epoch": 0.7778941854374017,
|
|
"grad_norm": 0.5070710235211495,
|
|
"learning_rate": 3.9985255208170065e-05,
|
|
"loss": 0.1322,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07832945883274078,
|
|
"step": 1485,
|
|
"valid_targets_mean": 3548.9,
|
|
"valid_targets_min": 2189
|
|
},
|
|
{
|
|
"epoch": 0.7805133577789418,
|
|
"grad_norm": 0.46229267557092607,
|
|
"learning_rate": 3.99842352365181e-05,
|
|
"loss": 0.1443,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05179700627923012,
|
|
"step": 1490,
|
|
"valid_targets_mean": 3564.6,
|
|
"valid_targets_min": 827
|
|
},
|
|
{
|
|
"epoch": 0.7831325301204819,
|
|
"grad_norm": 0.5527045613471703,
|
|
"learning_rate": 3.9983181170326835e-05,
|
|
"loss": 0.1586,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07232558727264404,
|
|
"step": 1495,
|
|
"valid_targets_mean": 1936.6,
|
|
"valid_targets_min": 597
|
|
},
|
|
{
|
|
"epoch": 0.785751702462022,
|
|
"grad_norm": 0.5620881757133277,
|
|
"learning_rate": 3.998209301139458e-05,
|
|
"loss": 0.1536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06416729837656021,
|
|
"step": 1500,
|
|
"valid_targets_mean": 1747.5,
|
|
"valid_targets_min": 701
|
|
},
|
|
{
|
|
"epoch": 0.788370874803562,
|
|
"grad_norm": 0.605616319073172,
|
|
"learning_rate": 3.998097076157781e-05,
|
|
"loss": 0.1786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1605839878320694,
|
|
"step": 1505,
|
|
"valid_targets_mean": 3329.6,
|
|
"valid_targets_min": 1652
|
|
},
|
|
{
|
|
"epoch": 0.7909900471451021,
|
|
"grad_norm": 0.46922548355435045,
|
|
"learning_rate": 3.997981442279117e-05,
|
|
"loss": 0.1549,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06859530508518219,
|
|
"step": 1510,
|
|
"valid_targets_mean": 3117.1,
|
|
"valid_targets_min": 786
|
|
},
|
|
{
|
|
"epoch": 0.7936092194866422,
|
|
"grad_norm": 0.7021593930917235,
|
|
"learning_rate": 3.997862399700744e-05,
|
|
"loss": 0.1323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08121612668037415,
|
|
"step": 1515,
|
|
"valid_targets_mean": 2574.4,
|
|
"valid_targets_min": 971
|
|
},
|
|
{
|
|
"epoch": 0.7962283918281823,
|
|
"grad_norm": 0.4561349661914808,
|
|
"learning_rate": 3.99773994862576e-05,
|
|
"loss": 0.1846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0743626058101654,
|
|
"step": 1520,
|
|
"valid_targets_mean": 3320.0,
|
|
"valid_targets_min": 2610
|
|
},
|
|
{
|
|
"epoch": 0.7988475641697224,
|
|
"grad_norm": 0.5086685356121075,
|
|
"learning_rate": 3.997614089263073e-05,
|
|
"loss": 0.1582,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051420897245407104,
|
|
"step": 1525,
|
|
"valid_targets_mean": 2580.6,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 0.8014667365112624,
|
|
"grad_norm": 0.6117088309504906,
|
|
"learning_rate": 3.997484821827409e-05,
|
|
"loss": 0.1444,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06646887958049774,
|
|
"step": 1530,
|
|
"valid_targets_mean": 1537.1,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 0.8040859088528025,
|
|
"grad_norm": 0.5949023934852837,
|
|
"learning_rate": 3.9973521465393075e-05,
|
|
"loss": 0.3669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15084227919578552,
|
|
"step": 1535,
|
|
"valid_targets_mean": 2271.2,
|
|
"valid_targets_min": 987
|
|
},
|
|
{
|
|
"epoch": 0.8067050811943426,
|
|
"grad_norm": 0.7502710001993674,
|
|
"learning_rate": 3.997216063625121e-05,
|
|
"loss": 0.1631,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07774441689252853,
|
|
"step": 1540,
|
|
"valid_targets_mean": 1351.5,
|
|
"valid_targets_min": 329
|
|
},
|
|
{
|
|
"epoch": 0.8093242535358827,
|
|
"grad_norm": 0.606702731963554,
|
|
"learning_rate": 3.997076573317019e-05,
|
|
"loss": 0.1445,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06768237799406052,
|
|
"step": 1545,
|
|
"valid_targets_mean": 1553.9,
|
|
"valid_targets_min": 608
|
|
},
|
|
{
|
|
"epoch": 0.8119434258774227,
|
|
"grad_norm": 0.5191457016879443,
|
|
"learning_rate": 3.9969336758529804e-05,
|
|
"loss": 0.2531,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0957370176911354,
|
|
"step": 1550,
|
|
"valid_targets_mean": 3645.6,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 0.8145625982189628,
|
|
"grad_norm": 0.3155287206576497,
|
|
"learning_rate": 3.996787371476799e-05,
|
|
"loss": 0.1225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034252800047397614,
|
|
"step": 1555,
|
|
"valid_targets_mean": 4720.6,
|
|
"valid_targets_min": 3529
|
|
},
|
|
{
|
|
"epoch": 0.8171817705605029,
|
|
"grad_norm": 0.4212640247312681,
|
|
"learning_rate": 3.9966376604380795e-05,
|
|
"loss": 0.1411,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07607945054769516,
|
|
"step": 1560,
|
|
"valid_targets_mean": 4437.6,
|
|
"valid_targets_min": 2336
|
|
},
|
|
{
|
|
"epoch": 0.819800942902043,
|
|
"grad_norm": 0.39620114331915346,
|
|
"learning_rate": 3.9964845429922405e-05,
|
|
"loss": 0.1372,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06097068265080452,
|
|
"step": 1565,
|
|
"valid_targets_mean": 3533.1,
|
|
"valid_targets_min": 1226
|
|
},
|
|
{
|
|
"epoch": 0.822420115243583,
|
|
"grad_norm": 0.41995034506751305,
|
|
"learning_rate": 3.996328019400512e-05,
|
|
"loss": 0.1334,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06280517578125,
|
|
"step": 1570,
|
|
"valid_targets_mean": 3299.2,
|
|
"valid_targets_min": 798
|
|
},
|
|
{
|
|
"epoch": 0.8250392875851231,
|
|
"grad_norm": 0.46326861768093075,
|
|
"learning_rate": 3.996168089929934e-05,
|
|
"loss": 0.1262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052490413188934326,
|
|
"step": 1575,
|
|
"valid_targets_mean": 1836.9,
|
|
"valid_targets_min": 723
|
|
},
|
|
{
|
|
"epoch": 0.8276584599266632,
|
|
"grad_norm": 0.5055429997269286,
|
|
"learning_rate": 3.996004754853356e-05,
|
|
"loss": 0.1288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.063631571829319,
|
|
"step": 1580,
|
|
"valid_targets_mean": 3316.0,
|
|
"valid_targets_min": 898
|
|
},
|
|
{
|
|
"epoch": 0.8302776322682033,
|
|
"grad_norm": 0.3514177365249329,
|
|
"learning_rate": 3.995838014449442e-05,
|
|
"loss": 0.1212,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047119513154029846,
|
|
"step": 1585,
|
|
"valid_targets_mean": 4027.2,
|
|
"valid_targets_min": 3278
|
|
},
|
|
{
|
|
"epoch": 0.8328968046097434,
|
|
"grad_norm": 0.5531182262228151,
|
|
"learning_rate": 3.9956678690026625e-05,
|
|
"loss": 0.132,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061042577028274536,
|
|
"step": 1590,
|
|
"valid_targets_mean": 2052.8,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 0.8355159769512834,
|
|
"grad_norm": 0.7755795660820791,
|
|
"learning_rate": 3.995494318803297e-05,
|
|
"loss": 0.174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07252795249223709,
|
|
"step": 1595,
|
|
"valid_targets_mean": 1574.9,
|
|
"valid_targets_min": 538
|
|
},
|
|
{
|
|
"epoch": 0.8381351492928235,
|
|
"grad_norm": 0.625425729205124,
|
|
"learning_rate": 3.995317364147434e-05,
|
|
"loss": 0.1096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050821200013160706,
|
|
"step": 1600,
|
|
"valid_targets_mean": 1956.9,
|
|
"valid_targets_min": 654
|
|
},
|
|
{
|
|
"epoch": 0.8407543216343636,
|
|
"grad_norm": 0.5634458011817362,
|
|
"learning_rate": 3.9951370053369726e-05,
|
|
"loss": 0.1296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07513399422168732,
|
|
"step": 1605,
|
|
"valid_targets_mean": 2735.4,
|
|
"valid_targets_min": 930
|
|
},
|
|
{
|
|
"epoch": 0.8433734939759037,
|
|
"grad_norm": 0.48760076249445794,
|
|
"learning_rate": 3.994953242679617e-05,
|
|
"loss": 0.1162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06778737157583237,
|
|
"step": 1610,
|
|
"valid_targets_mean": 2965.0,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 0.8459926663174437,
|
|
"grad_norm": 0.4572772548812646,
|
|
"learning_rate": 3.994766076488879e-05,
|
|
"loss": 0.1866,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06816256791353226,
|
|
"step": 1615,
|
|
"valid_targets_mean": 3344.6,
|
|
"valid_targets_min": 999
|
|
},
|
|
{
|
|
"epoch": 0.8486118386589837,
|
|
"grad_norm": 0.65505384257651,
|
|
"learning_rate": 3.994575507084078e-05,
|
|
"loss": 0.1418,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07841216027736664,
|
|
"step": 1620,
|
|
"valid_targets_mean": 2445.4,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 0.8512310110005238,
|
|
"grad_norm": 0.4517191751461857,
|
|
"learning_rate": 3.994381534790339e-05,
|
|
"loss": 0.159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07189537584781647,
|
|
"step": 1625,
|
|
"valid_targets_mean": 3514.9,
|
|
"valid_targets_min": 1068
|
|
},
|
|
{
|
|
"epoch": 0.8538501833420639,
|
|
"grad_norm": 0.5158046190436185,
|
|
"learning_rate": 3.994184159938593e-05,
|
|
"loss": 0.1412,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060847245156764984,
|
|
"step": 1630,
|
|
"valid_targets_mean": 1602.8,
|
|
"valid_targets_min": 536
|
|
},
|
|
{
|
|
"epoch": 0.8564693556836039,
|
|
"grad_norm": 0.5277377242693408,
|
|
"learning_rate": 3.993983382865576e-05,
|
|
"loss": 0.1975,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14055420458316803,
|
|
"step": 1635,
|
|
"valid_targets_mean": 3568.8,
|
|
"valid_targets_min": 1507
|
|
},
|
|
{
|
|
"epoch": 0.859088528025144,
|
|
"grad_norm": 0.5069279427908884,
|
|
"learning_rate": 3.993779203913827e-05,
|
|
"loss": 0.1428,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06706821918487549,
|
|
"step": 1640,
|
|
"valid_targets_mean": 3292.2,
|
|
"valid_targets_min": 2342
|
|
},
|
|
{
|
|
"epoch": 0.8617077003666841,
|
|
"grad_norm": 0.6963105091913101,
|
|
"learning_rate": 3.99357162343169e-05,
|
|
"loss": 0.1733,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08217508345842361,
|
|
"step": 1645,
|
|
"valid_targets_mean": 1567.5,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 0.8643268727082242,
|
|
"grad_norm": 0.626309530814196,
|
|
"learning_rate": 3.993360641773313e-05,
|
|
"loss": 0.1332,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0423566959798336,
|
|
"step": 1650,
|
|
"valid_targets_mean": 1145.4,
|
|
"valid_targets_min": 449
|
|
},
|
|
{
|
|
"epoch": 0.8669460450497642,
|
|
"grad_norm": 0.42519785233247775,
|
|
"learning_rate": 3.993146259298646e-05,
|
|
"loss": 0.1385,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05637018382549286,
|
|
"step": 1655,
|
|
"valid_targets_mean": 3191.0,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 0.8695652173913043,
|
|
"grad_norm": 0.7088187288825465,
|
|
"learning_rate": 3.99292847637344e-05,
|
|
"loss": 0.1386,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08269469439983368,
|
|
"step": 1660,
|
|
"valid_targets_mean": 1924.8,
|
|
"valid_targets_min": 577
|
|
},
|
|
{
|
|
"epoch": 0.8721843897328444,
|
|
"grad_norm": 0.3732596807976658,
|
|
"learning_rate": 3.9927072933692484e-05,
|
|
"loss": 0.1135,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05370637774467468,
|
|
"step": 1665,
|
|
"valid_targets_mean": 3992.5,
|
|
"valid_targets_min": 726
|
|
},
|
|
{
|
|
"epoch": 0.8748035620743845,
|
|
"grad_norm": 0.4134721656759607,
|
|
"learning_rate": 3.992482710663426e-05,
|
|
"loss": 0.1138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06572000682353973,
|
|
"step": 1670,
|
|
"valid_targets_mean": 3626.4,
|
|
"valid_targets_min": 2252
|
|
},
|
|
{
|
|
"epoch": 0.8774227344159246,
|
|
"grad_norm": 0.5202339996629215,
|
|
"learning_rate": 3.992254728639127e-05,
|
|
"loss": 0.1246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05162326991558075,
|
|
"step": 1675,
|
|
"valid_targets_mean": 1556.2,
|
|
"valid_targets_min": 531
|
|
},
|
|
{
|
|
"epoch": 0.8800419067574646,
|
|
"grad_norm": 0.4644389455013823,
|
|
"learning_rate": 3.9920233476853035e-05,
|
|
"loss": 0.1503,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06630605459213257,
|
|
"step": 1680,
|
|
"valid_targets_mean": 2410.5,
|
|
"valid_targets_min": 744
|
|
},
|
|
{
|
|
"epoch": 0.8826610790990047,
|
|
"grad_norm": 0.45822935132382786,
|
|
"learning_rate": 3.99178856819671e-05,
|
|
"loss": 0.1395,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07550811767578125,
|
|
"step": 1685,
|
|
"valid_targets_mean": 3795.1,
|
|
"valid_targets_min": 3136
|
|
},
|
|
{
|
|
"epoch": 0.8852802514405448,
|
|
"grad_norm": 0.37909483938475036,
|
|
"learning_rate": 3.991550390573897e-05,
|
|
"loss": 0.1227,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052796460688114166,
|
|
"step": 1690,
|
|
"valid_targets_mean": 3324.4,
|
|
"valid_targets_min": 2366
|
|
},
|
|
{
|
|
"epoch": 0.8878994237820849,
|
|
"grad_norm": 0.4170391242135034,
|
|
"learning_rate": 3.9913088152232105e-05,
|
|
"loss": 0.1212,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04679000750184059,
|
|
"step": 1695,
|
|
"valid_targets_mean": 1668.1,
|
|
"valid_targets_min": 669
|
|
},
|
|
{
|
|
"epoch": 0.8905185961236249,
|
|
"grad_norm": 0.47835404906149454,
|
|
"learning_rate": 3.9910638425567975e-05,
|
|
"loss": 0.149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07802337408065796,
|
|
"step": 1700,
|
|
"valid_targets_mean": 3529.8,
|
|
"valid_targets_min": 1785
|
|
},
|
|
{
|
|
"epoch": 0.893137768465165,
|
|
"grad_norm": 0.31806325971087684,
|
|
"learning_rate": 3.990815472992598e-05,
|
|
"loss": 0.109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05754898488521576,
|
|
"step": 1705,
|
|
"valid_targets_mean": 4120.1,
|
|
"valid_targets_min": 3862
|
|
},
|
|
{
|
|
"epoch": 0.8957569408067051,
|
|
"grad_norm": 0.4547597012813676,
|
|
"learning_rate": 3.9905637069543486e-05,
|
|
"loss": 0.1318,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059976283460855484,
|
|
"step": 1710,
|
|
"valid_targets_mean": 3062.2,
|
|
"valid_targets_min": 539
|
|
},
|
|
{
|
|
"epoch": 0.8983761131482452,
|
|
"grad_norm": 0.4378450620162749,
|
|
"learning_rate": 3.9903085448715805e-05,
|
|
"loss": 0.1389,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05159597843885422,
|
|
"step": 1715,
|
|
"valid_targets_mean": 2781.5,
|
|
"valid_targets_min": 1016
|
|
},
|
|
{
|
|
"epoch": 0.9009952854897852,
|
|
"grad_norm": 0.7506102404125042,
|
|
"learning_rate": 3.990049987179618e-05,
|
|
"loss": 0.1495,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08937472850084305,
|
|
"step": 1720,
|
|
"valid_targets_mean": 1102.9,
|
|
"valid_targets_min": 888
|
|
},
|
|
{
|
|
"epoch": 0.9036144578313253,
|
|
"grad_norm": 0.5309387154494348,
|
|
"learning_rate": 3.98978803431958e-05,
|
|
"loss": 0.1366,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05678325146436691,
|
|
"step": 1725,
|
|
"valid_targets_mean": 2602.5,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 0.9062336301728654,
|
|
"grad_norm": 0.6742771489158552,
|
|
"learning_rate": 3.989522686738375e-05,
|
|
"loss": 0.1378,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08440111577510834,
|
|
"step": 1730,
|
|
"valid_targets_mean": 1592.9,
|
|
"valid_targets_min": 605
|
|
},
|
|
{
|
|
"epoch": 0.9088528025144055,
|
|
"grad_norm": 0.4610770926209317,
|
|
"learning_rate": 3.989253944888707e-05,
|
|
"loss": 0.1341,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05471458286046982,
|
|
"step": 1735,
|
|
"valid_targets_mean": 2899.1,
|
|
"valid_targets_min": 1117
|
|
},
|
|
{
|
|
"epoch": 0.9114719748559456,
|
|
"grad_norm": 0.4341204746225806,
|
|
"learning_rate": 3.988981809229067e-05,
|
|
"loss": 0.1598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059293102473020554,
|
|
"step": 1740,
|
|
"valid_targets_mean": 3376.6,
|
|
"valid_targets_min": 1388
|
|
},
|
|
{
|
|
"epoch": 0.9140911471974856,
|
|
"grad_norm": 0.407443066879257,
|
|
"learning_rate": 3.98870628022374e-05,
|
|
"loss": 0.1287,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05706991255283356,
|
|
"step": 1745,
|
|
"valid_targets_mean": 3243.4,
|
|
"valid_targets_min": 524
|
|
},
|
|
{
|
|
"epoch": 0.9167103195390257,
|
|
"grad_norm": 0.4697544449354311,
|
|
"learning_rate": 3.988427358342797e-05,
|
|
"loss": 0.1175,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05848868936300278,
|
|
"step": 1750,
|
|
"valid_targets_mean": 2468.0,
|
|
"valid_targets_min": 970
|
|
},
|
|
{
|
|
"epoch": 0.9193294918805658,
|
|
"grad_norm": 0.48937162451685806,
|
|
"learning_rate": 3.9881450440620994e-05,
|
|
"loss": 0.1428,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06698908656835556,
|
|
"step": 1755,
|
|
"valid_targets_mean": 2959.6,
|
|
"valid_targets_min": 877
|
|
},
|
|
{
|
|
"epoch": 0.9219486642221059,
|
|
"grad_norm": 0.35050547101950813,
|
|
"learning_rate": 3.987859337863295e-05,
|
|
"loss": 0.1385,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05743839591741562,
|
|
"step": 1760,
|
|
"valid_targets_mean": 3886.5,
|
|
"valid_targets_min": 3251
|
|
},
|
|
{
|
|
"epoch": 0.9245678365636459,
|
|
"grad_norm": 0.5895188068394133,
|
|
"learning_rate": 3.98757024023382e-05,
|
|
"loss": 0.1152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06867806613445282,
|
|
"step": 1765,
|
|
"valid_targets_mean": 2695.6,
|
|
"valid_targets_min": 564
|
|
},
|
|
{
|
|
"epoch": 0.927187008905186,
|
|
"grad_norm": 0.3473075470152696,
|
|
"learning_rate": 3.9872777516668935e-05,
|
|
"loss": 0.1228,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06826956570148468,
|
|
"step": 1770,
|
|
"valid_targets_mean": 4404.8,
|
|
"valid_targets_min": 2555
|
|
},
|
|
{
|
|
"epoch": 0.929806181246726,
|
|
"grad_norm": 0.3421684036177156,
|
|
"learning_rate": 3.9869818726615245e-05,
|
|
"loss": 0.1133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05443980544805527,
|
|
"step": 1775,
|
|
"valid_targets_mean": 4410.5,
|
|
"valid_targets_min": 3261
|
|
},
|
|
{
|
|
"epoch": 0.9324253535882661,
|
|
"grad_norm": 0.4417010057915673,
|
|
"learning_rate": 3.9866826037225025e-05,
|
|
"loss": 0.1046,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04898755997419357,
|
|
"step": 1780,
|
|
"valid_targets_mean": 1768.5,
|
|
"valid_targets_min": 618
|
|
},
|
|
{
|
|
"epoch": 0.9350445259298061,
|
|
"grad_norm": 0.3543811897370335,
|
|
"learning_rate": 3.9863799453604026e-05,
|
|
"loss": 0.1331,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05469723045825958,
|
|
"step": 1785,
|
|
"valid_targets_mean": 3985.8,
|
|
"valid_targets_min": 3213
|
|
},
|
|
{
|
|
"epoch": 0.9376636982713462,
|
|
"grad_norm": 0.4264038979036985,
|
|
"learning_rate": 3.986073898091581e-05,
|
|
"loss": 0.1289,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.118505097925663,
|
|
"step": 1790,
|
|
"valid_targets_mean": 3149.1,
|
|
"valid_targets_min": 1539
|
|
},
|
|
{
|
|
"epoch": 0.9402828706128863,
|
|
"grad_norm": 0.3789345301334967,
|
|
"learning_rate": 3.985764462438176e-05,
|
|
"loss": 0.1392,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04531242698431015,
|
|
"step": 1795,
|
|
"valid_targets_mean": 2094.8,
|
|
"valid_targets_min": 509
|
|
},
|
|
{
|
|
"epoch": 0.9429020429544264,
|
|
"grad_norm": 0.34121951059745015,
|
|
"learning_rate": 3.985451638928108e-05,
|
|
"loss": 0.1291,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0372382327914238,
|
|
"step": 1800,
|
|
"valid_targets_mean": 2441.4,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 0.9455212152959664,
|
|
"grad_norm": 0.42942031033018574,
|
|
"learning_rate": 3.9851354280950756e-05,
|
|
"loss": 0.1558,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06137951835989952,
|
|
"step": 1805,
|
|
"valid_targets_mean": 2740.9,
|
|
"valid_targets_min": 737
|
|
},
|
|
{
|
|
"epoch": 0.9481403876375065,
|
|
"grad_norm": 0.36405462270761896,
|
|
"learning_rate": 3.9848158304785576e-05,
|
|
"loss": 0.1141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056062668561935425,
|
|
"step": 1810,
|
|
"valid_targets_mean": 2726.9,
|
|
"valid_targets_min": 492
|
|
},
|
|
{
|
|
"epoch": 0.9507595599790466,
|
|
"grad_norm": 0.6623352936503045,
|
|
"learning_rate": 3.984492846623811e-05,
|
|
"loss": 0.1427,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08910174667835236,
|
|
"step": 1815,
|
|
"valid_targets_mean": 1607.9,
|
|
"valid_targets_min": 833
|
|
},
|
|
{
|
|
"epoch": 0.9533787323205867,
|
|
"grad_norm": 0.35061832091732376,
|
|
"learning_rate": 3.984166477081868e-05,
|
|
"loss": 0.1194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03973294422030449,
|
|
"step": 1820,
|
|
"valid_targets_mean": 3502.6,
|
|
"valid_targets_min": 923
|
|
},
|
|
{
|
|
"epoch": 0.9559979046621268,
|
|
"grad_norm": 0.6962754339523737,
|
|
"learning_rate": 3.983836722409539e-05,
|
|
"loss": 0.1792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08447998017072678,
|
|
"step": 1825,
|
|
"valid_targets_mean": 1553.9,
|
|
"valid_targets_min": 531
|
|
},
|
|
{
|
|
"epoch": 0.9586170770036668,
|
|
"grad_norm": 0.5601831525746735,
|
|
"learning_rate": 3.98350358316941e-05,
|
|
"loss": 0.1617,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10630452632904053,
|
|
"step": 1830,
|
|
"valid_targets_mean": 1825.6,
|
|
"valid_targets_min": 575
|
|
},
|
|
{
|
|
"epoch": 0.9612362493452069,
|
|
"grad_norm": 0.5004613077270998,
|
|
"learning_rate": 3.9831670599298394e-05,
|
|
"loss": 0.1264,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06777000427246094,
|
|
"step": 1835,
|
|
"valid_targets_mean": 3701.9,
|
|
"valid_targets_min": 2784
|
|
},
|
|
{
|
|
"epoch": 0.963855421686747,
|
|
"grad_norm": 0.38095623986806926,
|
|
"learning_rate": 3.9828271532649595e-05,
|
|
"loss": 0.1248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05625955015420914,
|
|
"step": 1840,
|
|
"valid_targets_mean": 3575.8,
|
|
"valid_targets_min": 2525
|
|
},
|
|
{
|
|
"epoch": 0.9664745940282871,
|
|
"grad_norm": 0.3645503223771157,
|
|
"learning_rate": 3.9824838637546774e-05,
|
|
"loss": 0.122,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04756051301956177,
|
|
"step": 1845,
|
|
"valid_targets_mean": 3081.0,
|
|
"valid_targets_min": 970
|
|
},
|
|
{
|
|
"epoch": 0.9690937663698271,
|
|
"grad_norm": 0.49645775656429314,
|
|
"learning_rate": 3.982137191984668e-05,
|
|
"loss": 0.1184,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06135290116071701,
|
|
"step": 1850,
|
|
"valid_targets_mean": 1963.5,
|
|
"valid_targets_min": 817
|
|
},
|
|
{
|
|
"epoch": 0.9717129387113672,
|
|
"grad_norm": 0.7974570176042606,
|
|
"learning_rate": 3.98178713854638e-05,
|
|
"loss": 0.1544,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10152144730091095,
|
|
"step": 1855,
|
|
"valid_targets_mean": 1270.9,
|
|
"valid_targets_min": 637
|
|
},
|
|
{
|
|
"epoch": 0.9743321110529073,
|
|
"grad_norm": 0.4046465137561532,
|
|
"learning_rate": 3.981433704037027e-05,
|
|
"loss": 0.1109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05127967149019241,
|
|
"step": 1860,
|
|
"valid_targets_mean": 2594.1,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 0.9769512833944474,
|
|
"grad_norm": 0.3493509574031833,
|
|
"learning_rate": 3.981076889059596e-05,
|
|
"loss": 0.116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05416160076856613,
|
|
"step": 1865,
|
|
"valid_targets_mean": 3222.2,
|
|
"valid_targets_min": 1606
|
|
},
|
|
{
|
|
"epoch": 0.9795704557359874,
|
|
"grad_norm": 0.4105765185711659,
|
|
"learning_rate": 3.980716694222838e-05,
|
|
"loss": 0.1234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06041751429438591,
|
|
"step": 1870,
|
|
"valid_targets_mean": 3023.8,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 0.9821896280775275,
|
|
"grad_norm": 0.3137716102916606,
|
|
"learning_rate": 3.9803531201412716e-05,
|
|
"loss": 0.1285,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0552176833152771,
|
|
"step": 1875,
|
|
"valid_targets_mean": 4263.8,
|
|
"valid_targets_min": 1078
|
|
},
|
|
{
|
|
"epoch": 0.9848088004190676,
|
|
"grad_norm": 0.38076869047499096,
|
|
"learning_rate": 3.9799861674351794e-05,
|
|
"loss": 0.1091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06345896422863007,
|
|
"step": 1880,
|
|
"valid_targets_mean": 4150.0,
|
|
"valid_targets_min": 761
|
|
},
|
|
{
|
|
"epoch": 0.9874279727606077,
|
|
"grad_norm": 0.4944113710787087,
|
|
"learning_rate": 3.9796158367306095e-05,
|
|
"loss": 0.1231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06043882295489311,
|
|
"step": 1885,
|
|
"valid_targets_mean": 1368.4,
|
|
"valid_targets_min": 538
|
|
},
|
|
{
|
|
"epoch": 0.9900471451021478,
|
|
"grad_norm": 0.5854892130133516,
|
|
"learning_rate": 3.979242128659373e-05,
|
|
"loss": 0.1399,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11408171057701111,
|
|
"step": 1890,
|
|
"valid_targets_mean": 2630.4,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 0.9926663174436878,
|
|
"grad_norm": 0.37841290083701706,
|
|
"learning_rate": 3.9788650438590425e-05,
|
|
"loss": 0.2078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051241979002952576,
|
|
"step": 1895,
|
|
"valid_targets_mean": 3353.9,
|
|
"valid_targets_min": 2347
|
|
},
|
|
{
|
|
"epoch": 0.9952854897852279,
|
|
"grad_norm": 0.43927741316702157,
|
|
"learning_rate": 3.9784845829729516e-05,
|
|
"loss": 0.1205,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07244317978620529,
|
|
"step": 1900,
|
|
"valid_targets_mean": 3880.2,
|
|
"valid_targets_min": 2473
|
|
},
|
|
{
|
|
"epoch": 0.997904662126768,
|
|
"grad_norm": 0.4138312053437435,
|
|
"learning_rate": 3.978100746650194e-05,
|
|
"loss": 0.119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06079757958650589,
|
|
"step": 1905,
|
|
"valid_targets_mean": 3556.4,
|
|
"valid_targets_min": 2245
|
|
},
|
|
{
|
|
"epoch": 1.000523834468308,
|
|
"grad_norm": 1.194369845882971,
|
|
"learning_rate": 3.977713535545622e-05,
|
|
"loss": 0.1952,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1503610908985138,
|
|
"step": 1910,
|
|
"valid_targets_mean": 1492.2,
|
|
"valid_targets_min": 544
|
|
},
|
|
{
|
|
"epoch": 1.0031430068098481,
|
|
"grad_norm": 0.9818625513655856,
|
|
"learning_rate": 3.9773229503198455e-05,
|
|
"loss": 0.2881,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13344977796077728,
|
|
"step": 1915,
|
|
"valid_targets_mean": 1270.4,
|
|
"valid_targets_min": 632
|
|
},
|
|
{
|
|
"epoch": 1.0057621791513882,
|
|
"grad_norm": 1.0052963104092862,
|
|
"learning_rate": 3.976928991639231e-05,
|
|
"loss": 0.2637,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11686259508132935,
|
|
"step": 1920,
|
|
"valid_targets_mean": 1358.4,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 1.0083813514929283,
|
|
"grad_norm": 0.9433402217216043,
|
|
"learning_rate": 3.9765316601759e-05,
|
|
"loss": 0.2602,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13791446387767792,
|
|
"step": 1925,
|
|
"valid_targets_mean": 1413.0,
|
|
"valid_targets_min": 701
|
|
},
|
|
{
|
|
"epoch": 1.0110005238344684,
|
|
"grad_norm": 0.7703734046875843,
|
|
"learning_rate": 3.976130956607729e-05,
|
|
"loss": 0.2488,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15123215317726135,
|
|
"step": 1930,
|
|
"valid_targets_mean": 1745.6,
|
|
"valid_targets_min": 754
|
|
},
|
|
{
|
|
"epoch": 1.0136196961760084,
|
|
"grad_norm": 0.8156969109616492,
|
|
"learning_rate": 3.9757268816183464e-05,
|
|
"loss": 0.2496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13424146175384521,
|
|
"step": 1935,
|
|
"valid_targets_mean": 1337.9,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 1.0162388685175485,
|
|
"grad_norm": 0.762245798336208,
|
|
"learning_rate": 3.975319435897134e-05,
|
|
"loss": 0.2569,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13536062836647034,
|
|
"step": 1940,
|
|
"valid_targets_mean": 1728.8,
|
|
"valid_targets_min": 970
|
|
},
|
|
{
|
|
"epoch": 1.0188580408590886,
|
|
"grad_norm": 0.7655927328602858,
|
|
"learning_rate": 3.9749086201392224e-05,
|
|
"loss": 0.2354,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1405276507139206,
|
|
"step": 1945,
|
|
"valid_targets_mean": 1569.5,
|
|
"valid_targets_min": 648
|
|
},
|
|
{
|
|
"epoch": 1.0214772132006287,
|
|
"grad_norm": 0.8729326295398971,
|
|
"learning_rate": 3.9744944350454935e-05,
|
|
"loss": 0.2598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15222492814064026,
|
|
"step": 1950,
|
|
"valid_targets_mean": 1538.8,
|
|
"valid_targets_min": 727
|
|
},
|
|
{
|
|
"epoch": 1.0240963855421688,
|
|
"grad_norm": 0.724871270354201,
|
|
"learning_rate": 3.9740768813225765e-05,
|
|
"loss": 0.2335,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09660930931568146,
|
|
"step": 1955,
|
|
"valid_targets_mean": 1519.5,
|
|
"valid_targets_min": 788
|
|
},
|
|
{
|
|
"epoch": 1.0267155578837088,
|
|
"grad_norm": 0.807519620375925,
|
|
"learning_rate": 3.973655959682847e-05,
|
|
"loss": 0.2419,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1028185710310936,
|
|
"step": 1960,
|
|
"valid_targets_mean": 1325.1,
|
|
"valid_targets_min": 627
|
|
},
|
|
{
|
|
"epoch": 1.029334730225249,
|
|
"grad_norm": 0.8629622918174555,
|
|
"learning_rate": 3.9732316708444306e-05,
|
|
"loss": 0.2469,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15766587853431702,
|
|
"step": 1965,
|
|
"valid_targets_mean": 1732.4,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 1.031953902566789,
|
|
"grad_norm": 0.7916278926192974,
|
|
"learning_rate": 3.9728040155311906e-05,
|
|
"loss": 0.2407,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13446414470672607,
|
|
"step": 1970,
|
|
"valid_targets_mean": 1525.6,
|
|
"valid_targets_min": 923
|
|
},
|
|
{
|
|
"epoch": 1.034573074908329,
|
|
"grad_norm": 0.8942876139817988,
|
|
"learning_rate": 3.972372994472741e-05,
|
|
"loss": 0.231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12159556150436401,
|
|
"step": 1975,
|
|
"valid_targets_mean": 1603.5,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 1.0371922472498691,
|
|
"grad_norm": 0.778377599100849,
|
|
"learning_rate": 3.971938608404432e-05,
|
|
"loss": 0.229,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11791212856769562,
|
|
"step": 1980,
|
|
"valid_targets_mean": 1757.0,
|
|
"valid_targets_min": 895
|
|
},
|
|
{
|
|
"epoch": 1.0398114195914092,
|
|
"grad_norm": 0.7636176226363194,
|
|
"learning_rate": 3.9715008580673595e-05,
|
|
"loss": 0.2227,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11560601741075516,
|
|
"step": 1985,
|
|
"valid_targets_mean": 1391.5,
|
|
"valid_targets_min": 1005
|
|
},
|
|
{
|
|
"epoch": 1.0424305919329493,
|
|
"grad_norm": 0.8014038325436506,
|
|
"learning_rate": 3.9710597442083555e-05,
|
|
"loss": 0.2463,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08947145193815231,
|
|
"step": 1990,
|
|
"valid_targets_mean": 1217.1,
|
|
"valid_targets_min": 870
|
|
},
|
|
{
|
|
"epoch": 1.0450497642744894,
|
|
"grad_norm": 0.8699500894472375,
|
|
"learning_rate": 3.970615267579993e-05,
|
|
"loss": 0.2442,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12444935739040375,
|
|
"step": 1995,
|
|
"valid_targets_mean": 1619.1,
|
|
"valid_targets_min": 879
|
|
},
|
|
{
|
|
"epoch": 1.0476689366160294,
|
|
"grad_norm": 0.7436468038034245,
|
|
"learning_rate": 3.97016742894058e-05,
|
|
"loss": 0.2417,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1077117770910263,
|
|
"step": 2000,
|
|
"valid_targets_mean": 1507.5,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 1.0502881089575693,
|
|
"grad_norm": 0.80509332534856,
|
|
"learning_rate": 3.969716229054162e-05,
|
|
"loss": 0.2357,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14222241938114166,
|
|
"step": 2005,
|
|
"valid_targets_mean": 1707.1,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 1.0529072812991094,
|
|
"grad_norm": 0.7612458704908245,
|
|
"learning_rate": 3.969261668690518e-05,
|
|
"loss": 0.232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10558472573757172,
|
|
"step": 2010,
|
|
"valid_targets_mean": 1747.9,
|
|
"valid_targets_min": 832
|
|
},
|
|
{
|
|
"epoch": 1.0555264536406495,
|
|
"grad_norm": 0.6753537286936889,
|
|
"learning_rate": 3.9688037486251615e-05,
|
|
"loss": 0.2381,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09377558529376984,
|
|
"step": 2015,
|
|
"valid_targets_mean": 1581.9,
|
|
"valid_targets_min": 891
|
|
},
|
|
{
|
|
"epoch": 1.0581456259821895,
|
|
"grad_norm": 0.7688444067608663,
|
|
"learning_rate": 3.9683424696393355e-05,
|
|
"loss": 0.2395,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1449514925479889,
|
|
"step": 2020,
|
|
"valid_targets_mean": 1746.8,
|
|
"valid_targets_min": 1052
|
|
},
|
|
{
|
|
"epoch": 1.0607647983237296,
|
|
"grad_norm": 0.7971853412990318,
|
|
"learning_rate": 3.9678778325200164e-05,
|
|
"loss": 0.2225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11333002150058746,
|
|
"step": 2025,
|
|
"valid_targets_mean": 1469.5,
|
|
"valid_targets_min": 737
|
|
},
|
|
{
|
|
"epoch": 1.0633839706652697,
|
|
"grad_norm": 0.7484228655586582,
|
|
"learning_rate": 3.967409838059908e-05,
|
|
"loss": 0.233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11924304068088531,
|
|
"step": 2030,
|
|
"valid_targets_mean": 1762.2,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 1.0660031430068098,
|
|
"grad_norm": 0.7785912990514914,
|
|
"learning_rate": 3.966938487057442e-05,
|
|
"loss": 0.2223,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12280270457267761,
|
|
"step": 2035,
|
|
"valid_targets_mean": 1482.1,
|
|
"valid_targets_min": 805
|
|
},
|
|
{
|
|
"epoch": 1.0686223153483498,
|
|
"grad_norm": 0.6860369029184382,
|
|
"learning_rate": 3.9664637803167775e-05,
|
|
"loss": 0.23,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12503057718276978,
|
|
"step": 2040,
|
|
"valid_targets_mean": 1962.8,
|
|
"valid_targets_min": 1253
|
|
},
|
|
{
|
|
"epoch": 1.07124148768989,
|
|
"grad_norm": 0.7513660699268863,
|
|
"learning_rate": 3.965985718647798e-05,
|
|
"loss": 0.2308,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13035273551940918,
|
|
"step": 2045,
|
|
"valid_targets_mean": 1958.0,
|
|
"valid_targets_min": 1264
|
|
},
|
|
{
|
|
"epoch": 1.07386066003143,
|
|
"grad_norm": 0.8647787964953808,
|
|
"learning_rate": 3.965504302866112e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09549619257450104,
|
|
"step": 2050,
|
|
"valid_targets_mean": 1201.4,
|
|
"valid_targets_min": 613
|
|
},
|
|
{
|
|
"epoch": 1.07647983237297,
|
|
"grad_norm": 0.8658928820184512,
|
|
"learning_rate": 3.965019533793048e-05,
|
|
"loss": 0.2257,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11795791238546371,
|
|
"step": 2055,
|
|
"valid_targets_mean": 1559.0,
|
|
"valid_targets_min": 768
|
|
},
|
|
{
|
|
"epoch": 1.0790990047145101,
|
|
"grad_norm": 0.8542433100391335,
|
|
"learning_rate": 3.964531412255657e-05,
|
|
"loss": 0.225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11437257379293442,
|
|
"step": 2060,
|
|
"valid_targets_mean": 1210.0,
|
|
"valid_targets_min": 714
|
|
},
|
|
{
|
|
"epoch": 1.0817181770560502,
|
|
"grad_norm": 0.859546874634809,
|
|
"learning_rate": 3.96403993908671e-05,
|
|
"loss": 0.2346,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09374754875898361,
|
|
"step": 2065,
|
|
"valid_targets_mean": 1242.2,
|
|
"valid_targets_min": 841
|
|
},
|
|
{
|
|
"epoch": 1.0843373493975903,
|
|
"grad_norm": 0.7231260646891682,
|
|
"learning_rate": 3.963545115124695e-05,
|
|
"loss": 0.2227,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1218329519033432,
|
|
"step": 2070,
|
|
"valid_targets_mean": 1708.1,
|
|
"valid_targets_min": 936
|
|
},
|
|
{
|
|
"epoch": 1.0869565217391304,
|
|
"grad_norm": 1.0871171462012361,
|
|
"learning_rate": 3.963046941213818e-05,
|
|
"loss": 0.2224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10440310835838318,
|
|
"step": 2075,
|
|
"valid_targets_mean": 1163.1,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 1.0895756940806705,
|
|
"grad_norm": 0.8386367512159356,
|
|
"learning_rate": 3.962545418203999e-05,
|
|
"loss": 0.2079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1122165396809578,
|
|
"step": 2080,
|
|
"valid_targets_mean": 1380.4,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 1.0921948664222105,
|
|
"grad_norm": 0.7611728018818952,
|
|
"learning_rate": 3.9620405469508714e-05,
|
|
"loss": 0.2239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14091750979423523,
|
|
"step": 2085,
|
|
"valid_targets_mean": 1625.4,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 1.0948140387637506,
|
|
"grad_norm": 0.7671597940815614,
|
|
"learning_rate": 3.9615323283157825e-05,
|
|
"loss": 0.2215,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08596023917198181,
|
|
"step": 2090,
|
|
"valid_targets_mean": 1153.4,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 1.0974332111052907,
|
|
"grad_norm": 1.0130613386512692,
|
|
"learning_rate": 3.96102076316579e-05,
|
|
"loss": 0.2213,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12400378286838531,
|
|
"step": 2095,
|
|
"valid_targets_mean": 1487.1,
|
|
"valid_targets_min": 831
|
|
},
|
|
{
|
|
"epoch": 1.1000523834468308,
|
|
"grad_norm": 0.7349219723401507,
|
|
"learning_rate": 3.96050585237366e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1018712967634201,
|
|
"step": 2100,
|
|
"valid_targets_mean": 1479.1,
|
|
"valid_targets_min": 1015
|
|
},
|
|
{
|
|
"epoch": 1.1026715557883708,
|
|
"grad_norm": 0.7265585575593051,
|
|
"learning_rate": 3.959987596817868e-05,
|
|
"loss": 0.2255,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1054868996143341,
|
|
"step": 2105,
|
|
"valid_targets_mean": 1366.5,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 1.105290728129911,
|
|
"grad_norm": 0.7170020320852712,
|
|
"learning_rate": 3.9594659973825956e-05,
|
|
"loss": 0.2183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07551706582307816,
|
|
"step": 2110,
|
|
"valid_targets_mean": 1318.1,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 1.107909900471451,
|
|
"grad_norm": 0.8668127695322749,
|
|
"learning_rate": 3.958941054957728e-05,
|
|
"loss": 0.2256,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13176393508911133,
|
|
"step": 2115,
|
|
"valid_targets_mean": 1541.2,
|
|
"valid_targets_min": 931
|
|
},
|
|
{
|
|
"epoch": 1.110529072812991,
|
|
"grad_norm": 0.7426866355102264,
|
|
"learning_rate": 3.958412770438854e-05,
|
|
"loss": 0.2298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11150750517845154,
|
|
"step": 2120,
|
|
"valid_targets_mean": 1790.2,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 1.1131482451545311,
|
|
"grad_norm": 0.79793910226458,
|
|
"learning_rate": 3.957881144727266e-05,
|
|
"loss": 0.2365,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12302378565073013,
|
|
"step": 2125,
|
|
"valid_targets_mean": 1511.6,
|
|
"valid_targets_min": 748
|
|
},
|
|
{
|
|
"epoch": 1.1157674174960712,
|
|
"grad_norm": 0.8529625169049818,
|
|
"learning_rate": 3.957346178729956e-05,
|
|
"loss": 0.2214,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11573164165019989,
|
|
"step": 2130,
|
|
"valid_targets_mean": 1568.1,
|
|
"valid_targets_min": 699
|
|
},
|
|
{
|
|
"epoch": 1.1183865898376113,
|
|
"grad_norm": 0.852495053036662,
|
|
"learning_rate": 3.956807873359613e-05,
|
|
"loss": 0.2162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09075269103050232,
|
|
"step": 2135,
|
|
"valid_targets_mean": 1002.4,
|
|
"valid_targets_min": 534
|
|
},
|
|
{
|
|
"epoch": 1.1210057621791514,
|
|
"grad_norm": 0.7757396558566536,
|
|
"learning_rate": 3.9562662295346256e-05,
|
|
"loss": 0.2281,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10594692826271057,
|
|
"step": 2140,
|
|
"valid_targets_mean": 1367.1,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 1.1236249345206915,
|
|
"grad_norm": 0.7465660688593253,
|
|
"learning_rate": 3.955721248179076e-05,
|
|
"loss": 0.2211,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08780340850353241,
|
|
"step": 2145,
|
|
"valid_targets_mean": 1259.8,
|
|
"valid_targets_min": 835
|
|
},
|
|
{
|
|
"epoch": 1.1262441068622315,
|
|
"grad_norm": 0.7539584899370625,
|
|
"learning_rate": 3.9551729302227436e-05,
|
|
"loss": 0.2138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10159491002559662,
|
|
"step": 2150,
|
|
"valid_targets_mean": 1330.9,
|
|
"valid_targets_min": 829
|
|
},
|
|
{
|
|
"epoch": 1.1288632792037716,
|
|
"grad_norm": 0.8467666052896952,
|
|
"learning_rate": 3.9546212766010956e-05,
|
|
"loss": 0.2124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11529254913330078,
|
|
"step": 2155,
|
|
"valid_targets_mean": 1756.4,
|
|
"valid_targets_min": 1138
|
|
},
|
|
{
|
|
"epoch": 1.1314824515453117,
|
|
"grad_norm": 0.6862128095142289,
|
|
"learning_rate": 3.9540662882552944e-05,
|
|
"loss": 0.227,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13152453303337097,
|
|
"step": 2160,
|
|
"valid_targets_mean": 1799.8,
|
|
"valid_targets_min": 1174
|
|
},
|
|
{
|
|
"epoch": 1.1341016238868518,
|
|
"grad_norm": 0.8123483728194234,
|
|
"learning_rate": 3.95350796613219e-05,
|
|
"loss": 0.2204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10572031140327454,
|
|
"step": 2165,
|
|
"valid_targets_mean": 1524.5,
|
|
"valid_targets_min": 554
|
|
},
|
|
{
|
|
"epoch": 1.1367207962283918,
|
|
"grad_norm": 0.747590190784486,
|
|
"learning_rate": 3.9529463111843186e-05,
|
|
"loss": 0.2185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10925063490867615,
|
|
"step": 2170,
|
|
"valid_targets_mean": 1367.2,
|
|
"valid_targets_min": 843
|
|
},
|
|
{
|
|
"epoch": 1.139339968569932,
|
|
"grad_norm": 0.9388007763174887,
|
|
"learning_rate": 3.952381324369905e-05,
|
|
"loss": 0.2129,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10343623161315918,
|
|
"step": 2175,
|
|
"valid_targets_mean": 1673.1,
|
|
"valid_targets_min": 849
|
|
},
|
|
{
|
|
"epoch": 1.141959140911472,
|
|
"grad_norm": 0.8653252025215202,
|
|
"learning_rate": 3.951813006652856e-05,
|
|
"loss": 0.2246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11683325469493866,
|
|
"step": 2180,
|
|
"valid_targets_mean": 1509.0,
|
|
"valid_targets_min": 612
|
|
},
|
|
{
|
|
"epoch": 1.144578313253012,
|
|
"grad_norm": 0.9278457407371539,
|
|
"learning_rate": 3.951241359002764e-05,
|
|
"loss": 0.2157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10242593288421631,
|
|
"step": 2185,
|
|
"valid_targets_mean": 1401.0,
|
|
"valid_targets_min": 841
|
|
},
|
|
{
|
|
"epoch": 1.1471974855945521,
|
|
"grad_norm": 0.8022037075812134,
|
|
"learning_rate": 3.950666382394901e-05,
|
|
"loss": 0.2157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13213659822940826,
|
|
"step": 2190,
|
|
"valid_targets_mean": 1564.8,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 1.1498166579360922,
|
|
"grad_norm": 0.7516270705204289,
|
|
"learning_rate": 3.9500880778102163e-05,
|
|
"loss": 0.2107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08958139270544052,
|
|
"step": 2195,
|
|
"valid_targets_mean": 1293.1,
|
|
"valid_targets_min": 822
|
|
},
|
|
{
|
|
"epoch": 1.1524358302776323,
|
|
"grad_norm": 0.7547102416422085,
|
|
"learning_rate": 3.949506446235341e-05,
|
|
"loss": 0.1997,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10334242880344391,
|
|
"step": 2200,
|
|
"valid_targets_mean": 1731.0,
|
|
"valid_targets_min": 670
|
|
},
|
|
{
|
|
"epoch": 1.1550550026191724,
|
|
"grad_norm": 1.1197584829502087,
|
|
"learning_rate": 3.9489214886625794e-05,
|
|
"loss": 0.2185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12101603299379349,
|
|
"step": 2205,
|
|
"valid_targets_mean": 1624.1,
|
|
"valid_targets_min": 647
|
|
},
|
|
{
|
|
"epoch": 1.1576741749607125,
|
|
"grad_norm": 1.0909042034318837,
|
|
"learning_rate": 3.948333206089912e-05,
|
|
"loss": 0.2118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08762732148170471,
|
|
"step": 2210,
|
|
"valid_targets_mean": 1429.1,
|
|
"valid_targets_min": 952
|
|
},
|
|
{
|
|
"epoch": 1.1602933473022525,
|
|
"grad_norm": 0.8287334070305442,
|
|
"learning_rate": 3.947741599520989e-05,
|
|
"loss": 0.2141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09083770215511322,
|
|
"step": 2215,
|
|
"valid_targets_mean": 1222.5,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 1.1629125196437926,
|
|
"grad_norm": 0.7422438956164562,
|
|
"learning_rate": 3.947146669965136e-05,
|
|
"loss": 0.2147,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09345854818820953,
|
|
"step": 2220,
|
|
"valid_targets_mean": 1417.8,
|
|
"valid_targets_min": 892
|
|
},
|
|
{
|
|
"epoch": 1.1655316919853327,
|
|
"grad_norm": 0.8067699975678834,
|
|
"learning_rate": 3.9465484184373445e-05,
|
|
"loss": 0.2205,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12448067218065262,
|
|
"step": 2225,
|
|
"valid_targets_mean": 1636.0,
|
|
"valid_targets_min": 985
|
|
},
|
|
{
|
|
"epoch": 1.1681508643268728,
|
|
"grad_norm": 0.7594608853046841,
|
|
"learning_rate": 3.945946845958274e-05,
|
|
"loss": 0.2124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10024244338274002,
|
|
"step": 2230,
|
|
"valid_targets_mean": 1505.0,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 1.1707700366684128,
|
|
"grad_norm": 0.7324083576368025,
|
|
"learning_rate": 3.945341953554251e-05,
|
|
"loss": 0.2139,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09119733422994614,
|
|
"step": 2235,
|
|
"valid_targets_mean": 1286.5,
|
|
"valid_targets_min": 641
|
|
},
|
|
{
|
|
"epoch": 1.173389209009953,
|
|
"grad_norm": 0.7323386488755297,
|
|
"learning_rate": 3.944733742257266e-05,
|
|
"loss": 0.2201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12102872133255005,
|
|
"step": 2240,
|
|
"valid_targets_mean": 1567.5,
|
|
"valid_targets_min": 1020
|
|
},
|
|
{
|
|
"epoch": 1.176008381351493,
|
|
"grad_norm": 0.795772920577501,
|
|
"learning_rate": 3.9441222131049696e-05,
|
|
"loss": 0.2186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09501123428344727,
|
|
"step": 2245,
|
|
"valid_targets_mean": 1252.4,
|
|
"valid_targets_min": 729
|
|
},
|
|
{
|
|
"epoch": 1.178627553693033,
|
|
"grad_norm": 0.829365981248302,
|
|
"learning_rate": 3.943507367140676e-05,
|
|
"loss": 0.2271,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13544471561908722,
|
|
"step": 2250,
|
|
"valid_targets_mean": 1518.2,
|
|
"valid_targets_min": 682
|
|
},
|
|
{
|
|
"epoch": 1.1812467260345731,
|
|
"grad_norm": 0.6910621458994263,
|
|
"learning_rate": 3.9428892054133546e-05,
|
|
"loss": 0.2218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11135991662740707,
|
|
"step": 2255,
|
|
"valid_targets_mean": 1619.2,
|
|
"valid_targets_min": 1260
|
|
},
|
|
{
|
|
"epoch": 1.1838658983761132,
|
|
"grad_norm": 0.7852480038349146,
|
|
"learning_rate": 3.942267728977635e-05,
|
|
"loss": 0.221,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14127010107040405,
|
|
"step": 2260,
|
|
"valid_targets_mean": 1575.4,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 1.1864850707176533,
|
|
"grad_norm": 0.7630474982212205,
|
|
"learning_rate": 3.941642938893801e-05,
|
|
"loss": 0.2185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11815932393074036,
|
|
"step": 2265,
|
|
"valid_targets_mean": 1608.1,
|
|
"valid_targets_min": 801
|
|
},
|
|
{
|
|
"epoch": 1.1891042430591934,
|
|
"grad_norm": 0.6949087723774056,
|
|
"learning_rate": 3.9410148362277884e-05,
|
|
"loss": 0.2097,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1087980642914772,
|
|
"step": 2270,
|
|
"valid_targets_mean": 1414.8,
|
|
"valid_targets_min": 564
|
|
},
|
|
{
|
|
"epoch": 1.1917234154007335,
|
|
"grad_norm": 0.7277531366323761,
|
|
"learning_rate": 3.940383422051185e-05,
|
|
"loss": 0.212,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10609061270952225,
|
|
"step": 2275,
|
|
"valid_targets_mean": 1512.5,
|
|
"valid_targets_min": 651
|
|
},
|
|
{
|
|
"epoch": 1.1943425877422735,
|
|
"grad_norm": 0.730344588104479,
|
|
"learning_rate": 3.93974869744123e-05,
|
|
"loss": 0.2217,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09614178538322449,
|
|
"step": 2280,
|
|
"valid_targets_mean": 1262.4,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 1.1969617600838136,
|
|
"grad_norm": 0.7297937012909113,
|
|
"learning_rate": 3.939110663480809e-05,
|
|
"loss": 0.2102,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09250747412443161,
|
|
"step": 2285,
|
|
"valid_targets_mean": 1230.0,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 1.1995809324253537,
|
|
"grad_norm": 0.7617396957992164,
|
|
"learning_rate": 3.9384693212584524e-05,
|
|
"loss": 0.2052,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0926002711057663,
|
|
"step": 2290,
|
|
"valid_targets_mean": 1497.5,
|
|
"valid_targets_min": 1251
|
|
},
|
|
{
|
|
"epoch": 1.2022001047668938,
|
|
"grad_norm": 0.7539149132777555,
|
|
"learning_rate": 3.9378246718683374e-05,
|
|
"loss": 0.2098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10801123082637787,
|
|
"step": 2295,
|
|
"valid_targets_mean": 1470.8,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 1.2048192771084336,
|
|
"grad_norm": 0.9524084561164579,
|
|
"learning_rate": 3.937176716410281e-05,
|
|
"loss": 0.2294,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10383504629135132,
|
|
"step": 2300,
|
|
"valid_targets_mean": 1469.4,
|
|
"valid_targets_min": 846
|
|
},
|
|
{
|
|
"epoch": 1.207438449449974,
|
|
"grad_norm": 0.7372489587943898,
|
|
"learning_rate": 3.9365254559897426e-05,
|
|
"loss": 0.2046,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11510239541530609,
|
|
"step": 2305,
|
|
"valid_targets_mean": 1774.4,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 1.2100576217915138,
|
|
"grad_norm": 0.6890782437617907,
|
|
"learning_rate": 3.9358708917178194e-05,
|
|
"loss": 0.2135,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09536474198102951,
|
|
"step": 2310,
|
|
"valid_targets_mean": 1558.9,
|
|
"valid_targets_min": 823
|
|
},
|
|
{
|
|
"epoch": 1.212676794133054,
|
|
"grad_norm": 0.6239930558716202,
|
|
"learning_rate": 3.9352130247112444e-05,
|
|
"loss": 0.1889,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1046057790517807,
|
|
"step": 2315,
|
|
"valid_targets_mean": 1687.6,
|
|
"valid_targets_min": 595
|
|
},
|
|
{
|
|
"epoch": 1.215295966474594,
|
|
"grad_norm": 0.7836318798739531,
|
|
"learning_rate": 3.934551856092386e-05,
|
|
"loss": 0.2101,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10032084584236145,
|
|
"step": 2320,
|
|
"valid_targets_mean": 1336.9,
|
|
"valid_targets_min": 868
|
|
},
|
|
{
|
|
"epoch": 1.2179151388161342,
|
|
"grad_norm": 0.8601781857987808,
|
|
"learning_rate": 3.933887386989245e-05,
|
|
"loss": 0.2095,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11940158158540726,
|
|
"step": 2325,
|
|
"valid_targets_mean": 1266.6,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 1.220534311157674,
|
|
"grad_norm": 0.8009716500949716,
|
|
"learning_rate": 3.933219618535454e-05,
|
|
"loss": 0.212,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.129446879029274,
|
|
"step": 2330,
|
|
"valid_targets_mean": 1976.8,
|
|
"valid_targets_min": 1080
|
|
},
|
|
{
|
|
"epoch": 1.2231534834992142,
|
|
"grad_norm": 0.7492745681712601,
|
|
"learning_rate": 3.932548551870273e-05,
|
|
"loss": 0.2084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11948906630277634,
|
|
"step": 2335,
|
|
"valid_targets_mean": 1554.1,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 1.2257726558407542,
|
|
"grad_norm": 0.7850723118104741,
|
|
"learning_rate": 3.9318741881385906e-05,
|
|
"loss": 0.2141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09871551394462585,
|
|
"step": 2340,
|
|
"valid_targets_mean": 1499.6,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 1.2283918281822943,
|
|
"grad_norm": 0.6772448031374775,
|
|
"learning_rate": 3.9311965284909205e-05,
|
|
"loss": 0.2215,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10974691808223724,
|
|
"step": 2345,
|
|
"valid_targets_mean": 1475.5,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 1.2310110005238344,
|
|
"grad_norm": 0.5016573935407069,
|
|
"learning_rate": 3.930515574083397e-05,
|
|
"loss": 0.2265,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12183630466461182,
|
|
"step": 2350,
|
|
"valid_targets_mean": 7107.2,
|
|
"valid_targets_min": 5295
|
|
},
|
|
{
|
|
"epoch": 1.2336301728653745,
|
|
"grad_norm": 0.42133670248430216,
|
|
"learning_rate": 3.929831326077779e-05,
|
|
"loss": 0.2323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12615281343460083,
|
|
"step": 2355,
|
|
"valid_targets_mean": 7098.2,
|
|
"valid_targets_min": 5066
|
|
},
|
|
{
|
|
"epoch": 1.2362493452069145,
|
|
"grad_norm": 0.41383673694225553,
|
|
"learning_rate": 3.929143785641441e-05,
|
|
"loss": 0.2232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1073240339756012,
|
|
"step": 2360,
|
|
"valid_targets_mean": 6838.1,
|
|
"valid_targets_min": 4923
|
|
},
|
|
{
|
|
"epoch": 1.2388685175484546,
|
|
"grad_norm": 0.9018915755468442,
|
|
"learning_rate": 3.9284529539473765e-05,
|
|
"loss": 0.2192,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10324497520923615,
|
|
"step": 2365,
|
|
"valid_targets_mean": 5701.2,
|
|
"valid_targets_min": 3922
|
|
},
|
|
{
|
|
"epoch": 1.2414876898899947,
|
|
"grad_norm": 0.39388164766272543,
|
|
"learning_rate": 3.9277588321741954e-05,
|
|
"loss": 0.2148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10737532377243042,
|
|
"step": 2370,
|
|
"valid_targets_mean": 7340.9,
|
|
"valid_targets_min": 4797
|
|
},
|
|
{
|
|
"epoch": 1.2441068622315348,
|
|
"grad_norm": 0.43312552283028277,
|
|
"learning_rate": 3.927061421506118e-05,
|
|
"loss": 0.2366,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12764379382133484,
|
|
"step": 2375,
|
|
"valid_targets_mean": 6343.9,
|
|
"valid_targets_min": 3853
|
|
},
|
|
{
|
|
"epoch": 1.2467260345730748,
|
|
"grad_norm": 0.4270684045685589,
|
|
"learning_rate": 3.926360723132977e-05,
|
|
"loss": 0.2232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10858067870140076,
|
|
"step": 2380,
|
|
"valid_targets_mean": 6566.2,
|
|
"valid_targets_min": 4586
|
|
},
|
|
{
|
|
"epoch": 1.249345206914615,
|
|
"grad_norm": 0.4067994069431279,
|
|
"learning_rate": 3.9256567382502145e-05,
|
|
"loss": 0.2133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12333779782056808,
|
|
"step": 2385,
|
|
"valid_targets_mean": 7153.6,
|
|
"valid_targets_min": 4728
|
|
},
|
|
{
|
|
"epoch": 1.251964379256155,
|
|
"grad_norm": 0.49740636629506035,
|
|
"learning_rate": 3.9249494680588786e-05,
|
|
"loss": 0.2479,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19038259983062744,
|
|
"step": 2390,
|
|
"valid_targets_mean": 6562.0,
|
|
"valid_targets_min": 5203
|
|
},
|
|
{
|
|
"epoch": 1.254583551597695,
|
|
"grad_norm": 0.3797774326216875,
|
|
"learning_rate": 3.924238913765624e-05,
|
|
"loss": 0.2055,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11123651266098022,
|
|
"step": 2395,
|
|
"valid_targets_mean": 7098.9,
|
|
"valid_targets_min": 4561
|
|
},
|
|
{
|
|
"epoch": 1.2572027239392352,
|
|
"grad_norm": 0.5778136233412741,
|
|
"learning_rate": 3.9235250765827075e-05,
|
|
"loss": 0.2224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13788042962551117,
|
|
"step": 2400,
|
|
"valid_targets_mean": 4960.2,
|
|
"valid_targets_min": 3439
|
|
},
|
|
{
|
|
"epoch": 1.2598218962807752,
|
|
"grad_norm": 0.49608254313307765,
|
|
"learning_rate": 3.922807957727985e-05,
|
|
"loss": 0.2106,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09457670152187347,
|
|
"step": 2405,
|
|
"valid_targets_mean": 5384.2,
|
|
"valid_targets_min": 3635
|
|
},
|
|
{
|
|
"epoch": 1.2624410686223153,
|
|
"grad_norm": 0.41188277730720274,
|
|
"learning_rate": 3.9220875584249135e-05,
|
|
"loss": 0.2181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10995665937662125,
|
|
"step": 2410,
|
|
"valid_targets_mean": 6940.2,
|
|
"valid_targets_min": 4393
|
|
},
|
|
{
|
|
"epoch": 1.2650602409638554,
|
|
"grad_norm": 0.3848453621339323,
|
|
"learning_rate": 3.921363879902546e-05,
|
|
"loss": 0.202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11147625744342804,
|
|
"step": 2415,
|
|
"valid_targets_mean": 6812.4,
|
|
"valid_targets_min": 4749
|
|
},
|
|
{
|
|
"epoch": 1.2676794133053955,
|
|
"grad_norm": 0.48623178931329286,
|
|
"learning_rate": 3.9206369233955304e-05,
|
|
"loss": 0.1989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044521696865558624,
|
|
"step": 2420,
|
|
"valid_targets_mean": 2363.0,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 1.2702985856469355,
|
|
"grad_norm": 0.41502605372146767,
|
|
"learning_rate": 3.919906690144106e-05,
|
|
"loss": 0.1997,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09982247650623322,
|
|
"step": 2425,
|
|
"valid_targets_mean": 5841.2,
|
|
"valid_targets_min": 3119
|
|
},
|
|
{
|
|
"epoch": 1.2729177579884756,
|
|
"grad_norm": 0.4296180929373709,
|
|
"learning_rate": 3.919173181394103e-05,
|
|
"loss": 0.2129,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11030298471450806,
|
|
"step": 2430,
|
|
"valid_targets_mean": 5761.4,
|
|
"valid_targets_min": 3774
|
|
},
|
|
{
|
|
"epoch": 1.2755369303300157,
|
|
"grad_norm": 0.7229441580441829,
|
|
"learning_rate": 3.9184363983969405e-05,
|
|
"loss": 0.2084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10802686214447021,
|
|
"step": 2435,
|
|
"valid_targets_mean": 6412.4,
|
|
"valid_targets_min": 5293
|
|
},
|
|
{
|
|
"epoch": 1.2781561026715558,
|
|
"grad_norm": 0.3671482076221174,
|
|
"learning_rate": 3.917696342409623e-05,
|
|
"loss": 0.1891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10334502905607224,
|
|
"step": 2440,
|
|
"valid_targets_mean": 6709.1,
|
|
"valid_targets_min": 5665
|
|
},
|
|
{
|
|
"epoch": 1.2807752750130958,
|
|
"grad_norm": 0.44540104176071077,
|
|
"learning_rate": 3.9169530146947396e-05,
|
|
"loss": 0.1959,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10827546566724777,
|
|
"step": 2445,
|
|
"valid_targets_mean": 7344.6,
|
|
"valid_targets_min": 4428
|
|
},
|
|
{
|
|
"epoch": 1.283394447354636,
|
|
"grad_norm": 0.4618878085458157,
|
|
"learning_rate": 3.916206416520459e-05,
|
|
"loss": 0.219,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10369175672531128,
|
|
"step": 2450,
|
|
"valid_targets_mean": 5780.6,
|
|
"valid_targets_min": 4690
|
|
},
|
|
{
|
|
"epoch": 1.286013619696176,
|
|
"grad_norm": 0.4317398263786524,
|
|
"learning_rate": 3.915456549160533e-05,
|
|
"loss": 0.2199,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09932570159435272,
|
|
"step": 2455,
|
|
"valid_targets_mean": 5854.5,
|
|
"valid_targets_min": 4335
|
|
},
|
|
{
|
|
"epoch": 1.288632792037716,
|
|
"grad_norm": 0.41441258686101656,
|
|
"learning_rate": 3.914703413894289e-05,
|
|
"loss": 0.2004,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1022871732711792,
|
|
"step": 2460,
|
|
"valid_targets_mean": 6378.9,
|
|
"valid_targets_min": 5342
|
|
},
|
|
{
|
|
"epoch": 1.2912519643792562,
|
|
"grad_norm": 0.43417833329432376,
|
|
"learning_rate": 3.9139470120066295e-05,
|
|
"loss": 0.2091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0962183028459549,
|
|
"step": 2465,
|
|
"valid_targets_mean": 5756.5,
|
|
"valid_targets_min": 3840
|
|
},
|
|
{
|
|
"epoch": 1.2938711367207962,
|
|
"grad_norm": 0.4349955432381067,
|
|
"learning_rate": 3.9131873447880296e-05,
|
|
"loss": 0.2253,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12103915959596634,
|
|
"step": 2470,
|
|
"valid_targets_mean": 6438.6,
|
|
"valid_targets_min": 5104
|
|
},
|
|
{
|
|
"epoch": 1.2964903090623363,
|
|
"grad_norm": 0.5636207800862163,
|
|
"learning_rate": 3.912424413534537e-05,
|
|
"loss": 0.2422,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09318073838949203,
|
|
"step": 2475,
|
|
"valid_targets_mean": 5190.4,
|
|
"valid_targets_min": 4224
|
|
},
|
|
{
|
|
"epoch": 1.2991094814038764,
|
|
"grad_norm": 0.41352492240745525,
|
|
"learning_rate": 3.9116582195477676e-05,
|
|
"loss": 0.2299,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10306873917579651,
|
|
"step": 2480,
|
|
"valid_targets_mean": 5932.8,
|
|
"valid_targets_min": 4318
|
|
},
|
|
{
|
|
"epoch": 1.3017286537454165,
|
|
"grad_norm": 0.3912057260667177,
|
|
"learning_rate": 3.910888764134902e-05,
|
|
"loss": 0.2104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10825017094612122,
|
|
"step": 2485,
|
|
"valid_targets_mean": 6348.5,
|
|
"valid_targets_min": 4783
|
|
},
|
|
{
|
|
"epoch": 1.3043478260869565,
|
|
"grad_norm": 0.4242389807432637,
|
|
"learning_rate": 3.9101160486086874e-05,
|
|
"loss": 0.2126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10528185218572617,
|
|
"step": 2490,
|
|
"valid_targets_mean": 7200.1,
|
|
"valid_targets_min": 4412
|
|
},
|
|
{
|
|
"epoch": 1.3069669984284966,
|
|
"grad_norm": 0.5655384901625239,
|
|
"learning_rate": 3.909340074287431e-05,
|
|
"loss": 0.2178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11472706496715546,
|
|
"step": 2495,
|
|
"valid_targets_mean": 6137.5,
|
|
"valid_targets_min": 5033
|
|
},
|
|
{
|
|
"epoch": 1.3095861707700367,
|
|
"grad_norm": 0.45341216141074125,
|
|
"learning_rate": 3.908560842495002e-05,
|
|
"loss": 0.2104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09808801114559174,
|
|
"step": 2500,
|
|
"valid_targets_mean": 5695.5,
|
|
"valid_targets_min": 4822
|
|
},
|
|
{
|
|
"epoch": 1.3122053431115768,
|
|
"grad_norm": 0.47046598181440774,
|
|
"learning_rate": 3.9077783545608244e-05,
|
|
"loss": 0.2132,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11489500105381012,
|
|
"step": 2505,
|
|
"valid_targets_mean": 5906.1,
|
|
"valid_targets_min": 2364
|
|
},
|
|
{
|
|
"epoch": 1.3148245154531168,
|
|
"grad_norm": 0.5262879452966152,
|
|
"learning_rate": 3.906992611819879e-05,
|
|
"loss": 0.2192,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11242279410362244,
|
|
"step": 2510,
|
|
"valid_targets_mean": 6359.0,
|
|
"valid_targets_min": 4793
|
|
},
|
|
{
|
|
"epoch": 1.317443687794657,
|
|
"grad_norm": 0.5424578345324054,
|
|
"learning_rate": 3.9062036156127e-05,
|
|
"loss": 0.2113,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11259541660547256,
|
|
"step": 2515,
|
|
"valid_targets_mean": 6104.9,
|
|
"valid_targets_min": 4733
|
|
},
|
|
{
|
|
"epoch": 1.320062860136197,
|
|
"grad_norm": 0.3852376094367663,
|
|
"learning_rate": 3.905411367285371e-05,
|
|
"loss": 0.1932,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09048554301261902,
|
|
"step": 2520,
|
|
"valid_targets_mean": 6780.8,
|
|
"valid_targets_min": 5388
|
|
},
|
|
{
|
|
"epoch": 1.322682032477737,
|
|
"grad_norm": 0.441983088883469,
|
|
"learning_rate": 3.9046158681895245e-05,
|
|
"loss": 0.1885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10433575510978699,
|
|
"step": 2525,
|
|
"valid_targets_mean": 5956.2,
|
|
"valid_targets_min": 3182
|
|
},
|
|
{
|
|
"epoch": 1.3253012048192772,
|
|
"grad_norm": 0.39460410967625675,
|
|
"learning_rate": 3.90381711968234e-05,
|
|
"loss": 0.2164,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1174294576048851,
|
|
"step": 2530,
|
|
"valid_targets_mean": 6389.4,
|
|
"valid_targets_min": 5077
|
|
},
|
|
{
|
|
"epoch": 1.3279203771608172,
|
|
"grad_norm": 0.5421430229515332,
|
|
"learning_rate": 3.903015123126538e-05,
|
|
"loss": 0.2172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12288212776184082,
|
|
"step": 2535,
|
|
"valid_targets_mean": 5607.8,
|
|
"valid_targets_min": 3663
|
|
},
|
|
{
|
|
"epoch": 1.3305395495023573,
|
|
"grad_norm": 0.43894542574203654,
|
|
"learning_rate": 3.9022098798903846e-05,
|
|
"loss": 0.2039,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09502558410167694,
|
|
"step": 2540,
|
|
"valid_targets_mean": 5616.4,
|
|
"valid_targets_min": 4105
|
|
},
|
|
{
|
|
"epoch": 1.3331587218438974,
|
|
"grad_norm": 0.3773569176447065,
|
|
"learning_rate": 3.90140139134768e-05,
|
|
"loss": 0.2049,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10998235642910004,
|
|
"step": 2545,
|
|
"valid_targets_mean": 7271.0,
|
|
"valid_targets_min": 5201
|
|
},
|
|
{
|
|
"epoch": 1.3357778941854375,
|
|
"grad_norm": 0.4156053262329772,
|
|
"learning_rate": 3.900589658877765e-05,
|
|
"loss": 0.1946,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11112470924854279,
|
|
"step": 2550,
|
|
"valid_targets_mean": 6797.6,
|
|
"valid_targets_min": 5069
|
|
},
|
|
{
|
|
"epoch": 1.3383970665269775,
|
|
"grad_norm": 0.41824839708799705,
|
|
"learning_rate": 3.899774683865513e-05,
|
|
"loss": 0.2065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10233832150697708,
|
|
"step": 2555,
|
|
"valid_targets_mean": 5787.4,
|
|
"valid_targets_min": 4729
|
|
},
|
|
{
|
|
"epoch": 1.3410162388685176,
|
|
"grad_norm": 0.44548469125969525,
|
|
"learning_rate": 3.898956467701331e-05,
|
|
"loss": 0.2124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10589951276779175,
|
|
"step": 2560,
|
|
"valid_targets_mean": 5989.0,
|
|
"valid_targets_min": 4567
|
|
},
|
|
{
|
|
"epoch": 1.3436354112100577,
|
|
"grad_norm": 0.41009003012121287,
|
|
"learning_rate": 3.8981350117811525e-05,
|
|
"loss": 0.1973,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09308187663555145,
|
|
"step": 2565,
|
|
"valid_targets_mean": 5873.4,
|
|
"valid_targets_min": 4464
|
|
},
|
|
{
|
|
"epoch": 1.3462545835515978,
|
|
"grad_norm": 0.4212542692971494,
|
|
"learning_rate": 3.897310317506441e-05,
|
|
"loss": 0.2176,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10250398516654968,
|
|
"step": 2570,
|
|
"valid_targets_mean": 5373.8,
|
|
"valid_targets_min": 4908
|
|
},
|
|
{
|
|
"epoch": 1.3488737558931378,
|
|
"grad_norm": 0.4236756568668427,
|
|
"learning_rate": 3.8964823862841855e-05,
|
|
"loss": 0.1932,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0861271470785141,
|
|
"step": 2575,
|
|
"valid_targets_mean": 5441.6,
|
|
"valid_targets_min": 4443
|
|
},
|
|
{
|
|
"epoch": 1.351492928234678,
|
|
"grad_norm": 0.4347939680190683,
|
|
"learning_rate": 3.8956512195268936e-05,
|
|
"loss": 0.1779,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08970703184604645,
|
|
"step": 2580,
|
|
"valid_targets_mean": 6286.8,
|
|
"valid_targets_min": 3381
|
|
},
|
|
{
|
|
"epoch": 1.3541121005762178,
|
|
"grad_norm": 0.40095729731188445,
|
|
"learning_rate": 3.894816818652598e-05,
|
|
"loss": 0.1836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08964194357395172,
|
|
"step": 2585,
|
|
"valid_targets_mean": 6162.8,
|
|
"valid_targets_min": 4790
|
|
},
|
|
{
|
|
"epoch": 1.356731272917758,
|
|
"grad_norm": 0.40418046170075766,
|
|
"learning_rate": 3.893979185084844e-05,
|
|
"loss": 0.1879,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09120337665081024,
|
|
"step": 2590,
|
|
"valid_targets_mean": 6336.1,
|
|
"valid_targets_min": 5161
|
|
},
|
|
{
|
|
"epoch": 1.359350445259298,
|
|
"grad_norm": 0.4242269315409445,
|
|
"learning_rate": 3.893138320252698e-05,
|
|
"loss": 0.2068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09119582176208496,
|
|
"step": 2595,
|
|
"valid_targets_mean": 6038.0,
|
|
"valid_targets_min": 4491
|
|
},
|
|
{
|
|
"epoch": 1.3619696176008382,
|
|
"grad_norm": 0.9708595853734081,
|
|
"learning_rate": 3.8922942255907316e-05,
|
|
"loss": 0.2407,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1414804458618164,
|
|
"step": 2600,
|
|
"valid_targets_mean": 1423.0,
|
|
"valid_targets_min": 817
|
|
},
|
|
{
|
|
"epoch": 1.364588789942378,
|
|
"grad_norm": 0.4477758125594203,
|
|
"learning_rate": 3.891446902539033e-05,
|
|
"loss": 0.2182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09902373701334,
|
|
"step": 2605,
|
|
"valid_targets_mean": 5531.1,
|
|
"valid_targets_min": 3893
|
|
},
|
|
{
|
|
"epoch": 1.3672079622839184,
|
|
"grad_norm": 0.44661837506019453,
|
|
"learning_rate": 3.8905963525431975e-05,
|
|
"loss": 0.1975,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1093834713101387,
|
|
"step": 2610,
|
|
"valid_targets_mean": 6276.2,
|
|
"valid_targets_min": 4622
|
|
},
|
|
{
|
|
"epoch": 1.3698271346254582,
|
|
"grad_norm": 0.395648111157185,
|
|
"learning_rate": 3.889742577054321e-05,
|
|
"loss": 0.2055,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10112433135509491,
|
|
"step": 2615,
|
|
"valid_targets_mean": 8019.8,
|
|
"valid_targets_min": 6108
|
|
},
|
|
{
|
|
"epoch": 1.3724463069669985,
|
|
"grad_norm": 0.4283627037962096,
|
|
"learning_rate": 3.888885577529009e-05,
|
|
"loss": 0.2093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11028099060058594,
|
|
"step": 2620,
|
|
"valid_targets_mean": 6822.1,
|
|
"valid_targets_min": 4253
|
|
},
|
|
{
|
|
"epoch": 1.3750654793085384,
|
|
"grad_norm": 0.40795304311294595,
|
|
"learning_rate": 3.888025355429362e-05,
|
|
"loss": 0.2198,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10736887156963348,
|
|
"step": 2625,
|
|
"valid_targets_mean": 6800.6,
|
|
"valid_targets_min": 4990
|
|
},
|
|
{
|
|
"epoch": 1.3776846516500787,
|
|
"grad_norm": 0.4158131427344537,
|
|
"learning_rate": 3.8871619122229816e-05,
|
|
"loss": 0.2038,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12992602586746216,
|
|
"step": 2630,
|
|
"valid_targets_mean": 6860.5,
|
|
"valid_targets_min": 4957
|
|
},
|
|
{
|
|
"epoch": 1.3803038239916186,
|
|
"grad_norm": 0.3956084333567644,
|
|
"learning_rate": 3.886295249382964e-05,
|
|
"loss": 0.2086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09402551501989365,
|
|
"step": 2635,
|
|
"valid_targets_mean": 5820.2,
|
|
"valid_targets_min": 4775
|
|
},
|
|
{
|
|
"epoch": 1.3829229963331588,
|
|
"grad_norm": 0.4269057881551316,
|
|
"learning_rate": 3.885425368387898e-05,
|
|
"loss": 0.1963,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07762826979160309,
|
|
"step": 2640,
|
|
"valid_targets_mean": 5628.8,
|
|
"valid_targets_min": 4098
|
|
},
|
|
{
|
|
"epoch": 1.3855421686746987,
|
|
"grad_norm": 0.41321789055034214,
|
|
"learning_rate": 3.8845522707218626e-05,
|
|
"loss": 0.1895,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10863015055656433,
|
|
"step": 2645,
|
|
"valid_targets_mean": 6170.6,
|
|
"valid_targets_min": 4801
|
|
},
|
|
{
|
|
"epoch": 1.388161341016239,
|
|
"grad_norm": 0.5113842812158871,
|
|
"learning_rate": 3.8836759578744255e-05,
|
|
"loss": 0.2435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21723729372024536,
|
|
"step": 2650,
|
|
"valid_targets_mean": 6196.0,
|
|
"valid_targets_min": 4734
|
|
},
|
|
{
|
|
"epoch": 1.3907805133577789,
|
|
"grad_norm": 0.4380085783797243,
|
|
"learning_rate": 3.882796431340639e-05,
|
|
"loss": 0.2045,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11256542056798935,
|
|
"step": 2655,
|
|
"valid_targets_mean": 6086.4,
|
|
"valid_targets_min": 4400
|
|
},
|
|
{
|
|
"epoch": 1.393399685699319,
|
|
"grad_norm": 0.43940376502954864,
|
|
"learning_rate": 3.881913692621039e-05,
|
|
"loss": 0.1934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10712681710720062,
|
|
"step": 2660,
|
|
"valid_targets_mean": 6544.9,
|
|
"valid_targets_min": 4919
|
|
},
|
|
{
|
|
"epoch": 1.396018858040859,
|
|
"grad_norm": 0.4177688964837717,
|
|
"learning_rate": 3.8810277432216404e-05,
|
|
"loss": 0.2061,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11651991307735443,
|
|
"step": 2665,
|
|
"valid_targets_mean": 6052.2,
|
|
"valid_targets_min": 4611
|
|
},
|
|
{
|
|
"epoch": 1.398638030382399,
|
|
"grad_norm": 0.4027821543625251,
|
|
"learning_rate": 3.880138584653937e-05,
|
|
"loss": 0.2079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09147213399410248,
|
|
"step": 2670,
|
|
"valid_targets_mean": 6934.9,
|
|
"valid_targets_min": 2275
|
|
},
|
|
{
|
|
"epoch": 1.4012572027239392,
|
|
"grad_norm": 0.415163749331779,
|
|
"learning_rate": 3.879246218434896e-05,
|
|
"loss": 0.1836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09008123725652695,
|
|
"step": 2675,
|
|
"valid_targets_mean": 5878.5,
|
|
"valid_targets_min": 4826
|
|
},
|
|
{
|
|
"epoch": 1.4038763750654792,
|
|
"grad_norm": 0.3753579362033299,
|
|
"learning_rate": 3.8783506460869605e-05,
|
|
"loss": 0.1984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08899456262588501,
|
|
"step": 2680,
|
|
"valid_targets_mean": 6784.9,
|
|
"valid_targets_min": 4707
|
|
},
|
|
{
|
|
"epoch": 1.4064955474070193,
|
|
"grad_norm": 0.37948171111073,
|
|
"learning_rate": 3.877451869138039e-05,
|
|
"loss": 0.1824,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12983255088329315,
|
|
"step": 2685,
|
|
"valid_targets_mean": 7245.8,
|
|
"valid_targets_min": 5421
|
|
},
|
|
{
|
|
"epoch": 1.4091147197485594,
|
|
"grad_norm": 0.3919828815276863,
|
|
"learning_rate": 3.876549889121511e-05,
|
|
"loss": 0.1923,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0971972793340683,
|
|
"step": 2690,
|
|
"valid_targets_mean": 7122.0,
|
|
"valid_targets_min": 4949
|
|
},
|
|
{
|
|
"epoch": 1.4117338920900995,
|
|
"grad_norm": 0.37599481909327476,
|
|
"learning_rate": 3.875644707576218e-05,
|
|
"loss": 0.1913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10371506959199905,
|
|
"step": 2695,
|
|
"valid_targets_mean": 6444.5,
|
|
"valid_targets_min": 4831
|
|
},
|
|
{
|
|
"epoch": 1.4143530644316396,
|
|
"grad_norm": 0.42157672366288373,
|
|
"learning_rate": 3.8747363260464644e-05,
|
|
"loss": 0.1975,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11276239156723022,
|
|
"step": 2700,
|
|
"valid_targets_mean": 6454.9,
|
|
"valid_targets_min": 4018
|
|
},
|
|
{
|
|
"epoch": 1.4169722367731796,
|
|
"grad_norm": 0.590647973505511,
|
|
"learning_rate": 3.873824746082016e-05,
|
|
"loss": 0.1874,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07755673676729202,
|
|
"step": 2705,
|
|
"valid_targets_mean": 6851.4,
|
|
"valid_targets_min": 5034
|
|
},
|
|
{
|
|
"epoch": 1.4195914091147197,
|
|
"grad_norm": 0.415589389082563,
|
|
"learning_rate": 3.872909969238092e-05,
|
|
"loss": 0.1966,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11099794507026672,
|
|
"step": 2710,
|
|
"valid_targets_mean": 6674.2,
|
|
"valid_targets_min": 4682
|
|
},
|
|
{
|
|
"epoch": 1.4222105814562598,
|
|
"grad_norm": 0.4868111375104368,
|
|
"learning_rate": 3.871991997075368e-05,
|
|
"loss": 0.2075,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10624338686466217,
|
|
"step": 2715,
|
|
"valid_targets_mean": 5518.1,
|
|
"valid_targets_min": 4764
|
|
},
|
|
{
|
|
"epoch": 1.4248297537977999,
|
|
"grad_norm": 0.5715378448201193,
|
|
"learning_rate": 3.87107083115997e-05,
|
|
"loss": 0.2113,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11575331538915634,
|
|
"step": 2720,
|
|
"valid_targets_mean": 5329.2,
|
|
"valid_targets_min": 2481
|
|
},
|
|
{
|
|
"epoch": 1.42744892613934,
|
|
"grad_norm": 0.5199610461866421,
|
|
"learning_rate": 3.8701464730634734e-05,
|
|
"loss": 0.2856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16606278717517853,
|
|
"step": 2725,
|
|
"valid_targets_mean": 4792.6,
|
|
"valid_targets_min": 702
|
|
},
|
|
{
|
|
"epoch": 1.43006809848088,
|
|
"grad_norm": 0.5535968504065762,
|
|
"learning_rate": 3.8692189243629e-05,
|
|
"loss": 0.3189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17155292630195618,
|
|
"step": 2730,
|
|
"valid_targets_mean": 5322.8,
|
|
"valid_targets_min": 1005
|
|
},
|
|
{
|
|
"epoch": 1.43268727082242,
|
|
"grad_norm": 0.4988420804817056,
|
|
"learning_rate": 3.868288186640714e-05,
|
|
"loss": 0.307,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11296539008617401,
|
|
"step": 2735,
|
|
"valid_targets_mean": 3584.6,
|
|
"valid_targets_min": 2054
|
|
},
|
|
{
|
|
"epoch": 1.4353064431639602,
|
|
"grad_norm": 0.6110701721483975,
|
|
"learning_rate": 3.86735426148482e-05,
|
|
"loss": 0.3078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1350294053554535,
|
|
"step": 2740,
|
|
"valid_targets_mean": 3685.6,
|
|
"valid_targets_min": 1751
|
|
},
|
|
{
|
|
"epoch": 1.4379256155055002,
|
|
"grad_norm": 0.4989938534479867,
|
|
"learning_rate": 3.8664171504885625e-05,
|
|
"loss": 0.3017,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18602502346038818,
|
|
"step": 2745,
|
|
"valid_targets_mean": 5944.2,
|
|
"valid_targets_min": 1970
|
|
},
|
|
{
|
|
"epoch": 1.4405447878470403,
|
|
"grad_norm": 0.47585566879088936,
|
|
"learning_rate": 3.86547685525072e-05,
|
|
"loss": 0.3178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14865697920322418,
|
|
"step": 2750,
|
|
"valid_targets_mean": 5916.0,
|
|
"valid_targets_min": 920
|
|
},
|
|
{
|
|
"epoch": 1.4431639601885804,
|
|
"grad_norm": 0.48408636593936255,
|
|
"learning_rate": 3.864533377375503e-05,
|
|
"loss": 0.298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14260733127593994,
|
|
"step": 2755,
|
|
"valid_targets_mean": 6475.4,
|
|
"valid_targets_min": 2231
|
|
},
|
|
{
|
|
"epoch": 1.4457831325301205,
|
|
"grad_norm": 0.5158666232835837,
|
|
"learning_rate": 3.8635867184725516e-05,
|
|
"loss": 0.302,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11541718989610672,
|
|
"step": 2760,
|
|
"valid_targets_mean": 4043.4,
|
|
"valid_targets_min": 467
|
|
},
|
|
{
|
|
"epoch": 1.4484023048716606,
|
|
"grad_norm": 0.510309128453629,
|
|
"learning_rate": 3.862636880156936e-05,
|
|
"loss": 0.313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16205619275569916,
|
|
"step": 2765,
|
|
"valid_targets_mean": 5549.8,
|
|
"valid_targets_min": 3440
|
|
},
|
|
{
|
|
"epoch": 1.4510214772132006,
|
|
"grad_norm": 0.5221182088210312,
|
|
"learning_rate": 3.8616838640491474e-05,
|
|
"loss": 0.313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1546589732170105,
|
|
"step": 2770,
|
|
"valid_targets_mean": 4736.9,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 1.4536406495547407,
|
|
"grad_norm": 0.5833107024383283,
|
|
"learning_rate": 3.860727671775099e-05,
|
|
"loss": 0.3083,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18122589588165283,
|
|
"step": 2775,
|
|
"valid_targets_mean": 4838.2,
|
|
"valid_targets_min": 1223
|
|
},
|
|
{
|
|
"epoch": 1.4562598218962808,
|
|
"grad_norm": 0.5228591268026902,
|
|
"learning_rate": 3.859768304966125e-05,
|
|
"loss": 0.3041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15430261194705963,
|
|
"step": 2780,
|
|
"valid_targets_mean": 5466.9,
|
|
"valid_targets_min": 1552
|
|
},
|
|
{
|
|
"epoch": 1.4588789942378209,
|
|
"grad_norm": 0.5353294956306439,
|
|
"learning_rate": 3.8588057652589724e-05,
|
|
"loss": 0.31,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17821773886680603,
|
|
"step": 2785,
|
|
"valid_targets_mean": 5166.0,
|
|
"valid_targets_min": 1546
|
|
},
|
|
{
|
|
"epoch": 1.461498166579361,
|
|
"grad_norm": 0.5503796718154574,
|
|
"learning_rate": 3.857840054295804e-05,
|
|
"loss": 0.3037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14941097795963287,
|
|
"step": 2790,
|
|
"valid_targets_mean": 3848.2,
|
|
"valid_targets_min": 1957
|
|
},
|
|
{
|
|
"epoch": 1.464117338920901,
|
|
"grad_norm": 0.5916325559924084,
|
|
"learning_rate": 3.8568711737241925e-05,
|
|
"loss": 0.3174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14574086666107178,
|
|
"step": 2795,
|
|
"valid_targets_mean": 3487.5,
|
|
"valid_targets_min": 1538
|
|
},
|
|
{
|
|
"epoch": 1.466736511262441,
|
|
"grad_norm": 0.5925249004918991,
|
|
"learning_rate": 3.855899125197116e-05,
|
|
"loss": 0.3112,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15048345923423767,
|
|
"step": 2800,
|
|
"valid_targets_mean": 3397.2,
|
|
"valid_targets_min": 1645
|
|
},
|
|
{
|
|
"epoch": 1.4693556836039812,
|
|
"grad_norm": 0.6334952039976055,
|
|
"learning_rate": 3.8549239103729606e-05,
|
|
"loss": 0.3109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13147863745689392,
|
|
"step": 2805,
|
|
"valid_targets_mean": 2859.4,
|
|
"valid_targets_min": 935
|
|
},
|
|
{
|
|
"epoch": 1.4719748559455212,
|
|
"grad_norm": 0.6237687179523252,
|
|
"learning_rate": 3.8539455309155116e-05,
|
|
"loss": 0.3216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16066870093345642,
|
|
"step": 2810,
|
|
"valid_targets_mean": 3490.6,
|
|
"valid_targets_min": 1269
|
|
},
|
|
{
|
|
"epoch": 1.4745940282870613,
|
|
"grad_norm": 0.6604809060385983,
|
|
"learning_rate": 3.852963988493955e-05,
|
|
"loss": 0.3179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14912638068199158,
|
|
"step": 2815,
|
|
"valid_targets_mean": 2860.0,
|
|
"valid_targets_min": 1173
|
|
},
|
|
{
|
|
"epoch": 1.4772132006286014,
|
|
"grad_norm": 0.5470019823441967,
|
|
"learning_rate": 3.851979284782873e-05,
|
|
"loss": 0.3128,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1375359147787094,
|
|
"step": 2820,
|
|
"valid_targets_mean": 3480.9,
|
|
"valid_targets_min": 1392
|
|
},
|
|
{
|
|
"epoch": 1.4798323729701415,
|
|
"grad_norm": 0.6191223562860483,
|
|
"learning_rate": 3.850991421462241e-05,
|
|
"loss": 0.3155,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1769552230834961,
|
|
"step": 2825,
|
|
"valid_targets_mean": 3878.6,
|
|
"valid_targets_min": 1883
|
|
},
|
|
{
|
|
"epoch": 1.4824515453116816,
|
|
"grad_norm": 0.6184574495831373,
|
|
"learning_rate": 3.850000400217424e-05,
|
|
"loss": 0.3064,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17909373342990875,
|
|
"step": 2830,
|
|
"valid_targets_mean": 3721.4,
|
|
"valid_targets_min": 2106
|
|
},
|
|
{
|
|
"epoch": 1.4850707176532216,
|
|
"grad_norm": 0.5242297805984385,
|
|
"learning_rate": 3.849006222739175e-05,
|
|
"loss": 0.2833,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10997025668621063,
|
|
"step": 2835,
|
|
"valid_targets_mean": 3868.5,
|
|
"valid_targets_min": 2222
|
|
},
|
|
{
|
|
"epoch": 1.4876898899947617,
|
|
"grad_norm": 0.631553110720061,
|
|
"learning_rate": 3.848008890723633e-05,
|
|
"loss": 0.3109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14636048674583435,
|
|
"step": 2840,
|
|
"valid_targets_mean": 3010.1,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 1.4903090623363018,
|
|
"grad_norm": 0.5927771226679842,
|
|
"learning_rate": 3.8470084058723174e-05,
|
|
"loss": 0.3086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15214566886425018,
|
|
"step": 2845,
|
|
"valid_targets_mean": 3789.1,
|
|
"valid_targets_min": 771
|
|
},
|
|
{
|
|
"epoch": 1.4929282346778419,
|
|
"grad_norm": 0.4971923591509714,
|
|
"learning_rate": 3.846004769892128e-05,
|
|
"loss": 0.3146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10021981596946716,
|
|
"step": 2850,
|
|
"valid_targets_mean": 3578.9,
|
|
"valid_targets_min": 1774
|
|
},
|
|
{
|
|
"epoch": 1.495547407019382,
|
|
"grad_norm": 0.6577345008187645,
|
|
"learning_rate": 3.844997984495338e-05,
|
|
"loss": 0.3015,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14464107155799866,
|
|
"step": 2855,
|
|
"valid_targets_mean": 2545.8,
|
|
"valid_targets_min": 997
|
|
},
|
|
{
|
|
"epoch": 1.498166579360922,
|
|
"grad_norm": 0.5793133638446858,
|
|
"learning_rate": 3.843988051399598e-05,
|
|
"loss": 0.2951,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15637826919555664,
|
|
"step": 2860,
|
|
"valid_targets_mean": 3667.1,
|
|
"valid_targets_min": 1741
|
|
},
|
|
{
|
|
"epoch": 1.500785751702462,
|
|
"grad_norm": 0.6542856165355548,
|
|
"learning_rate": 3.842974972327925e-05,
|
|
"loss": 0.3092,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15415236353874207,
|
|
"step": 2865,
|
|
"valid_targets_mean": 3606.1,
|
|
"valid_targets_min": 863
|
|
},
|
|
{
|
|
"epoch": 1.503404924044002,
|
|
"grad_norm": 0.5452351556476359,
|
|
"learning_rate": 3.841958749008704e-05,
|
|
"loss": 0.3018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15664133429527283,
|
|
"step": 2870,
|
|
"valid_targets_mean": 3876.9,
|
|
"valid_targets_min": 1854
|
|
},
|
|
{
|
|
"epoch": 1.5060240963855422,
|
|
"grad_norm": 0.6084567799593589,
|
|
"learning_rate": 3.8409393831756857e-05,
|
|
"loss": 0.2932,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14668738842010498,
|
|
"step": 2875,
|
|
"valid_targets_mean": 4763.2,
|
|
"valid_targets_min": 2040
|
|
},
|
|
{
|
|
"epoch": 1.508643268727082,
|
|
"grad_norm": 0.6095390383279379,
|
|
"learning_rate": 3.839916876567981e-05,
|
|
"loss": 0.3043,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1429491937160492,
|
|
"step": 2880,
|
|
"valid_targets_mean": 3521.9,
|
|
"valid_targets_min": 1036
|
|
},
|
|
{
|
|
"epoch": 1.5112624410686224,
|
|
"grad_norm": 0.7525866704807807,
|
|
"learning_rate": 3.838891230930061e-05,
|
|
"loss": 0.3079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15864211320877075,
|
|
"step": 2885,
|
|
"valid_targets_mean": 4304.8,
|
|
"valid_targets_min": 1279
|
|
},
|
|
{
|
|
"epoch": 1.5138816134101623,
|
|
"grad_norm": 0.6314225340741985,
|
|
"learning_rate": 3.837862448011748e-05,
|
|
"loss": 0.2977,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15632294118404388,
|
|
"step": 2890,
|
|
"valid_targets_mean": 3600.0,
|
|
"valid_targets_min": 1798
|
|
},
|
|
{
|
|
"epoch": 1.5165007857517026,
|
|
"grad_norm": 0.566585761284272,
|
|
"learning_rate": 3.836830529568222e-05,
|
|
"loss": 0.3109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13838429749011993,
|
|
"step": 2895,
|
|
"valid_targets_mean": 3067.2,
|
|
"valid_targets_min": 1446
|
|
},
|
|
{
|
|
"epoch": 1.5191199580932424,
|
|
"grad_norm": 0.5597242375298772,
|
|
"learning_rate": 3.835795477360008e-05,
|
|
"loss": 0.3039,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15407949686050415,
|
|
"step": 2900,
|
|
"valid_targets_mean": 4183.2,
|
|
"valid_targets_min": 1270
|
|
},
|
|
{
|
|
"epoch": 1.5217391304347827,
|
|
"grad_norm": 0.492745220516234,
|
|
"learning_rate": 3.834757293152981e-05,
|
|
"loss": 0.303,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15897804498672485,
|
|
"step": 2905,
|
|
"valid_targets_mean": 5485.8,
|
|
"valid_targets_min": 1216
|
|
},
|
|
{
|
|
"epoch": 1.5243583027763226,
|
|
"grad_norm": 0.5832887181140126,
|
|
"learning_rate": 3.833715978718357e-05,
|
|
"loss": 0.2927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17947159707546234,
|
|
"step": 2910,
|
|
"valid_targets_mean": 3994.0,
|
|
"valid_targets_min": 2028
|
|
},
|
|
{
|
|
"epoch": 1.5269774751178629,
|
|
"grad_norm": 0.5992472868552041,
|
|
"learning_rate": 3.832671535832693e-05,
|
|
"loss": 0.3045,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17488516867160797,
|
|
"step": 2915,
|
|
"valid_targets_mean": 3700.6,
|
|
"valid_targets_min": 2329
|
|
},
|
|
{
|
|
"epoch": 1.5295966474594027,
|
|
"grad_norm": 0.608689936567635,
|
|
"learning_rate": 3.831623966277884e-05,
|
|
"loss": 0.3018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19074520468711853,
|
|
"step": 2920,
|
|
"valid_targets_mean": 4075.2,
|
|
"valid_targets_min": 1363
|
|
},
|
|
{
|
|
"epoch": 1.532215819800943,
|
|
"grad_norm": 0.5687522702510394,
|
|
"learning_rate": 3.830573271841157e-05,
|
|
"loss": 0.2939,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1393849402666092,
|
|
"step": 2925,
|
|
"valid_targets_mean": 3560.8,
|
|
"valid_targets_min": 1087
|
|
},
|
|
{
|
|
"epoch": 1.5348349921424829,
|
|
"grad_norm": 0.6557873431198882,
|
|
"learning_rate": 3.829519454315075e-05,
|
|
"loss": 0.308,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1506955921649933,
|
|
"step": 2930,
|
|
"valid_targets_mean": 3860.1,
|
|
"valid_targets_min": 2483
|
|
},
|
|
{
|
|
"epoch": 1.5374541644840232,
|
|
"grad_norm": 0.5802954065582627,
|
|
"learning_rate": 3.828462515497524e-05,
|
|
"loss": 0.3001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17755460739135742,
|
|
"step": 2935,
|
|
"valid_targets_mean": 4583.2,
|
|
"valid_targets_min": 2316
|
|
},
|
|
{
|
|
"epoch": 1.540073336825563,
|
|
"grad_norm": 0.5333037941904751,
|
|
"learning_rate": 3.8274024571917194e-05,
|
|
"loss": 0.2945,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14878308773040771,
|
|
"step": 2940,
|
|
"valid_targets_mean": 4198.9,
|
|
"valid_targets_min": 2022
|
|
},
|
|
{
|
|
"epoch": 1.5426925091671033,
|
|
"grad_norm": 0.5918564145559779,
|
|
"learning_rate": 3.826339281206195e-05,
|
|
"loss": 0.284,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11888591945171356,
|
|
"step": 2945,
|
|
"valid_targets_mean": 3228.9,
|
|
"valid_targets_min": 1294
|
|
},
|
|
{
|
|
"epoch": 1.5453116815086432,
|
|
"grad_norm": 0.7659151280372383,
|
|
"learning_rate": 3.8252729893548066e-05,
|
|
"loss": 0.3103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18149510025978088,
|
|
"step": 2950,
|
|
"valid_targets_mean": 3846.8,
|
|
"valid_targets_min": 2716
|
|
},
|
|
{
|
|
"epoch": 1.5479308538501835,
|
|
"grad_norm": 0.6781972012501118,
|
|
"learning_rate": 3.824203583456723e-05,
|
|
"loss": 0.2912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14599120616912842,
|
|
"step": 2955,
|
|
"valid_targets_mean": 3211.6,
|
|
"valid_targets_min": 2060
|
|
},
|
|
{
|
|
"epoch": 1.5505500261917233,
|
|
"grad_norm": 0.6459462222115947,
|
|
"learning_rate": 3.8231310653364293e-05,
|
|
"loss": 0.3044,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16775554418563843,
|
|
"step": 2960,
|
|
"valid_targets_mean": 3320.4,
|
|
"valid_targets_min": 2407
|
|
},
|
|
{
|
|
"epoch": 1.5531691985332636,
|
|
"grad_norm": 0.583093980020909,
|
|
"learning_rate": 3.822055436823717e-05,
|
|
"loss": 0.3053,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15554597973823547,
|
|
"step": 2965,
|
|
"valid_targets_mean": 4165.1,
|
|
"valid_targets_min": 1533
|
|
},
|
|
{
|
|
"epoch": 1.5557883708748035,
|
|
"grad_norm": 0.5786723076536124,
|
|
"learning_rate": 3.8209766997536867e-05,
|
|
"loss": 0.3031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18110856413841248,
|
|
"step": 2970,
|
|
"valid_targets_mean": 5201.2,
|
|
"valid_targets_min": 2189
|
|
},
|
|
{
|
|
"epoch": 1.5584075432163438,
|
|
"grad_norm": 0.5810527634474358,
|
|
"learning_rate": 3.81989485596674e-05,
|
|
"loss": 0.2958,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18378518521785736,
|
|
"step": 2975,
|
|
"valid_targets_mean": 4193.1,
|
|
"valid_targets_min": 1328
|
|
},
|
|
{
|
|
"epoch": 1.5610267155578836,
|
|
"grad_norm": 0.5793278438385654,
|
|
"learning_rate": 3.81880990730858e-05,
|
|
"loss": 0.31,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21407201886177063,
|
|
"step": 2980,
|
|
"valid_targets_mean": 4500.4,
|
|
"valid_targets_min": 1708
|
|
},
|
|
{
|
|
"epoch": 1.563645887899424,
|
|
"grad_norm": 0.5545862624490651,
|
|
"learning_rate": 3.817721855630209e-05,
|
|
"loss": 0.2871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11643369495868683,
|
|
"step": 2985,
|
|
"valid_targets_mean": 3494.0,
|
|
"valid_targets_min": 1912
|
|
},
|
|
{
|
|
"epoch": 1.5662650602409638,
|
|
"grad_norm": 0.5846290764905854,
|
|
"learning_rate": 3.816630702787919e-05,
|
|
"loss": 0.2841,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.176794171333313,
|
|
"step": 2990,
|
|
"valid_targets_mean": 4145.2,
|
|
"valid_targets_min": 2606
|
|
},
|
|
{
|
|
"epoch": 1.568884232582504,
|
|
"grad_norm": 0.5560977476076939,
|
|
"learning_rate": 3.815536450643296e-05,
|
|
"loss": 0.2833,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1391376554965973,
|
|
"step": 2995,
|
|
"valid_targets_mean": 4189.0,
|
|
"valid_targets_min": 1378
|
|
},
|
|
{
|
|
"epoch": 1.571503404924044,
|
|
"grad_norm": 0.5864475779631305,
|
|
"learning_rate": 3.814439101063212e-05,
|
|
"loss": 0.294,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16940513253211975,
|
|
"step": 3000,
|
|
"valid_targets_mean": 3983.4,
|
|
"valid_targets_min": 1219
|
|
},
|
|
{
|
|
"epoch": 1.574122577265584,
|
|
"grad_norm": 0.5486004743118648,
|
|
"learning_rate": 3.8133386559198255e-05,
|
|
"loss": 0.3001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1555742472410202,
|
|
"step": 3005,
|
|
"valid_targets_mean": 3997.0,
|
|
"valid_targets_min": 2631
|
|
},
|
|
{
|
|
"epoch": 1.576741749607124,
|
|
"grad_norm": 0.59745850047092,
|
|
"learning_rate": 3.812235117090574e-05,
|
|
"loss": 0.2942,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18192443251609802,
|
|
"step": 3010,
|
|
"valid_targets_mean": 4167.4,
|
|
"valid_targets_min": 1423
|
|
},
|
|
{
|
|
"epoch": 1.5793609219486642,
|
|
"grad_norm": 0.5491789317813656,
|
|
"learning_rate": 3.8111284864581734e-05,
|
|
"loss": 0.293,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11260128021240234,
|
|
"step": 3015,
|
|
"valid_targets_mean": 3381.9,
|
|
"valid_targets_min": 2320
|
|
},
|
|
{
|
|
"epoch": 1.5819800942902043,
|
|
"grad_norm": 0.5986153968033281,
|
|
"learning_rate": 3.8100187659106155e-05,
|
|
"loss": 0.3031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15462613105773926,
|
|
"step": 3020,
|
|
"valid_targets_mean": 3174.9,
|
|
"valid_targets_min": 1188
|
|
},
|
|
{
|
|
"epoch": 1.5845992666317443,
|
|
"grad_norm": 0.5360926088484862,
|
|
"learning_rate": 3.808905957341164e-05,
|
|
"loss": 0.3015,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1185324639081955,
|
|
"step": 3025,
|
|
"valid_targets_mean": 3378.6,
|
|
"valid_targets_min": 618
|
|
},
|
|
{
|
|
"epoch": 1.5872184389732844,
|
|
"grad_norm": 0.7183261171782832,
|
|
"learning_rate": 3.807790062648349e-05,
|
|
"loss": 0.2985,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1585368812084198,
|
|
"step": 3030,
|
|
"valid_targets_mean": 2988.1,
|
|
"valid_targets_min": 1364
|
|
},
|
|
{
|
|
"epoch": 1.5898376113148245,
|
|
"grad_norm": 0.6104807586407798,
|
|
"learning_rate": 3.806671083735967e-05,
|
|
"loss": 0.2864,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12601983547210693,
|
|
"step": 3035,
|
|
"valid_targets_mean": 3199.1,
|
|
"valid_targets_min": 1397
|
|
},
|
|
{
|
|
"epoch": 1.5924567836563646,
|
|
"grad_norm": 0.5254606303988886,
|
|
"learning_rate": 3.805549022513077e-05,
|
|
"loss": 0.2817,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1630454957485199,
|
|
"step": 3040,
|
|
"valid_targets_mean": 5348.0,
|
|
"valid_targets_min": 1778
|
|
},
|
|
{
|
|
"epoch": 1.5950759559979046,
|
|
"grad_norm": 0.6031622636026757,
|
|
"learning_rate": 3.804423880893995e-05,
|
|
"loss": 0.2902,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15097734332084656,
|
|
"step": 3045,
|
|
"valid_targets_mean": 3444.9,
|
|
"valid_targets_min": 621
|
|
},
|
|
{
|
|
"epoch": 1.5976951283394447,
|
|
"grad_norm": 0.5724648411597208,
|
|
"learning_rate": 3.803295660798294e-05,
|
|
"loss": 0.2763,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12265162169933319,
|
|
"step": 3050,
|
|
"valid_targets_mean": 3872.9,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 1.6003143006809848,
|
|
"grad_norm": 0.7795690843114012,
|
|
"learning_rate": 3.802164364150799e-05,
|
|
"loss": 0.2813,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13530173897743225,
|
|
"step": 3055,
|
|
"valid_targets_mean": 2546.6,
|
|
"valid_targets_min": 642
|
|
},
|
|
{
|
|
"epoch": 1.6029334730225249,
|
|
"grad_norm": 0.6840637976254862,
|
|
"learning_rate": 3.801029992881582e-05,
|
|
"loss": 0.2794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.149607315659523,
|
|
"step": 3060,
|
|
"valid_targets_mean": 3428.2,
|
|
"valid_targets_min": 1978
|
|
},
|
|
{
|
|
"epoch": 1.605552645364065,
|
|
"grad_norm": 0.6326176079391664,
|
|
"learning_rate": 3.799892548925963e-05,
|
|
"loss": 0.2958,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19109661877155304,
|
|
"step": 3065,
|
|
"valid_targets_mean": 4160.0,
|
|
"valid_targets_min": 2642
|
|
},
|
|
{
|
|
"epoch": 1.608171817705605,
|
|
"grad_norm": 0.6106544881530916,
|
|
"learning_rate": 3.798752034224502e-05,
|
|
"loss": 0.305,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11555371433496475,
|
|
"step": 3070,
|
|
"valid_targets_mean": 2937.1,
|
|
"valid_targets_min": 1510
|
|
},
|
|
{
|
|
"epoch": 1.610790990047145,
|
|
"grad_norm": 0.7650744871424319,
|
|
"learning_rate": 3.7976084507229993e-05,
|
|
"loss": 0.2901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1490192413330078,
|
|
"step": 3075,
|
|
"valid_targets_mean": 2744.2,
|
|
"valid_targets_min": 987
|
|
},
|
|
{
|
|
"epoch": 1.6134101623886852,
|
|
"grad_norm": 0.6605609372990109,
|
|
"learning_rate": 3.79646180037249e-05,
|
|
"loss": 0.2927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13533622026443481,
|
|
"step": 3080,
|
|
"valid_targets_mean": 3638.4,
|
|
"valid_targets_min": 1428
|
|
},
|
|
{
|
|
"epoch": 1.6160293347302253,
|
|
"grad_norm": 0.5621241160860045,
|
|
"learning_rate": 3.7953120851292434e-05,
|
|
"loss": 0.2865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1694134920835495,
|
|
"step": 3085,
|
|
"valid_targets_mean": 4731.5,
|
|
"valid_targets_min": 1700
|
|
},
|
|
{
|
|
"epoch": 1.6186485070717653,
|
|
"grad_norm": 0.4700900499169939,
|
|
"learning_rate": 3.794159306954754e-05,
|
|
"loss": 0.2795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13708017766475677,
|
|
"step": 3090,
|
|
"valid_targets_mean": 5699.6,
|
|
"valid_targets_min": 1767
|
|
},
|
|
{
|
|
"epoch": 1.6212676794133054,
|
|
"grad_norm": 0.49097823508639377,
|
|
"learning_rate": 3.793003467815746e-05,
|
|
"loss": 0.2907,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12990702688694,
|
|
"step": 3095,
|
|
"valid_targets_mean": 4101.5,
|
|
"valid_targets_min": 2684
|
|
},
|
|
{
|
|
"epoch": 1.6238868517548455,
|
|
"grad_norm": 0.6395190214185987,
|
|
"learning_rate": 3.791844569684163e-05,
|
|
"loss": 0.2807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12870249152183533,
|
|
"step": 3100,
|
|
"valid_targets_mean": 2761.2,
|
|
"valid_targets_min": 1542
|
|
},
|
|
{
|
|
"epoch": 1.6265060240963856,
|
|
"grad_norm": 0.6030725114729525,
|
|
"learning_rate": 3.790682614537168e-05,
|
|
"loss": 0.2891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11910074204206467,
|
|
"step": 3105,
|
|
"valid_targets_mean": 2765.8,
|
|
"valid_targets_min": 1476
|
|
},
|
|
{
|
|
"epoch": 1.6291251964379256,
|
|
"grad_norm": 0.5808334324956393,
|
|
"learning_rate": 3.7895176043571406e-05,
|
|
"loss": 0.2862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14055724442005157,
|
|
"step": 3110,
|
|
"valid_targets_mean": 3884.5,
|
|
"valid_targets_min": 1711
|
|
},
|
|
{
|
|
"epoch": 1.6317443687794657,
|
|
"grad_norm": 0.5901410595832308,
|
|
"learning_rate": 3.7883495411316716e-05,
|
|
"loss": 0.297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.147047758102417,
|
|
"step": 3115,
|
|
"valid_targets_mean": 3235.6,
|
|
"valid_targets_min": 1669
|
|
},
|
|
{
|
|
"epoch": 1.6343635411210058,
|
|
"grad_norm": 0.632321140577519,
|
|
"learning_rate": 3.787178426853561e-05,
|
|
"loss": 0.3023,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14954209327697754,
|
|
"step": 3120,
|
|
"valid_targets_mean": 3388.5,
|
|
"valid_targets_min": 1624
|
|
},
|
|
{
|
|
"epoch": 1.6369827134625459,
|
|
"grad_norm": 0.6492849577970939,
|
|
"learning_rate": 3.786004263520812e-05,
|
|
"loss": 0.2919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16489383578300476,
|
|
"step": 3125,
|
|
"valid_targets_mean": 3380.9,
|
|
"valid_targets_min": 1676
|
|
},
|
|
{
|
|
"epoch": 1.639601885804086,
|
|
"grad_norm": 0.6179749169137074,
|
|
"learning_rate": 3.784827053136634e-05,
|
|
"loss": 0.3115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12462452799081802,
|
|
"step": 3130,
|
|
"valid_targets_mean": 3148.4,
|
|
"valid_targets_min": 1500
|
|
},
|
|
{
|
|
"epoch": 1.642221058145626,
|
|
"grad_norm": 0.5304284636990658,
|
|
"learning_rate": 3.783646797709431e-05,
|
|
"loss": 0.2852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13992038369178772,
|
|
"step": 3135,
|
|
"valid_targets_mean": 4343.5,
|
|
"valid_targets_min": 2556
|
|
},
|
|
{
|
|
"epoch": 1.644840230487166,
|
|
"grad_norm": 0.5670035842137175,
|
|
"learning_rate": 3.782463499252804e-05,
|
|
"loss": 0.3041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11929184198379517,
|
|
"step": 3140,
|
|
"valid_targets_mean": 3299.5,
|
|
"valid_targets_min": 1241
|
|
},
|
|
{
|
|
"epoch": 1.6474594028287062,
|
|
"grad_norm": 0.6174416573018309,
|
|
"learning_rate": 3.781277159785544e-05,
|
|
"loss": 0.3095,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18445512652397156,
|
|
"step": 3145,
|
|
"valid_targets_mean": 3547.0,
|
|
"valid_targets_min": 992
|
|
},
|
|
{
|
|
"epoch": 1.6500785751702463,
|
|
"grad_norm": 0.5931150254553617,
|
|
"learning_rate": 3.780087781331632e-05,
|
|
"loss": 0.2893,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15835313498973846,
|
|
"step": 3150,
|
|
"valid_targets_mean": 4328.6,
|
|
"valid_targets_min": 1133
|
|
},
|
|
{
|
|
"epoch": 1.6526977475117863,
|
|
"grad_norm": 0.5789403612166983,
|
|
"learning_rate": 3.7788953659202346e-05,
|
|
"loss": 0.2915,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15773682296276093,
|
|
"step": 3155,
|
|
"valid_targets_mean": 4202.6,
|
|
"valid_targets_min": 2369
|
|
},
|
|
{
|
|
"epoch": 1.6553169198533264,
|
|
"grad_norm": 0.5600810789182009,
|
|
"learning_rate": 3.7776999155856954e-05,
|
|
"loss": 0.2977,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12007777392864227,
|
|
"step": 3160,
|
|
"valid_targets_mean": 3563.0,
|
|
"valid_targets_min": 1919
|
|
},
|
|
{
|
|
"epoch": 1.6579360921948663,
|
|
"grad_norm": 0.6972311923198078,
|
|
"learning_rate": 3.776501432367539e-05,
|
|
"loss": 0.3011,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16129465401172638,
|
|
"step": 3165,
|
|
"valid_targets_mean": 3975.5,
|
|
"valid_targets_min": 2902
|
|
},
|
|
{
|
|
"epoch": 1.6605552645364066,
|
|
"grad_norm": 0.6683336450697415,
|
|
"learning_rate": 3.775299918310466e-05,
|
|
"loss": 0.2989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18373434245586395,
|
|
"step": 3170,
|
|
"valid_targets_mean": 2975.5,
|
|
"valid_targets_min": 617
|
|
},
|
|
{
|
|
"epoch": 1.6631744368779464,
|
|
"grad_norm": 0.5676551411040464,
|
|
"learning_rate": 3.774095375464342e-05,
|
|
"loss": 0.2722,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12359879910945892,
|
|
"step": 3175,
|
|
"valid_targets_mean": 3290.4,
|
|
"valid_targets_min": 1349
|
|
},
|
|
{
|
|
"epoch": 1.6657936092194867,
|
|
"grad_norm": 0.6074011986936779,
|
|
"learning_rate": 3.772887805884207e-05,
|
|
"loss": 0.2822,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1485823392868042,
|
|
"step": 3180,
|
|
"valid_targets_mean": 3459.0,
|
|
"valid_targets_min": 1769
|
|
},
|
|
{
|
|
"epoch": 1.6684127815610266,
|
|
"grad_norm": 0.5214724890996397,
|
|
"learning_rate": 3.7716772116302585e-05,
|
|
"loss": 0.2848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13624684512615204,
|
|
"step": 3185,
|
|
"valid_targets_mean": 4186.1,
|
|
"valid_targets_min": 2097
|
|
},
|
|
{
|
|
"epoch": 1.6710319539025669,
|
|
"grad_norm": 0.6182861701326247,
|
|
"learning_rate": 3.770463594767859e-05,
|
|
"loss": 0.2928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15617293119430542,
|
|
"step": 3190,
|
|
"valid_targets_mean": 3304.0,
|
|
"valid_targets_min": 1066
|
|
},
|
|
{
|
|
"epoch": 1.6736511262441067,
|
|
"grad_norm": 0.5120195811384913,
|
|
"learning_rate": 3.769246957367525e-05,
|
|
"loss": 0.2819,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13785767555236816,
|
|
"step": 3195,
|
|
"valid_targets_mean": 4457.9,
|
|
"valid_targets_min": 2635
|
|
},
|
|
{
|
|
"epoch": 1.676270298585647,
|
|
"grad_norm": 0.5692443938196813,
|
|
"learning_rate": 3.768027301504928e-05,
|
|
"loss": 0.3019,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15586672723293304,
|
|
"step": 3200,
|
|
"valid_targets_mean": 4296.2,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 1.6788894709271869,
|
|
"grad_norm": 0.5517023110408084,
|
|
"learning_rate": 3.7668046292608885e-05,
|
|
"loss": 0.2793,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13861095905303955,
|
|
"step": 3205,
|
|
"valid_targets_mean": 4020.1,
|
|
"valid_targets_min": 2131
|
|
},
|
|
{
|
|
"epoch": 1.6815086432687272,
|
|
"grad_norm": 0.7075129242556276,
|
|
"learning_rate": 3.765578942721372e-05,
|
|
"loss": 0.2917,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14743347465991974,
|
|
"step": 3210,
|
|
"valid_targets_mean": 3541.0,
|
|
"valid_targets_min": 1353
|
|
},
|
|
{
|
|
"epoch": 1.684127815610267,
|
|
"grad_norm": 0.50982123500427,
|
|
"learning_rate": 3.764350243977488e-05,
|
|
"loss": 0.2858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11397814750671387,
|
|
"step": 3215,
|
|
"valid_targets_mean": 3362.0,
|
|
"valid_targets_min": 1509
|
|
},
|
|
{
|
|
"epoch": 1.6867469879518073,
|
|
"grad_norm": 0.5199391510603668,
|
|
"learning_rate": 3.763118535125486e-05,
|
|
"loss": 0.2752,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12962158024311066,
|
|
"step": 3220,
|
|
"valid_targets_mean": 3918.9,
|
|
"valid_targets_min": 2128
|
|
},
|
|
{
|
|
"epoch": 1.6893661602933472,
|
|
"grad_norm": 0.5455390582889343,
|
|
"learning_rate": 3.761883818266748e-05,
|
|
"loss": 0.2715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1064610630273819,
|
|
"step": 3225,
|
|
"valid_targets_mean": 2919.0,
|
|
"valid_targets_min": 1687
|
|
},
|
|
{
|
|
"epoch": 1.6919853326348875,
|
|
"grad_norm": 0.6119127564819121,
|
|
"learning_rate": 3.76064609550779e-05,
|
|
"loss": 0.2781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.161942720413208,
|
|
"step": 3230,
|
|
"valid_targets_mean": 3696.9,
|
|
"valid_targets_min": 1803
|
|
},
|
|
{
|
|
"epoch": 1.6946045049764273,
|
|
"grad_norm": 0.618712313943341,
|
|
"learning_rate": 3.759405368960257e-05,
|
|
"loss": 0.2781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13463839888572693,
|
|
"step": 3235,
|
|
"valid_targets_mean": 4102.0,
|
|
"valid_targets_min": 2263
|
|
},
|
|
{
|
|
"epoch": 1.6972236773179676,
|
|
"grad_norm": 0.5470521788680083,
|
|
"learning_rate": 3.7581616407409156e-05,
|
|
"loss": 0.2811,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16817307472229004,
|
|
"step": 3240,
|
|
"valid_targets_mean": 4621.4,
|
|
"valid_targets_min": 2232
|
|
},
|
|
{
|
|
"epoch": 1.6998428496595075,
|
|
"grad_norm": 0.600468764132397,
|
|
"learning_rate": 3.756914912971657e-05,
|
|
"loss": 0.2717,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16401325166225433,
|
|
"step": 3245,
|
|
"valid_targets_mean": 4051.6,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 1.7024620220010478,
|
|
"grad_norm": 0.5871156474593303,
|
|
"learning_rate": 3.755665187779487e-05,
|
|
"loss": 0.2772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12342692911624908,
|
|
"step": 3250,
|
|
"valid_targets_mean": 3340.1,
|
|
"valid_targets_min": 1942
|
|
},
|
|
{
|
|
"epoch": 1.7050811943425876,
|
|
"grad_norm": 0.618710048195054,
|
|
"learning_rate": 3.754412467296528e-05,
|
|
"loss": 0.2818,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14007382094860077,
|
|
"step": 3255,
|
|
"valid_targets_mean": 3179.4,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 1.707700366684128,
|
|
"grad_norm": 0.584669972621538,
|
|
"learning_rate": 3.7531567536600096e-05,
|
|
"loss": 0.2991,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14600828289985657,
|
|
"step": 3260,
|
|
"valid_targets_mean": 3492.2,
|
|
"valid_targets_min": 1695
|
|
},
|
|
{
|
|
"epoch": 1.7103195390256678,
|
|
"grad_norm": 0.6040311028107254,
|
|
"learning_rate": 3.75189804901227e-05,
|
|
"loss": 0.2834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13014143705368042,
|
|
"step": 3265,
|
|
"valid_targets_mean": 4043.9,
|
|
"valid_targets_min": 1712
|
|
},
|
|
{
|
|
"epoch": 1.712938711367208,
|
|
"grad_norm": 0.5891106352840271,
|
|
"learning_rate": 3.75063635550075e-05,
|
|
"loss": 0.2828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12240810692310333,
|
|
"step": 3270,
|
|
"valid_targets_mean": 2987.9,
|
|
"valid_targets_min": 993
|
|
},
|
|
{
|
|
"epoch": 1.715557883708748,
|
|
"grad_norm": 0.5299023195932661,
|
|
"learning_rate": 3.7493716752779885e-05,
|
|
"loss": 0.274,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14450840651988983,
|
|
"step": 3275,
|
|
"valid_targets_mean": 3854.8,
|
|
"valid_targets_min": 1679
|
|
},
|
|
{
|
|
"epoch": 1.7181770560502883,
|
|
"grad_norm": 0.540981281742683,
|
|
"learning_rate": 3.748104010501621e-05,
|
|
"loss": 0.2837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13237908482551575,
|
|
"step": 3280,
|
|
"valid_targets_mean": 3938.8,
|
|
"valid_targets_min": 2345
|
|
},
|
|
{
|
|
"epoch": 1.720796228391828,
|
|
"grad_norm": 0.5468890533600974,
|
|
"learning_rate": 3.746833363334374e-05,
|
|
"loss": 0.2848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14428837597370148,
|
|
"step": 3285,
|
|
"valid_targets_mean": 4507.9,
|
|
"valid_targets_min": 3068
|
|
},
|
|
{
|
|
"epoch": 1.7234154007333684,
|
|
"grad_norm": 0.5226147361815041,
|
|
"learning_rate": 3.745559735944064e-05,
|
|
"loss": 0.2876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13451524078845978,
|
|
"step": 3290,
|
|
"valid_targets_mean": 4080.9,
|
|
"valid_targets_min": 2965
|
|
},
|
|
{
|
|
"epoch": 1.7260345730749083,
|
|
"grad_norm": 0.5913684135118995,
|
|
"learning_rate": 3.74428313050359e-05,
|
|
"loss": 0.2848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1181318536400795,
|
|
"step": 3295,
|
|
"valid_targets_mean": 2993.5,
|
|
"valid_targets_min": 1817
|
|
},
|
|
{
|
|
"epoch": 1.7286537454164486,
|
|
"grad_norm": 0.5572804765560735,
|
|
"learning_rate": 3.743003549190933e-05,
|
|
"loss": 0.2828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13754969835281372,
|
|
"step": 3300,
|
|
"valid_targets_mean": 3795.2,
|
|
"valid_targets_min": 1727
|
|
},
|
|
{
|
|
"epoch": 1.7312729177579884,
|
|
"grad_norm": 0.5331802199450895,
|
|
"learning_rate": 3.741720994189149e-05,
|
|
"loss": 0.2812,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14808553457260132,
|
|
"step": 3305,
|
|
"valid_targets_mean": 4095.9,
|
|
"valid_targets_min": 2491
|
|
},
|
|
{
|
|
"epoch": 1.7338920900995285,
|
|
"grad_norm": 0.5780295247817255,
|
|
"learning_rate": 3.740435467686371e-05,
|
|
"loss": 0.2245,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12938253581523895,
|
|
"step": 3310,
|
|
"valid_targets_mean": 2046.1,
|
|
"valid_targets_min": 333
|
|
},
|
|
{
|
|
"epoch": 1.7365112624410686,
|
|
"grad_norm": 0.4291125213521186,
|
|
"learning_rate": 3.739146971875798e-05,
|
|
"loss": 0.1269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06559838354587555,
|
|
"step": 3315,
|
|
"valid_targets_mean": 3455.9,
|
|
"valid_targets_min": 712
|
|
},
|
|
{
|
|
"epoch": 1.7391304347826086,
|
|
"grad_norm": 0.49518588227739074,
|
|
"learning_rate": 3.737855508955696e-05,
|
|
"loss": 0.1147,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05264649912714958,
|
|
"step": 3320,
|
|
"valid_targets_mean": 2452.6,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 1.7417496071241487,
|
|
"grad_norm": 0.4791849034140046,
|
|
"learning_rate": 3.736561081129395e-05,
|
|
"loss": 0.1227,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053869135677814484,
|
|
"step": 3325,
|
|
"valid_targets_mean": 2228.9,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 1.7443687794656888,
|
|
"grad_norm": 0.38578294663538254,
|
|
"learning_rate": 3.7352636906052804e-05,
|
|
"loss": 0.1077,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05738483741879463,
|
|
"step": 3330,
|
|
"valid_targets_mean": 3038.6,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 1.7469879518072289,
|
|
"grad_norm": 0.4318151690490142,
|
|
"learning_rate": 3.733963339596795e-05,
|
|
"loss": 0.1175,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06107759475708008,
|
|
"step": 3335,
|
|
"valid_targets_mean": 3049.4,
|
|
"valid_targets_min": 1524
|
|
},
|
|
{
|
|
"epoch": 1.749607124148769,
|
|
"grad_norm": 0.7982300239873034,
|
|
"learning_rate": 3.7326600303224286e-05,
|
|
"loss": 0.1683,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07463137805461884,
|
|
"step": 3340,
|
|
"valid_targets_mean": 1256.1,
|
|
"valid_targets_min": 539
|
|
},
|
|
{
|
|
"epoch": 1.752226296490309,
|
|
"grad_norm": 0.39247918274137605,
|
|
"learning_rate": 3.731353765005723e-05,
|
|
"loss": 0.1285,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054993003606796265,
|
|
"step": 3345,
|
|
"valid_targets_mean": 3597.0,
|
|
"valid_targets_min": 2920
|
|
},
|
|
{
|
|
"epoch": 1.754845468831849,
|
|
"grad_norm": 0.6197858134037353,
|
|
"learning_rate": 3.730044545875258e-05,
|
|
"loss": 0.1259,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06043145805597305,
|
|
"step": 3350,
|
|
"valid_targets_mean": 2737.9,
|
|
"valid_targets_min": 1048
|
|
},
|
|
{
|
|
"epoch": 1.7574646411733892,
|
|
"grad_norm": 0.3204751526751142,
|
|
"learning_rate": 3.728732375164657e-05,
|
|
"loss": 0.1564,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04147946089506149,
|
|
"step": 3355,
|
|
"valid_targets_mean": 4340.1,
|
|
"valid_targets_min": 962
|
|
},
|
|
{
|
|
"epoch": 1.7600838135149293,
|
|
"grad_norm": 0.3381245996923048,
|
|
"learning_rate": 3.727417255112577e-05,
|
|
"loss": 0.0978,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06389027833938599,
|
|
"step": 3360,
|
|
"valid_targets_mean": 4283.1,
|
|
"valid_targets_min": 2091
|
|
},
|
|
{
|
|
"epoch": 1.7627029858564693,
|
|
"grad_norm": 0.3499036909764462,
|
|
"learning_rate": 3.7260991879627077e-05,
|
|
"loss": 0.1097,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05116523802280426,
|
|
"step": 3365,
|
|
"valid_targets_mean": 4320.9,
|
|
"valid_targets_min": 3349
|
|
},
|
|
{
|
|
"epoch": 1.7653221581980094,
|
|
"grad_norm": 0.34840836148116416,
|
|
"learning_rate": 3.724778175963765e-05,
|
|
"loss": 0.1115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05876196548342705,
|
|
"step": 3370,
|
|
"valid_targets_mean": 3895.9,
|
|
"valid_targets_min": 518
|
|
},
|
|
{
|
|
"epoch": 1.7679413305395495,
|
|
"grad_norm": 0.5146942827484791,
|
|
"learning_rate": 3.723454221369491e-05,
|
|
"loss": 0.1585,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06485970318317413,
|
|
"step": 3375,
|
|
"valid_targets_mean": 2219.6,
|
|
"valid_targets_min": 549
|
|
},
|
|
{
|
|
"epoch": 1.7705605028810896,
|
|
"grad_norm": 0.3655799217404372,
|
|
"learning_rate": 3.7221273264386476e-05,
|
|
"loss": 0.1015,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04247612878680229,
|
|
"step": 3380,
|
|
"valid_targets_mean": 3700.2,
|
|
"valid_targets_min": 2724
|
|
},
|
|
{
|
|
"epoch": 1.7731796752226296,
|
|
"grad_norm": 0.5511475710791736,
|
|
"learning_rate": 3.720797493435012e-05,
|
|
"loss": 0.1214,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057143114507198334,
|
|
"step": 3385,
|
|
"valid_targets_mean": 1672.5,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 1.7757988475641697,
|
|
"grad_norm": 0.3703889552448537,
|
|
"learning_rate": 3.7194647246273745e-05,
|
|
"loss": 0.0984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02953612431883812,
|
|
"step": 3390,
|
|
"valid_targets_mean": 3546.2,
|
|
"valid_targets_min": 3183
|
|
},
|
|
{
|
|
"epoch": 1.7784180199057098,
|
|
"grad_norm": 0.44206201540152384,
|
|
"learning_rate": 3.718129022289536e-05,
|
|
"loss": 0.116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052335482090711594,
|
|
"step": 3395,
|
|
"valid_targets_mean": 2473.2,
|
|
"valid_targets_min": 577
|
|
},
|
|
{
|
|
"epoch": 1.7810371922472499,
|
|
"grad_norm": 0.588365869244826,
|
|
"learning_rate": 3.7167903887002984e-05,
|
|
"loss": 0.1298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08485578745603561,
|
|
"step": 3400,
|
|
"valid_targets_mean": 1597.0,
|
|
"valid_targets_min": 557
|
|
},
|
|
{
|
|
"epoch": 1.78365636458879,
|
|
"grad_norm": 0.4247359837759308,
|
|
"learning_rate": 3.715448826143468e-05,
|
|
"loss": 0.1239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06827764213085175,
|
|
"step": 3405,
|
|
"valid_targets_mean": 3019.1,
|
|
"valid_targets_min": 774
|
|
},
|
|
{
|
|
"epoch": 1.78627553693033,
|
|
"grad_norm": 0.4115914246403156,
|
|
"learning_rate": 3.7141043369078466e-05,
|
|
"loss": 0.1267,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052259765565395355,
|
|
"step": 3410,
|
|
"valid_targets_mean": 1825.8,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 1.78889470927187,
|
|
"grad_norm": 0.5027405469936328,
|
|
"learning_rate": 3.712756923287229e-05,
|
|
"loss": 0.1726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08953357487916946,
|
|
"step": 3415,
|
|
"valid_targets_mean": 3686.9,
|
|
"valid_targets_min": 2819
|
|
},
|
|
{
|
|
"epoch": 1.7915138816134102,
|
|
"grad_norm": 0.341914931576673,
|
|
"learning_rate": 3.711406587580399e-05,
|
|
"loss": 0.1053,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04390271008014679,
|
|
"step": 3420,
|
|
"valid_targets_mean": 4137.8,
|
|
"valid_targets_min": 3927
|
|
},
|
|
{
|
|
"epoch": 1.7941330539549503,
|
|
"grad_norm": 0.5547878511262697,
|
|
"learning_rate": 3.710053332091128e-05,
|
|
"loss": 0.1191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07518318295478821,
|
|
"step": 3425,
|
|
"valid_targets_mean": 1376.1,
|
|
"valid_targets_min": 714
|
|
},
|
|
{
|
|
"epoch": 1.7967522262964903,
|
|
"grad_norm": 0.38780164210154305,
|
|
"learning_rate": 3.708697159128165e-05,
|
|
"loss": 0.1511,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053412362933158875,
|
|
"step": 3430,
|
|
"valid_targets_mean": 3432.0,
|
|
"valid_targets_min": 754
|
|
},
|
|
{
|
|
"epoch": 1.7993713986380304,
|
|
"grad_norm": 0.3622569328164762,
|
|
"learning_rate": 3.7073380710052394e-05,
|
|
"loss": 0.1338,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057908304035663605,
|
|
"step": 3435,
|
|
"valid_targets_mean": 3855.2,
|
|
"valid_targets_min": 3178
|
|
},
|
|
{
|
|
"epoch": 1.8019905709795705,
|
|
"grad_norm": 0.7660458310395545,
|
|
"learning_rate": 3.705976070041053e-05,
|
|
"loss": 0.1567,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1433018147945404,
|
|
"step": 3440,
|
|
"valid_targets_mean": 1566.6,
|
|
"valid_targets_min": 920
|
|
},
|
|
{
|
|
"epoch": 1.8046097433211106,
|
|
"grad_norm": 0.4264581704629832,
|
|
"learning_rate": 3.704611158559277e-05,
|
|
"loss": 0.3033,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057927440851926804,
|
|
"step": 3445,
|
|
"valid_targets_mean": 3478.0,
|
|
"valid_targets_min": 2314
|
|
},
|
|
{
|
|
"epoch": 1.8072289156626506,
|
|
"grad_norm": 0.421070094231295,
|
|
"learning_rate": 3.703243338888548e-05,
|
|
"loss": 0.1302,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04688467085361481,
|
|
"step": 3450,
|
|
"valid_targets_mean": 2561.1,
|
|
"valid_targets_min": 833
|
|
},
|
|
{
|
|
"epoch": 1.8098480880041907,
|
|
"grad_norm": 0.4449631252726508,
|
|
"learning_rate": 3.701872613362466e-05,
|
|
"loss": 0.1251,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0818537250161171,
|
|
"step": 3455,
|
|
"valid_targets_mean": 3509.2,
|
|
"valid_targets_min": 1803
|
|
},
|
|
{
|
|
"epoch": 1.8124672603457308,
|
|
"grad_norm": 0.40631985479418165,
|
|
"learning_rate": 3.7004989843195846e-05,
|
|
"loss": 0.2259,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07911115139722824,
|
|
"step": 3460,
|
|
"valid_targets_mean": 4291.4,
|
|
"valid_targets_min": 3089
|
|
},
|
|
{
|
|
"epoch": 1.8150864326872709,
|
|
"grad_norm": 0.2501241430639796,
|
|
"learning_rate": 3.6991224541034155e-05,
|
|
"loss": 0.0919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045642465353012085,
|
|
"step": 3465,
|
|
"valid_targets_mean": 4843.2,
|
|
"valid_targets_min": 2664
|
|
},
|
|
{
|
|
"epoch": 1.8177056050288107,
|
|
"grad_norm": 0.3398211835960956,
|
|
"learning_rate": 3.697743025062418e-05,
|
|
"loss": 0.1262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03657059744000435,
|
|
"step": 3470,
|
|
"valid_targets_mean": 3042.1,
|
|
"valid_targets_min": 560
|
|
},
|
|
{
|
|
"epoch": 1.820324777370351,
|
|
"grad_norm": 0.3910868523973532,
|
|
"learning_rate": 3.696360699549996e-05,
|
|
"loss": 0.1246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0625862404704094,
|
|
"step": 3475,
|
|
"valid_targets_mean": 3863.6,
|
|
"valid_targets_min": 1615
|
|
},
|
|
{
|
|
"epoch": 1.8229439497118909,
|
|
"grad_norm": 0.3456416958516906,
|
|
"learning_rate": 3.694975479924496e-05,
|
|
"loss": 0.1068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03799763694405556,
|
|
"step": 3480,
|
|
"valid_targets_mean": 3312.5,
|
|
"valid_targets_min": 2172
|
|
},
|
|
{
|
|
"epoch": 1.8255631220534312,
|
|
"grad_norm": 0.4206614882492888,
|
|
"learning_rate": 3.693587368549203e-05,
|
|
"loss": 0.1037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04803289473056793,
|
|
"step": 3485,
|
|
"valid_targets_mean": 3059.9,
|
|
"valid_targets_min": 536
|
|
},
|
|
{
|
|
"epoch": 1.828182294394971,
|
|
"grad_norm": 0.3680179904899294,
|
|
"learning_rate": 3.6921963677923336e-05,
|
|
"loss": 0.1096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05506697669625282,
|
|
"step": 3490,
|
|
"valid_targets_mean": 3881.0,
|
|
"valid_targets_min": 3295
|
|
},
|
|
{
|
|
"epoch": 1.8308014667365113,
|
|
"grad_norm": 0.3696429312740505,
|
|
"learning_rate": 3.6908024800270364e-05,
|
|
"loss": 0.1021,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045204613357782364,
|
|
"step": 3495,
|
|
"valid_targets_mean": 3287.2,
|
|
"valid_targets_min": 820
|
|
},
|
|
{
|
|
"epoch": 1.8334206390780512,
|
|
"grad_norm": 0.5566468816324691,
|
|
"learning_rate": 3.689405707631383e-05,
|
|
"loss": 0.1574,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1554620862007141,
|
|
"step": 3500,
|
|
"valid_targets_mean": 2566.6,
|
|
"valid_targets_min": 1046
|
|
},
|
|
{
|
|
"epoch": 1.8360398114195915,
|
|
"grad_norm": 0.33316738802021517,
|
|
"learning_rate": 3.688006052988366e-05,
|
|
"loss": 0.1002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05008360370993614,
|
|
"step": 3505,
|
|
"valid_targets_mean": 3590.1,
|
|
"valid_targets_min": 2719
|
|
},
|
|
{
|
|
"epoch": 1.8386589837611313,
|
|
"grad_norm": 0.6472921695898947,
|
|
"learning_rate": 3.6866035184858996e-05,
|
|
"loss": 0.1043,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08209703117609024,
|
|
"step": 3510,
|
|
"valid_targets_mean": 1638.1,
|
|
"valid_targets_min": 699
|
|
},
|
|
{
|
|
"epoch": 1.8412781561026716,
|
|
"grad_norm": 0.3288394336914071,
|
|
"learning_rate": 3.685198106516806e-05,
|
|
"loss": 0.0967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04378296434879303,
|
|
"step": 3515,
|
|
"valid_targets_mean": 3621.6,
|
|
"valid_targets_min": 2353
|
|
},
|
|
{
|
|
"epoch": 1.8438973284442115,
|
|
"grad_norm": 0.39765837004469934,
|
|
"learning_rate": 3.683789819478819e-05,
|
|
"loss": 0.0999,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05021953955292702,
|
|
"step": 3520,
|
|
"valid_targets_mean": 2328.1,
|
|
"valid_targets_min": 549
|
|
},
|
|
{
|
|
"epoch": 1.8465165007857518,
|
|
"grad_norm": 0.36074171564719126,
|
|
"learning_rate": 3.682378659774579e-05,
|
|
"loss": 0.1633,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04425311088562012,
|
|
"step": 3525,
|
|
"valid_targets_mean": 3330.4,
|
|
"valid_targets_min": 908
|
|
},
|
|
{
|
|
"epoch": 1.8491356731272917,
|
|
"grad_norm": 0.3865231180181791,
|
|
"learning_rate": 3.680964629811624e-05,
|
|
"loss": 0.1318,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08880967646837234,
|
|
"step": 3530,
|
|
"valid_targets_mean": 3458.4,
|
|
"valid_targets_min": 2196
|
|
},
|
|
{
|
|
"epoch": 1.851754845468832,
|
|
"grad_norm": 0.35925425435447844,
|
|
"learning_rate": 3.6795477320023915e-05,
|
|
"loss": 0.1228,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04197625070810318,
|
|
"step": 3535,
|
|
"valid_targets_mean": 1252.1,
|
|
"valid_targets_min": 598
|
|
},
|
|
{
|
|
"epoch": 1.8543740178103718,
|
|
"grad_norm": 0.8289333790703438,
|
|
"learning_rate": 3.6781279687642105e-05,
|
|
"loss": 0.1429,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1264137625694275,
|
|
"step": 3540,
|
|
"valid_targets_mean": 1353.1,
|
|
"valid_targets_min": 578
|
|
},
|
|
{
|
|
"epoch": 1.856993190151912,
|
|
"grad_norm": 0.6176220078989897,
|
|
"learning_rate": 3.676705342519299e-05,
|
|
"loss": 0.1667,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09037410467863083,
|
|
"step": 3545,
|
|
"valid_targets_mean": 1974.2,
|
|
"valid_targets_min": 695
|
|
},
|
|
{
|
|
"epoch": 1.859612362493452,
|
|
"grad_norm": 0.5105167569485973,
|
|
"learning_rate": 3.675279855694759e-05,
|
|
"loss": 0.1142,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049207840114831924,
|
|
"step": 3550,
|
|
"valid_targets_mean": 1724.5,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 1.8622315348349923,
|
|
"grad_norm": 0.3471757300765962,
|
|
"learning_rate": 3.673851510722574e-05,
|
|
"loss": 0.1396,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04608806222677231,
|
|
"step": 3555,
|
|
"valid_targets_mean": 3842.6,
|
|
"valid_targets_min": 2024
|
|
},
|
|
{
|
|
"epoch": 1.8648507071765321,
|
|
"grad_norm": 0.6162071645345287,
|
|
"learning_rate": 3.672420310039603e-05,
|
|
"loss": 0.1231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07409866899251938,
|
|
"step": 3560,
|
|
"valid_targets_mean": 1651.0,
|
|
"valid_targets_min": 523
|
|
},
|
|
{
|
|
"epoch": 1.8674698795180724,
|
|
"grad_norm": 0.31578727458606826,
|
|
"learning_rate": 3.670986256087577e-05,
|
|
"loss": 0.1079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04291585460305214,
|
|
"step": 3565,
|
|
"valid_targets_mean": 3532.5,
|
|
"valid_targets_min": 468
|
|
},
|
|
{
|
|
"epoch": 1.8700890518596123,
|
|
"grad_norm": 0.28444034402660834,
|
|
"learning_rate": 3.669549351313094e-05,
|
|
"loss": 0.1133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028033385053277016,
|
|
"step": 3570,
|
|
"valid_targets_mean": 5298.0,
|
|
"valid_targets_min": 1195
|
|
},
|
|
{
|
|
"epoch": 1.8727082242011526,
|
|
"grad_norm": 0.3262378658558787,
|
|
"learning_rate": 3.668109598167618e-05,
|
|
"loss": 0.1018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0372658334672451,
|
|
"step": 3575,
|
|
"valid_targets_mean": 3562.6,
|
|
"valid_targets_min": 459
|
|
},
|
|
{
|
|
"epoch": 1.8753273965426924,
|
|
"grad_norm": 0.3356185013987389,
|
|
"learning_rate": 3.666666999107471e-05,
|
|
"loss": 0.1013,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04638240486383438,
|
|
"step": 3580,
|
|
"valid_targets_mean": 3548.0,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 1.8779465688842327,
|
|
"grad_norm": 0.421021617725726,
|
|
"learning_rate": 3.6652215565938296e-05,
|
|
"loss": 0.114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06632312387228012,
|
|
"step": 3585,
|
|
"valid_targets_mean": 2130.1,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 1.8805657412257726,
|
|
"grad_norm": 0.4194821722809897,
|
|
"learning_rate": 3.663773273092723e-05,
|
|
"loss": 0.1235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05778392404317856,
|
|
"step": 3590,
|
|
"valid_targets_mean": 3661.6,
|
|
"valid_targets_min": 2759
|
|
},
|
|
{
|
|
"epoch": 1.8831849135673129,
|
|
"grad_norm": 0.5582186855630541,
|
|
"learning_rate": 3.662322151075027e-05,
|
|
"loss": 0.12,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059523727744817734,
|
|
"step": 3595,
|
|
"valid_targets_mean": 3909.5,
|
|
"valid_targets_min": 3391
|
|
},
|
|
{
|
|
"epoch": 1.8858040859088527,
|
|
"grad_norm": 0.3249548361125933,
|
|
"learning_rate": 3.660868193016459e-05,
|
|
"loss": 0.1049,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04768374562263489,
|
|
"step": 3600,
|
|
"valid_targets_mean": 3319.9,
|
|
"valid_targets_min": 1099
|
|
},
|
|
{
|
|
"epoch": 1.888423258250393,
|
|
"grad_norm": 0.3762130049174686,
|
|
"learning_rate": 3.6594114013975773e-05,
|
|
"loss": 0.107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05660580098628998,
|
|
"step": 3605,
|
|
"valid_targets_mean": 3717.0,
|
|
"valid_targets_min": 1874
|
|
},
|
|
{
|
|
"epoch": 1.8910424305919329,
|
|
"grad_norm": 0.3431053442027875,
|
|
"learning_rate": 3.6579517787037716e-05,
|
|
"loss": 0.1232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05007211118936539,
|
|
"step": 3610,
|
|
"valid_targets_mean": 3263.5,
|
|
"valid_targets_min": 724
|
|
},
|
|
{
|
|
"epoch": 1.8936616029334732,
|
|
"grad_norm": 0.28758993152211504,
|
|
"learning_rate": 3.656489327425264e-05,
|
|
"loss": 0.0923,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041469305753707886,
|
|
"step": 3615,
|
|
"valid_targets_mean": 3707.5,
|
|
"valid_targets_min": 2769
|
|
},
|
|
{
|
|
"epoch": 1.896280775275013,
|
|
"grad_norm": 0.41053107516302273,
|
|
"learning_rate": 3.6550240500571e-05,
|
|
"loss": 0.1204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05078805983066559,
|
|
"step": 3620,
|
|
"valid_targets_mean": 2815.8,
|
|
"valid_targets_min": 645
|
|
},
|
|
{
|
|
"epoch": 1.8988999476165531,
|
|
"grad_norm": 0.3576607627383553,
|
|
"learning_rate": 3.653555949099149e-05,
|
|
"loss": 0.1134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047421231865882874,
|
|
"step": 3625,
|
|
"valid_targets_mean": 3488.8,
|
|
"valid_targets_min": 600
|
|
},
|
|
{
|
|
"epoch": 1.9015191199580932,
|
|
"grad_norm": 0.378035252875767,
|
|
"learning_rate": 3.652085027056095e-05,
|
|
"loss": 0.1336,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05577198788523674,
|
|
"step": 3630,
|
|
"valid_targets_mean": 3373.4,
|
|
"valid_targets_min": 1648
|
|
},
|
|
{
|
|
"epoch": 1.9041382922996333,
|
|
"grad_norm": 0.4165422980003152,
|
|
"learning_rate": 3.6506112864374394e-05,
|
|
"loss": 0.1171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07040315866470337,
|
|
"step": 3635,
|
|
"valid_targets_mean": 2555.0,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 1.9067574646411733,
|
|
"grad_norm": 0.8415366072166318,
|
|
"learning_rate": 3.6491347297574874e-05,
|
|
"loss": 0.1323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07888585329055786,
|
|
"step": 3640,
|
|
"valid_targets_mean": 867.1,
|
|
"valid_targets_min": 612
|
|
},
|
|
{
|
|
"epoch": 1.9093766369827134,
|
|
"grad_norm": 0.5041458430821508,
|
|
"learning_rate": 3.647655359535351e-05,
|
|
"loss": 0.1152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08341926336288452,
|
|
"step": 3645,
|
|
"valid_targets_mean": 2202.1,
|
|
"valid_targets_min": 830
|
|
},
|
|
{
|
|
"epoch": 1.9119958093242535,
|
|
"grad_norm": 0.40226385036764234,
|
|
"learning_rate": 3.646173178294942e-05,
|
|
"loss": 0.1279,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05278739333152771,
|
|
"step": 3650,
|
|
"valid_targets_mean": 3027.0,
|
|
"valid_targets_min": 1146
|
|
},
|
|
{
|
|
"epoch": 1.9146149816657936,
|
|
"grad_norm": 0.4110868173650747,
|
|
"learning_rate": 3.644688188564968e-05,
|
|
"loss": 0.1134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0692356526851654,
|
|
"step": 3655,
|
|
"valid_targets_mean": 3296.6,
|
|
"valid_targets_min": 2760
|
|
},
|
|
{
|
|
"epoch": 1.9172341540073337,
|
|
"grad_norm": 0.592891977918973,
|
|
"learning_rate": 3.643200392878928e-05,
|
|
"loss": 0.1067,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07121525704860687,
|
|
"step": 3660,
|
|
"valid_targets_mean": 1469.5,
|
|
"valid_targets_min": 628
|
|
},
|
|
{
|
|
"epoch": 1.9198533263488737,
|
|
"grad_norm": 0.5921042005229938,
|
|
"learning_rate": 3.641709793775108e-05,
|
|
"loss": 0.1316,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10891153663396835,
|
|
"step": 3665,
|
|
"valid_targets_mean": 1752.1,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 1.9224724986904138,
|
|
"grad_norm": 0.2899199301344974,
|
|
"learning_rate": 3.640216393796578e-05,
|
|
"loss": 0.1038,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0305730439722538,
|
|
"step": 3670,
|
|
"valid_targets_mean": 3863.9,
|
|
"valid_targets_min": 3613
|
|
},
|
|
{
|
|
"epoch": 1.9250916710319539,
|
|
"grad_norm": 0.34772988219079837,
|
|
"learning_rate": 3.6387201954911835e-05,
|
|
"loss": 0.1022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051469407975673676,
|
|
"step": 3675,
|
|
"valid_targets_mean": 4304.2,
|
|
"valid_targets_min": 1773
|
|
},
|
|
{
|
|
"epoch": 1.927710843373494,
|
|
"grad_norm": 0.362820734701823,
|
|
"learning_rate": 3.637221201411549e-05,
|
|
"loss": 0.1102,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058442264795303345,
|
|
"step": 3680,
|
|
"valid_targets_mean": 1971.9,
|
|
"valid_targets_min": 762
|
|
},
|
|
{
|
|
"epoch": 1.930330015715034,
|
|
"grad_norm": 0.39901476818103243,
|
|
"learning_rate": 3.6357194141150636e-05,
|
|
"loss": 0.0967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0717681273818016,
|
|
"step": 3685,
|
|
"valid_targets_mean": 4554.1,
|
|
"valid_targets_min": 946
|
|
},
|
|
{
|
|
"epoch": 1.9329491880565741,
|
|
"grad_norm": 0.36834707097157554,
|
|
"learning_rate": 3.634214836163885e-05,
|
|
"loss": 0.0921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0641884058713913,
|
|
"step": 3690,
|
|
"valid_targets_mean": 5015.5,
|
|
"valid_targets_min": 4104
|
|
},
|
|
{
|
|
"epoch": 1.9355683603981142,
|
|
"grad_norm": 0.37260637980081013,
|
|
"learning_rate": 3.632707470124931e-05,
|
|
"loss": 0.1159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05261462181806564,
|
|
"step": 3695,
|
|
"valid_targets_mean": 3064.4,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 1.9381875327396543,
|
|
"grad_norm": 0.44098208940525413,
|
|
"learning_rate": 3.631197318569877e-05,
|
|
"loss": 0.1288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05642792209982872,
|
|
"step": 3700,
|
|
"valid_targets_mean": 3863.6,
|
|
"valid_targets_min": 3550
|
|
},
|
|
{
|
|
"epoch": 1.9408067050811943,
|
|
"grad_norm": 0.3466346466006047,
|
|
"learning_rate": 3.629684384075148e-05,
|
|
"loss": 0.1094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06260032206773758,
|
|
"step": 3705,
|
|
"valid_targets_mean": 3474.9,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 1.9434258774227344,
|
|
"grad_norm": 0.3386632212967048,
|
|
"learning_rate": 3.6281686692219215e-05,
|
|
"loss": 0.1107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04761439934372902,
|
|
"step": 3710,
|
|
"valid_targets_mean": 2568.8,
|
|
"valid_targets_min": 774
|
|
},
|
|
{
|
|
"epoch": 1.9460450497642745,
|
|
"grad_norm": 0.29558164946962245,
|
|
"learning_rate": 3.626650176596114e-05,
|
|
"loss": 0.1344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045535311102867126,
|
|
"step": 3715,
|
|
"valid_targets_mean": 3486.2,
|
|
"valid_targets_min": 888
|
|
},
|
|
{
|
|
"epoch": 1.9486642221058146,
|
|
"grad_norm": 0.3685079925522479,
|
|
"learning_rate": 3.625128908788383e-05,
|
|
"loss": 0.1055,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05522261559963226,
|
|
"step": 3720,
|
|
"valid_targets_mean": 3371.6,
|
|
"valid_targets_min": 1048
|
|
},
|
|
{
|
|
"epoch": 1.9512833944473547,
|
|
"grad_norm": 0.38643874766158975,
|
|
"learning_rate": 3.6236048683941207e-05,
|
|
"loss": 0.12,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05579588934779167,
|
|
"step": 3725,
|
|
"valid_targets_mean": 2941.5,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 1.9539025667888947,
|
|
"grad_norm": 0.40259470466238484,
|
|
"learning_rate": 3.62207805801345e-05,
|
|
"loss": 0.1144,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06654039025306702,
|
|
"step": 3730,
|
|
"valid_targets_mean": 2359.2,
|
|
"valid_targets_min": 865
|
|
},
|
|
{
|
|
"epoch": 1.9565217391304348,
|
|
"grad_norm": 0.48631724845097407,
|
|
"learning_rate": 3.620548480251218e-05,
|
|
"loss": 0.1605,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04295656830072403,
|
|
"step": 3735,
|
|
"valid_targets_mean": 1747.0,
|
|
"valid_targets_min": 570
|
|
},
|
|
{
|
|
"epoch": 1.9591409114719749,
|
|
"grad_norm": 0.4102309246391661,
|
|
"learning_rate": 3.6190161377169946e-05,
|
|
"loss": 0.1329,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06898366659879684,
|
|
"step": 3740,
|
|
"valid_targets_mean": 3704.2,
|
|
"valid_targets_min": 2811
|
|
},
|
|
{
|
|
"epoch": 1.961760083813515,
|
|
"grad_norm": 0.41579990829819685,
|
|
"learning_rate": 3.6174810330250664e-05,
|
|
"loss": 0.1099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047321662306785583,
|
|
"step": 3745,
|
|
"valid_targets_mean": 1749.1,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 1.964379256155055,
|
|
"grad_norm": 0.36412973997585185,
|
|
"learning_rate": 3.615943168794432e-05,
|
|
"loss": 0.1076,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05411919951438904,
|
|
"step": 3750,
|
|
"valid_targets_mean": 2985.1,
|
|
"valid_targets_min": 1591
|
|
},
|
|
{
|
|
"epoch": 1.9669984284965951,
|
|
"grad_norm": 0.4142314566337443,
|
|
"learning_rate": 3.6144025476487976e-05,
|
|
"loss": 0.1064,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04836519807577133,
|
|
"step": 3755,
|
|
"valid_targets_mean": 2912.6,
|
|
"valid_targets_min": 800
|
|
},
|
|
{
|
|
"epoch": 1.9696176008381352,
|
|
"grad_norm": 0.35907542692044886,
|
|
"learning_rate": 3.612859172216574e-05,
|
|
"loss": 0.0994,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055306024849414825,
|
|
"step": 3760,
|
|
"valid_targets_mean": 3084.9,
|
|
"valid_targets_min": 860
|
|
},
|
|
{
|
|
"epoch": 1.9722367731796753,
|
|
"grad_norm": 0.4488371386465046,
|
|
"learning_rate": 3.61131304513087e-05,
|
|
"loss": 0.1388,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06375181674957275,
|
|
"step": 3765,
|
|
"valid_targets_mean": 2482.6,
|
|
"valid_targets_min": 878
|
|
},
|
|
{
|
|
"epoch": 1.9748559455212154,
|
|
"grad_norm": 0.3495903817194169,
|
|
"learning_rate": 3.60976416902949e-05,
|
|
"loss": 0.0939,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047893598675727844,
|
|
"step": 3770,
|
|
"valid_targets_mean": 3403.2,
|
|
"valid_targets_min": 1024
|
|
},
|
|
{
|
|
"epoch": 1.9774751178627552,
|
|
"grad_norm": 0.4355827963253253,
|
|
"learning_rate": 3.608212546554927e-05,
|
|
"loss": 0.1071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05321928486227989,
|
|
"step": 3775,
|
|
"valid_targets_mean": 2173.1,
|
|
"valid_targets_min": 786
|
|
},
|
|
{
|
|
"epoch": 1.9800942902042955,
|
|
"grad_norm": 0.43601912121827463,
|
|
"learning_rate": 3.6066581803543605e-05,
|
|
"loss": 0.1098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06638313829898834,
|
|
"step": 3780,
|
|
"valid_targets_mean": 3233.9,
|
|
"valid_targets_min": 2573
|
|
},
|
|
{
|
|
"epoch": 1.9827134625458354,
|
|
"grad_norm": 0.307649421102193,
|
|
"learning_rate": 3.605101073079651e-05,
|
|
"loss": 0.1085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056634437292814255,
|
|
"step": 3785,
|
|
"valid_targets_mean": 3821.9,
|
|
"valid_targets_min": 2274
|
|
},
|
|
{
|
|
"epoch": 1.9853326348873757,
|
|
"grad_norm": 0.3786213196256713,
|
|
"learning_rate": 3.6035412273873355e-05,
|
|
"loss": 0.0957,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0659828782081604,
|
|
"step": 3790,
|
|
"valid_targets_mean": 3791.2,
|
|
"valid_targets_min": 892
|
|
},
|
|
{
|
|
"epoch": 1.9879518072289155,
|
|
"grad_norm": 0.3037949300453433,
|
|
"learning_rate": 3.601978645938623e-05,
|
|
"loss": 0.1047,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042338401079177856,
|
|
"step": 3795,
|
|
"valid_targets_mean": 4820.4,
|
|
"valid_targets_min": 4491
|
|
},
|
|
{
|
|
"epoch": 1.9905709795704558,
|
|
"grad_norm": 0.516157150218108,
|
|
"learning_rate": 3.6004133313993874e-05,
|
|
"loss": 0.1482,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1160566657781601,
|
|
"step": 3800,
|
|
"valid_targets_mean": 2159.1,
|
|
"valid_targets_min": 709
|
|
},
|
|
{
|
|
"epoch": 1.9931901519119957,
|
|
"grad_norm": 0.30142097086660685,
|
|
"learning_rate": 3.5988452864401695e-05,
|
|
"loss": 0.1598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034394338726997375,
|
|
"step": 3805,
|
|
"valid_targets_mean": 3430.8,
|
|
"valid_targets_min": 1097
|
|
},
|
|
{
|
|
"epoch": 1.995809324253536,
|
|
"grad_norm": 0.33310269310201696,
|
|
"learning_rate": 3.597274513736167e-05,
|
|
"loss": 0.1094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04524791240692139,
|
|
"step": 3810,
|
|
"valid_targets_mean": 3032.0,
|
|
"valid_targets_min": 840
|
|
},
|
|
{
|
|
"epoch": 1.9984284965950758,
|
|
"grad_norm": 0.31566090994172014,
|
|
"learning_rate": 3.595701015967229e-05,
|
|
"loss": 0.1029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04328145459294319,
|
|
"step": 3815,
|
|
"valid_targets_mean": 3812.2,
|
|
"valid_targets_min": 3106
|
|
},
|
|
{
|
|
"epoch": 2.001047668936616,
|
|
"grad_norm": 0.8926428358074082,
|
|
"learning_rate": 3.5941247958178574e-05,
|
|
"loss": 0.1922,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1102590411901474,
|
|
"step": 3820,
|
|
"valid_targets_mean": 1653.2,
|
|
"valid_targets_min": 777
|
|
},
|
|
{
|
|
"epoch": 2.003666841278156,
|
|
"grad_norm": 0.8223433508863462,
|
|
"learning_rate": 3.5925458559771954e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10741463303565979,
|
|
"step": 3825,
|
|
"valid_targets_mean": 1349.8,
|
|
"valid_targets_min": 788
|
|
},
|
|
{
|
|
"epoch": 2.0062860136196963,
|
|
"grad_norm": 0.710236684697301,
|
|
"learning_rate": 3.590964199139029e-05,
|
|
"loss": 0.2042,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08701279759407043,
|
|
"step": 3830,
|
|
"valid_targets_mean": 1582.2,
|
|
"valid_targets_min": 914
|
|
},
|
|
{
|
|
"epoch": 2.008905185961236,
|
|
"grad_norm": 0.7585428118511915,
|
|
"learning_rate": 3.589379828001777e-05,
|
|
"loss": 0.202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09916609525680542,
|
|
"step": 3835,
|
|
"valid_targets_mean": 1468.5,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 2.0115243583027764,
|
|
"grad_norm": 0.6791949417976332,
|
|
"learning_rate": 3.587792745268492e-05,
|
|
"loss": 0.202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09496802091598511,
|
|
"step": 3840,
|
|
"valid_targets_mean": 1762.2,
|
|
"valid_targets_min": 1058
|
|
},
|
|
{
|
|
"epoch": 2.0141435306443163,
|
|
"grad_norm": 0.749855232106419,
|
|
"learning_rate": 3.586202953646848e-05,
|
|
"loss": 0.1994,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09410624951124191,
|
|
"step": 3845,
|
|
"valid_targets_mean": 1261.5,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 2.0167627029858566,
|
|
"grad_norm": 0.7015288073990621,
|
|
"learning_rate": 3.5846104558491465e-05,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10790140926837921,
|
|
"step": 3850,
|
|
"valid_targets_mean": 1791.1,
|
|
"valid_targets_min": 930
|
|
},
|
|
{
|
|
"epoch": 2.0193818753273964,
|
|
"grad_norm": 0.7924206807083869,
|
|
"learning_rate": 3.5830152545923024e-05,
|
|
"loss": 0.1861,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11440248787403107,
|
|
"step": 3855,
|
|
"valid_targets_mean": 1598.8,
|
|
"valid_targets_min": 737
|
|
},
|
|
{
|
|
"epoch": 2.0220010476689367,
|
|
"grad_norm": 0.7386126975367331,
|
|
"learning_rate": 3.581417352597843e-05,
|
|
"loss": 0.2111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08926434814929962,
|
|
"step": 3860,
|
|
"valid_targets_mean": 1449.5,
|
|
"valid_targets_min": 759
|
|
},
|
|
{
|
|
"epoch": 2.0246202200104766,
|
|
"grad_norm": 0.9632976534987181,
|
|
"learning_rate": 3.579816752591903e-05,
|
|
"loss": 0.1951,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09514357894659042,
|
|
"step": 3865,
|
|
"valid_targets_mean": 1652.8,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 2.027239392352017,
|
|
"grad_norm": 0.7834421185090436,
|
|
"learning_rate": 3.578213457305222e-05,
|
|
"loss": 0.1983,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12458643317222595,
|
|
"step": 3870,
|
|
"valid_targets_mean": 1675.9,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 2.0298585646935567,
|
|
"grad_norm": 0.748004283273961,
|
|
"learning_rate": 3.5766074694731374e-05,
|
|
"loss": 0.2036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07764267176389694,
|
|
"step": 3875,
|
|
"valid_targets_mean": 1187.5,
|
|
"valid_targets_min": 679
|
|
},
|
|
{
|
|
"epoch": 2.032477737035097,
|
|
"grad_norm": 0.7663537229597992,
|
|
"learning_rate": 3.574998791835578e-05,
|
|
"loss": 0.2001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08690992742776871,
|
|
"step": 3880,
|
|
"valid_targets_mean": 1200.9,
|
|
"valid_targets_min": 771
|
|
},
|
|
{
|
|
"epoch": 2.035096909376637,
|
|
"grad_norm": 0.6898331140298751,
|
|
"learning_rate": 3.573387427137064e-05,
|
|
"loss": 0.1842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08832302689552307,
|
|
"step": 3885,
|
|
"valid_targets_mean": 1593.0,
|
|
"valid_targets_min": 986
|
|
},
|
|
{
|
|
"epoch": 2.037716081718177,
|
|
"grad_norm": 0.708967320449303,
|
|
"learning_rate": 3.5717733781266994e-05,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10770932585000992,
|
|
"step": 3890,
|
|
"valid_targets_mean": 1591.4,
|
|
"valid_targets_min": 1103
|
|
},
|
|
{
|
|
"epoch": 2.040335254059717,
|
|
"grad_norm": 0.7089081171481766,
|
|
"learning_rate": 3.5701566475581675e-05,
|
|
"loss": 0.1954,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08699934929609299,
|
|
"step": 3895,
|
|
"valid_targets_mean": 1343.2,
|
|
"valid_targets_min": 675
|
|
},
|
|
{
|
|
"epoch": 2.0429544264012574,
|
|
"grad_norm": 0.6949969333301061,
|
|
"learning_rate": 3.568537238189727e-05,
|
|
"loss": 0.1966,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10397947579622269,
|
|
"step": 3900,
|
|
"valid_targets_mean": 1491.6,
|
|
"valid_targets_min": 941
|
|
},
|
|
{
|
|
"epoch": 2.045573598742797,
|
|
"grad_norm": 0.7473021370579906,
|
|
"learning_rate": 3.566915152784206e-05,
|
|
"loss": 0.2023,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1123787984251976,
|
|
"step": 3905,
|
|
"valid_targets_mean": 1921.4,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 2.0481927710843375,
|
|
"grad_norm": 0.7142355552447804,
|
|
"learning_rate": 3.565290394108998e-05,
|
|
"loss": 0.2001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08229781687259674,
|
|
"step": 3910,
|
|
"valid_targets_mean": 1375.2,
|
|
"valid_targets_min": 1023
|
|
},
|
|
{
|
|
"epoch": 2.0508119434258774,
|
|
"grad_norm": 0.75568883902217,
|
|
"learning_rate": 3.56366296493606e-05,
|
|
"loss": 0.2029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09430685639381409,
|
|
"step": 3915,
|
|
"valid_targets_mean": 1440.0,
|
|
"valid_targets_min": 660
|
|
},
|
|
{
|
|
"epoch": 2.0534311157674177,
|
|
"grad_norm": 0.6820924771511605,
|
|
"learning_rate": 3.562032868041901e-05,
|
|
"loss": 0.1923,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07597774267196655,
|
|
"step": 3920,
|
|
"valid_targets_mean": 1234.9,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 2.0560502881089575,
|
|
"grad_norm": 0.7222333855023938,
|
|
"learning_rate": 3.560400106207584e-05,
|
|
"loss": 0.2003,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07671467959880829,
|
|
"step": 3925,
|
|
"valid_targets_mean": 1198.9,
|
|
"valid_targets_min": 650
|
|
},
|
|
{
|
|
"epoch": 2.058669460450498,
|
|
"grad_norm": 0.7836793704441655,
|
|
"learning_rate": 3.558764682218717e-05,
|
|
"loss": 0.2016,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07579369843006134,
|
|
"step": 3930,
|
|
"valid_targets_mean": 1354.0,
|
|
"valid_targets_min": 863
|
|
},
|
|
{
|
|
"epoch": 2.0612886327920377,
|
|
"grad_norm": 0.7340836520402091,
|
|
"learning_rate": 3.557126598865452e-05,
|
|
"loss": 0.1849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10264989733695984,
|
|
"step": 3935,
|
|
"valid_targets_mean": 1517.6,
|
|
"valid_targets_min": 728
|
|
},
|
|
{
|
|
"epoch": 2.063907805133578,
|
|
"grad_norm": 0.7224732121013455,
|
|
"learning_rate": 3.5554858589424755e-05,
|
|
"loss": 0.1956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09368730336427689,
|
|
"step": 3940,
|
|
"valid_targets_mean": 1649.9,
|
|
"valid_targets_min": 686
|
|
},
|
|
{
|
|
"epoch": 2.066526977475118,
|
|
"grad_norm": 0.725587499393139,
|
|
"learning_rate": 3.553842465249009e-05,
|
|
"loss": 0.1893,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09984948486089706,
|
|
"step": 3945,
|
|
"valid_targets_mean": 1512.1,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 2.069146149816658,
|
|
"grad_norm": 0.6655391235263616,
|
|
"learning_rate": 3.552196420588799e-05,
|
|
"loss": 0.1889,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07939508557319641,
|
|
"step": 3950,
|
|
"valid_targets_mean": 1428.5,
|
|
"valid_targets_min": 779
|
|
},
|
|
{
|
|
"epoch": 2.071765322158198,
|
|
"grad_norm": 0.7260574699338164,
|
|
"learning_rate": 3.550547727770115e-05,
|
|
"loss": 0.2007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10762490332126617,
|
|
"step": 3955,
|
|
"valid_targets_mean": 1555.8,
|
|
"valid_targets_min": 1266
|
|
},
|
|
{
|
|
"epoch": 2.0743844944997383,
|
|
"grad_norm": 0.7434906694477653,
|
|
"learning_rate": 3.548896389605748e-05,
|
|
"loss": 0.1874,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0947028249502182,
|
|
"step": 3960,
|
|
"valid_targets_mean": 1495.4,
|
|
"valid_targets_min": 1033
|
|
},
|
|
{
|
|
"epoch": 2.077003666841278,
|
|
"grad_norm": 0.7109038144568908,
|
|
"learning_rate": 3.547242408912997e-05,
|
|
"loss": 0.1944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07691630721092224,
|
|
"step": 3965,
|
|
"valid_targets_mean": 1375.0,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 2.0796228391828184,
|
|
"grad_norm": 0.7311476978095995,
|
|
"learning_rate": 3.545585788513672e-05,
|
|
"loss": 0.1917,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09924264997243881,
|
|
"step": 3970,
|
|
"valid_targets_mean": 1515.8,
|
|
"valid_targets_min": 793
|
|
},
|
|
{
|
|
"epoch": 2.0822420115243583,
|
|
"grad_norm": 0.6821558270361922,
|
|
"learning_rate": 3.5439265312340874e-05,
|
|
"loss": 0.1949,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08497537672519684,
|
|
"step": 3975,
|
|
"valid_targets_mean": 1370.6,
|
|
"valid_targets_min": 806
|
|
},
|
|
{
|
|
"epoch": 2.0848611838658986,
|
|
"grad_norm": 0.6821998944070964,
|
|
"learning_rate": 3.542264639905054e-05,
|
|
"loss": 0.1933,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09639190137386322,
|
|
"step": 3980,
|
|
"valid_targets_mean": 1514.8,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 2.0874803562074384,
|
|
"grad_norm": 0.7289424882358152,
|
|
"learning_rate": 3.5406001173618785e-05,
|
|
"loss": 0.1918,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08458380401134491,
|
|
"step": 3985,
|
|
"valid_targets_mean": 1385.2,
|
|
"valid_targets_min": 800
|
|
},
|
|
{
|
|
"epoch": 2.0900995285489787,
|
|
"grad_norm": 0.675048563702559,
|
|
"learning_rate": 3.5389329664443546e-05,
|
|
"loss": 0.1771,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0678868219256401,
|
|
"step": 3990,
|
|
"valid_targets_mean": 1232.2,
|
|
"valid_targets_min": 837
|
|
},
|
|
{
|
|
"epoch": 2.0927187008905186,
|
|
"grad_norm": 0.678815702539696,
|
|
"learning_rate": 3.5372631899967626e-05,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08220085501670837,
|
|
"step": 3995,
|
|
"valid_targets_mean": 1317.6,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 2.095337873232059,
|
|
"grad_norm": 0.6740938563868182,
|
|
"learning_rate": 3.535590790867861e-05,
|
|
"loss": 0.1936,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07967671751976013,
|
|
"step": 4000,
|
|
"valid_targets_mean": 1312.8,
|
|
"valid_targets_min": 765
|
|
},
|
|
{
|
|
"epoch": 2.0979570455735987,
|
|
"grad_norm": 0.7141291054969721,
|
|
"learning_rate": 3.5339157719108825e-05,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09623882174491882,
|
|
"step": 4005,
|
|
"valid_targets_mean": 1511.5,
|
|
"valid_targets_min": 875
|
|
},
|
|
{
|
|
"epoch": 2.1005762179151386,
|
|
"grad_norm": 0.7336572112922154,
|
|
"learning_rate": 3.532238135983529e-05,
|
|
"loss": 0.1885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08519243448972702,
|
|
"step": 4010,
|
|
"valid_targets_mean": 1320.1,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 2.103195390256679,
|
|
"grad_norm": 0.7441682298736509,
|
|
"learning_rate": 3.530557885947968e-05,
|
|
"loss": 0.1963,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10714493691921234,
|
|
"step": 4015,
|
|
"valid_targets_mean": 1628.4,
|
|
"valid_targets_min": 669
|
|
},
|
|
{
|
|
"epoch": 2.1058145625982188,
|
|
"grad_norm": 0.7190417323186135,
|
|
"learning_rate": 3.528875024670827e-05,
|
|
"loss": 0.1828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09126546233892441,
|
|
"step": 4020,
|
|
"valid_targets_mean": 1412.2,
|
|
"valid_targets_min": 1046
|
|
},
|
|
{
|
|
"epoch": 2.108433734939759,
|
|
"grad_norm": 0.7170551586019916,
|
|
"learning_rate": 3.5271895550231874e-05,
|
|
"loss": 0.1982,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12611155211925507,
|
|
"step": 4025,
|
|
"valid_targets_mean": 2057.2,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 2.111052907281299,
|
|
"grad_norm": 0.6701651977269235,
|
|
"learning_rate": 3.525501479880583e-05,
|
|
"loss": 0.1953,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10826807469129562,
|
|
"step": 4030,
|
|
"valid_targets_mean": 1732.9,
|
|
"valid_targets_min": 907
|
|
},
|
|
{
|
|
"epoch": 2.113672079622839,
|
|
"grad_norm": 0.7644199373172864,
|
|
"learning_rate": 3.523810802122989e-05,
|
|
"loss": 0.2028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08702050149440765,
|
|
"step": 4035,
|
|
"valid_targets_mean": 1521.4,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 2.116291251964379,
|
|
"grad_norm": 0.6548591291141537,
|
|
"learning_rate": 3.522117524634824e-05,
|
|
"loss": 0.1863,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07724203914403915,
|
|
"step": 4040,
|
|
"valid_targets_mean": 1395.0,
|
|
"valid_targets_min": 822
|
|
},
|
|
{
|
|
"epoch": 2.1189104243059194,
|
|
"grad_norm": 0.6815393108692591,
|
|
"learning_rate": 3.520421650304942e-05,
|
|
"loss": 0.1858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10110010206699371,
|
|
"step": 4045,
|
|
"valid_targets_mean": 1654.0,
|
|
"valid_targets_min": 1291
|
|
},
|
|
{
|
|
"epoch": 2.121529596647459,
|
|
"grad_norm": 0.7724843355268597,
|
|
"learning_rate": 3.518723182026625e-05,
|
|
"loss": 0.1974,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10089796036481857,
|
|
"step": 4050,
|
|
"valid_targets_mean": 1573.2,
|
|
"valid_targets_min": 816
|
|
},
|
|
{
|
|
"epoch": 2.1241487689889995,
|
|
"grad_norm": 0.7093961681655364,
|
|
"learning_rate": 3.517022122697581e-05,
|
|
"loss": 0.1852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09187100827693939,
|
|
"step": 4055,
|
|
"valid_targets_mean": 1310.8,
|
|
"valid_targets_min": 721
|
|
},
|
|
{
|
|
"epoch": 2.1267679413305394,
|
|
"grad_norm": 0.6856877135989632,
|
|
"learning_rate": 3.5153184752199407e-05,
|
|
"loss": 0.1835,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08593648672103882,
|
|
"step": 4060,
|
|
"valid_targets_mean": 1413.8,
|
|
"valid_targets_min": 785
|
|
},
|
|
{
|
|
"epoch": 2.1293871136720797,
|
|
"grad_norm": 0.7008541080505509,
|
|
"learning_rate": 3.513612242500248e-05,
|
|
"loss": 0.1879,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.129766047000885,
|
|
"step": 4065,
|
|
"valid_targets_mean": 2005.6,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 2.1320062860136195,
|
|
"grad_norm": 0.6690436451371921,
|
|
"learning_rate": 3.511903427449459e-05,
|
|
"loss": 0.1922,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11221389472484589,
|
|
"step": 4070,
|
|
"valid_targets_mean": 1750.0,
|
|
"valid_targets_min": 744
|
|
},
|
|
{
|
|
"epoch": 2.13462545835516,
|
|
"grad_norm": 0.6992679651353025,
|
|
"learning_rate": 3.510192032982935e-05,
|
|
"loss": 0.1916,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11023158580064774,
|
|
"step": 4075,
|
|
"valid_targets_mean": 1856.8,
|
|
"valid_targets_min": 620
|
|
},
|
|
{
|
|
"epoch": 2.1372446306966997,
|
|
"grad_norm": 0.652670841490309,
|
|
"learning_rate": 3.508478062020437e-05,
|
|
"loss": 0.1839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1008521318435669,
|
|
"step": 4080,
|
|
"valid_targets_mean": 1614.4,
|
|
"valid_targets_min": 949
|
|
},
|
|
{
|
|
"epoch": 2.13986380303824,
|
|
"grad_norm": 0.6280896076662544,
|
|
"learning_rate": 3.506761517486123e-05,
|
|
"loss": 0.1915,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09381900727748871,
|
|
"step": 4085,
|
|
"valid_targets_mean": 1898.5,
|
|
"valid_targets_min": 1078
|
|
},
|
|
{
|
|
"epoch": 2.14248297537978,
|
|
"grad_norm": 0.6953352163842283,
|
|
"learning_rate": 3.505042402308541e-05,
|
|
"loss": 0.1854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09103520214557648,
|
|
"step": 4090,
|
|
"valid_targets_mean": 1538.6,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 2.14510214772132,
|
|
"grad_norm": 0.711371271604788,
|
|
"learning_rate": 3.5033207194206245e-05,
|
|
"loss": 0.1918,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10047143697738647,
|
|
"step": 4095,
|
|
"valid_targets_mean": 1591.0,
|
|
"valid_targets_min": 1122
|
|
},
|
|
{
|
|
"epoch": 2.14772132006286,
|
|
"grad_norm": 0.701579462853468,
|
|
"learning_rate": 3.501596471759689e-05,
|
|
"loss": 0.1866,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08872205764055252,
|
|
"step": 4100,
|
|
"valid_targets_mean": 1377.5,
|
|
"valid_targets_min": 674
|
|
},
|
|
{
|
|
"epoch": 2.1503404924044003,
|
|
"grad_norm": 0.6937603851188818,
|
|
"learning_rate": 3.499869662267423e-05,
|
|
"loss": 0.1803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07986070215702057,
|
|
"step": 4105,
|
|
"valid_targets_mean": 1198.6,
|
|
"valid_targets_min": 661
|
|
},
|
|
{
|
|
"epoch": 2.15295966474594,
|
|
"grad_norm": 0.7309748288880897,
|
|
"learning_rate": 3.498140293889889e-05,
|
|
"loss": 0.1782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11420120298862457,
|
|
"step": 4110,
|
|
"valid_targets_mean": 1806.1,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 2.1555788370874804,
|
|
"grad_norm": 0.6495847377174823,
|
|
"learning_rate": 3.4964083695775136e-05,
|
|
"loss": 0.1924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09754351526498795,
|
|
"step": 4115,
|
|
"valid_targets_mean": 1644.0,
|
|
"valid_targets_min": 1271
|
|
},
|
|
{
|
|
"epoch": 2.1581980094290203,
|
|
"grad_norm": 0.6451295156488667,
|
|
"learning_rate": 3.4946738922850825e-05,
|
|
"loss": 0.1832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08996035158634186,
|
|
"step": 4120,
|
|
"valid_targets_mean": 1391.1,
|
|
"valid_targets_min": 812
|
|
},
|
|
{
|
|
"epoch": 2.1608171817705606,
|
|
"grad_norm": 0.6577673028743707,
|
|
"learning_rate": 3.49293686497174e-05,
|
|
"loss": 0.1879,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09317929297685623,
|
|
"step": 4125,
|
|
"valid_targets_mean": 1529.6,
|
|
"valid_targets_min": 954
|
|
},
|
|
{
|
|
"epoch": 2.1634363541121004,
|
|
"grad_norm": 0.6870101085446985,
|
|
"learning_rate": 3.491197290600978e-05,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08796785771846771,
|
|
"step": 4130,
|
|
"valid_targets_mean": 1537.9,
|
|
"valid_targets_min": 1297
|
|
},
|
|
{
|
|
"epoch": 2.1660555264536407,
|
|
"grad_norm": 0.7033994409983768,
|
|
"learning_rate": 3.489455172140637e-05,
|
|
"loss": 0.1898,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0888407975435257,
|
|
"step": 4135,
|
|
"valid_targets_mean": 1411.9,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 2.1686746987951806,
|
|
"grad_norm": 0.68174779726752,
|
|
"learning_rate": 3.487710512562894e-05,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10044382512569427,
|
|
"step": 4140,
|
|
"valid_targets_mean": 1824.5,
|
|
"valid_targets_min": 592
|
|
},
|
|
{
|
|
"epoch": 2.171293871136721,
|
|
"grad_norm": 0.7058446912301533,
|
|
"learning_rate": 3.4859633148442644e-05,
|
|
"loss": 0.1881,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09607305377721786,
|
|
"step": 4145,
|
|
"valid_targets_mean": 1587.0,
|
|
"valid_targets_min": 662
|
|
},
|
|
{
|
|
"epoch": 2.1739130434782608,
|
|
"grad_norm": 0.7109324969687294,
|
|
"learning_rate": 3.4842135819655926e-05,
|
|
"loss": 0.1919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08893938362598419,
|
|
"step": 4150,
|
|
"valid_targets_mean": 1366.5,
|
|
"valid_targets_min": 622
|
|
},
|
|
{
|
|
"epoch": 2.176532215819801,
|
|
"grad_norm": 0.7091820730051697,
|
|
"learning_rate": 3.482461316912049e-05,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07064250111579895,
|
|
"step": 4155,
|
|
"valid_targets_mean": 1229.2,
|
|
"valid_targets_min": 625
|
|
},
|
|
{
|
|
"epoch": 2.179151388161341,
|
|
"grad_norm": 0.7304958338240904,
|
|
"learning_rate": 3.480706522673124e-05,
|
|
"loss": 0.1986,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09226875007152557,
|
|
"step": 4160,
|
|
"valid_targets_mean": 1548.0,
|
|
"valid_targets_min": 888
|
|
},
|
|
{
|
|
"epoch": 2.181770560502881,
|
|
"grad_norm": 0.8269923067529306,
|
|
"learning_rate": 3.47894920224262e-05,
|
|
"loss": 0.189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07171830534934998,
|
|
"step": 4165,
|
|
"valid_targets_mean": 1331.6,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 2.184389732844421,
|
|
"grad_norm": 0.7316829722718358,
|
|
"learning_rate": 3.477189358618655e-05,
|
|
"loss": 0.1957,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13178831338882446,
|
|
"step": 4170,
|
|
"valid_targets_mean": 1781.0,
|
|
"valid_targets_min": 522
|
|
},
|
|
{
|
|
"epoch": 2.1870089051859614,
|
|
"grad_norm": 0.6989346135672048,
|
|
"learning_rate": 3.4754269948036465e-05,
|
|
"loss": 0.1854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08098998665809631,
|
|
"step": 4175,
|
|
"valid_targets_mean": 1283.8,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 2.189628077527501,
|
|
"grad_norm": 0.6609140467070396,
|
|
"learning_rate": 3.473662113804314e-05,
|
|
"loss": 0.1806,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07369507849216461,
|
|
"step": 4180,
|
|
"valid_targets_mean": 1332.5,
|
|
"valid_targets_min": 869
|
|
},
|
|
{
|
|
"epoch": 2.1922472498690415,
|
|
"grad_norm": 0.7339790791732105,
|
|
"learning_rate": 3.4718947186316714e-05,
|
|
"loss": 0.1883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08820786327123642,
|
|
"step": 4185,
|
|
"valid_targets_mean": 1341.1,
|
|
"valid_targets_min": 769
|
|
},
|
|
{
|
|
"epoch": 2.1948664222105814,
|
|
"grad_norm": 0.6865017670917262,
|
|
"learning_rate": 3.4701248123010214e-05,
|
|
"loss": 0.189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08384440839290619,
|
|
"step": 4190,
|
|
"valid_targets_mean": 1329.4,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 2.1974855945521217,
|
|
"grad_norm": 0.6817303509529443,
|
|
"learning_rate": 3.4683523978319514e-05,
|
|
"loss": 0.1815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0841609388589859,
|
|
"step": 4195,
|
|
"valid_targets_mean": 1297.8,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 2.2001047668936615,
|
|
"grad_norm": 0.6522827159862312,
|
|
"learning_rate": 3.4665774782483275e-05,
|
|
"loss": 0.1805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11358039081096649,
|
|
"step": 4200,
|
|
"valid_targets_mean": 1903.4,
|
|
"valid_targets_min": 1323
|
|
},
|
|
{
|
|
"epoch": 2.202723939235202,
|
|
"grad_norm": 0.7825502975856193,
|
|
"learning_rate": 3.464800056578289e-05,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10468216240406036,
|
|
"step": 4205,
|
|
"valid_targets_mean": 1449.0,
|
|
"valid_targets_min": 851
|
|
},
|
|
{
|
|
"epoch": 2.2053431115767417,
|
|
"grad_norm": 0.6706125180804291,
|
|
"learning_rate": 3.4630201358542476e-05,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09529706090688705,
|
|
"step": 4210,
|
|
"valid_targets_mean": 1540.2,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 2.207962283918282,
|
|
"grad_norm": 0.6309444219938192,
|
|
"learning_rate": 3.461237719112872e-05,
|
|
"loss": 0.1788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09462803602218628,
|
|
"step": 4215,
|
|
"valid_targets_mean": 1497.5,
|
|
"valid_targets_min": 854
|
|
},
|
|
{
|
|
"epoch": 2.210581456259822,
|
|
"grad_norm": 0.5323564402683407,
|
|
"learning_rate": 3.459452809395097e-05,
|
|
"loss": 0.1783,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06672701984643936,
|
|
"step": 4220,
|
|
"valid_targets_mean": 1495.4,
|
|
"valid_targets_min": 917
|
|
},
|
|
{
|
|
"epoch": 2.213200628601362,
|
|
"grad_norm": 0.612533975177077,
|
|
"learning_rate": 3.457665409746103e-05,
|
|
"loss": 0.1688,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07375086843967438,
|
|
"step": 4225,
|
|
"valid_targets_mean": 1462.5,
|
|
"valid_targets_min": 930
|
|
},
|
|
{
|
|
"epoch": 2.215819800942902,
|
|
"grad_norm": 0.6502973513547888,
|
|
"learning_rate": 3.455875523215325e-05,
|
|
"loss": 0.1864,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08019331097602844,
|
|
"step": 4230,
|
|
"valid_targets_mean": 1464.4,
|
|
"valid_targets_min": 570
|
|
},
|
|
{
|
|
"epoch": 2.2184389732844423,
|
|
"grad_norm": 0.6362908081945943,
|
|
"learning_rate": 3.454083152856435e-05,
|
|
"loss": 0.1828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09245746582746506,
|
|
"step": 4235,
|
|
"valid_targets_mean": 1536.1,
|
|
"valid_targets_min": 870
|
|
},
|
|
{
|
|
"epoch": 2.221058145625982,
|
|
"grad_norm": 0.6704801346208137,
|
|
"learning_rate": 3.452288301727348e-05,
|
|
"loss": 0.1849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09293714165687561,
|
|
"step": 4240,
|
|
"valid_targets_mean": 1563.5,
|
|
"valid_targets_min": 1215
|
|
},
|
|
{
|
|
"epoch": 2.2236773179675224,
|
|
"grad_norm": 0.6738442410852162,
|
|
"learning_rate": 3.450490972890208e-05,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0857582688331604,
|
|
"step": 4245,
|
|
"valid_targets_mean": 1545.9,
|
|
"valid_targets_min": 1214
|
|
},
|
|
{
|
|
"epoch": 2.2262964903090623,
|
|
"grad_norm": 0.7121593190212068,
|
|
"learning_rate": 3.448691169411386e-05,
|
|
"loss": 0.1798,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11348661035299301,
|
|
"step": 4250,
|
|
"valid_targets_mean": 1675.2,
|
|
"valid_targets_min": 591
|
|
},
|
|
{
|
|
"epoch": 2.2289156626506026,
|
|
"grad_norm": 0.7297488641520847,
|
|
"learning_rate": 3.446888894361477e-05,
|
|
"loss": 0.1933,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09707330167293549,
|
|
"step": 4255,
|
|
"valid_targets_mean": 1329.6,
|
|
"valid_targets_min": 646
|
|
},
|
|
{
|
|
"epoch": 2.2315348349921424,
|
|
"grad_norm": 0.4715234550471819,
|
|
"learning_rate": 3.445084150815292e-05,
|
|
"loss": 0.1984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08680760860443115,
|
|
"step": 4260,
|
|
"valid_targets_mean": 5262.4,
|
|
"valid_targets_min": 4751
|
|
},
|
|
{
|
|
"epoch": 2.2341540073336827,
|
|
"grad_norm": 0.5179432162410191,
|
|
"learning_rate": 3.443276941851851e-05,
|
|
"loss": 0.1965,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10530431568622589,
|
|
"step": 4265,
|
|
"valid_targets_mean": 8557.5,
|
|
"valid_targets_min": 5164
|
|
},
|
|
{
|
|
"epoch": 2.2367731796752226,
|
|
"grad_norm": 0.35189088859712925,
|
|
"learning_rate": 3.4414672705543836e-05,
|
|
"loss": 0.1871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08838881552219391,
|
|
"step": 4270,
|
|
"valid_targets_mean": 6448.5,
|
|
"valid_targets_min": 4342
|
|
},
|
|
{
|
|
"epoch": 2.239392352016763,
|
|
"grad_norm": 0.3967730785170395,
|
|
"learning_rate": 3.43965514001032e-05,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08274708688259125,
|
|
"step": 4275,
|
|
"valid_targets_mean": 5300.6,
|
|
"valid_targets_min": 3853
|
|
},
|
|
{
|
|
"epoch": 2.2420115243583028,
|
|
"grad_norm": 0.3914698796402864,
|
|
"learning_rate": 3.437840553311282e-05,
|
|
"loss": 0.1944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1191055178642273,
|
|
"step": 4280,
|
|
"valid_targets_mean": 5632.0,
|
|
"valid_targets_min": 4669
|
|
},
|
|
{
|
|
"epoch": 2.244630696699843,
|
|
"grad_norm": 0.3826804658437182,
|
|
"learning_rate": 3.436023513553087e-05,
|
|
"loss": 0.1985,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10407496988773346,
|
|
"step": 4285,
|
|
"valid_targets_mean": 6666.6,
|
|
"valid_targets_min": 5144
|
|
},
|
|
{
|
|
"epoch": 2.247249869041383,
|
|
"grad_norm": 0.40704622259157347,
|
|
"learning_rate": 3.434204023835735e-05,
|
|
"loss": 0.194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10071707516908646,
|
|
"step": 4290,
|
|
"valid_targets_mean": 5754.6,
|
|
"valid_targets_min": 4350
|
|
},
|
|
{
|
|
"epoch": 2.249869041382923,
|
|
"grad_norm": 0.35141069529027763,
|
|
"learning_rate": 3.432382087263405e-05,
|
|
"loss": 0.1839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09793560206890106,
|
|
"step": 4295,
|
|
"valid_targets_mean": 6102.2,
|
|
"valid_targets_min": 3934
|
|
},
|
|
{
|
|
"epoch": 2.252488213724463,
|
|
"grad_norm": 0.41169499298812734,
|
|
"learning_rate": 3.430557706944452e-05,
|
|
"loss": 0.2143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08575282990932465,
|
|
"step": 4300,
|
|
"valid_targets_mean": 6471.4,
|
|
"valid_targets_min": 5231
|
|
},
|
|
{
|
|
"epoch": 2.255107386066003,
|
|
"grad_norm": 0.40526461109259,
|
|
"learning_rate": 3.4287308859914e-05,
|
|
"loss": 0.1874,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10314113646745682,
|
|
"step": 4305,
|
|
"valid_targets_mean": 6736.8,
|
|
"valid_targets_min": 5062
|
|
},
|
|
{
|
|
"epoch": 2.257726558407543,
|
|
"grad_norm": 0.6096771286476559,
|
|
"learning_rate": 3.426901627520937e-05,
|
|
"loss": 0.1948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0975448489189148,
|
|
"step": 4310,
|
|
"valid_targets_mean": 3416.1,
|
|
"valid_targets_min": 229
|
|
},
|
|
{
|
|
"epoch": 2.2603457307490835,
|
|
"grad_norm": 0.36027677446290235,
|
|
"learning_rate": 3.425069934653907e-05,
|
|
"loss": 0.1839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09143798798322678,
|
|
"step": 4315,
|
|
"valid_targets_mean": 6196.9,
|
|
"valid_targets_min": 4508
|
|
},
|
|
{
|
|
"epoch": 2.2629649030906234,
|
|
"grad_norm": 0.39418300333836775,
|
|
"learning_rate": 3.423235810515313e-05,
|
|
"loss": 0.1935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09388123452663422,
|
|
"step": 4320,
|
|
"valid_targets_mean": 5994.6,
|
|
"valid_targets_min": 4673
|
|
},
|
|
{
|
|
"epoch": 2.2655840754321632,
|
|
"grad_norm": 0.32210755336423735,
|
|
"learning_rate": 3.4213992582343e-05,
|
|
"loss": 0.1754,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0751730352640152,
|
|
"step": 4325,
|
|
"valid_targets_mean": 6293.1,
|
|
"valid_targets_min": 4995
|
|
},
|
|
{
|
|
"epoch": 2.2682032477737035,
|
|
"grad_norm": 0.44027247201770786,
|
|
"learning_rate": 3.419560280944158e-05,
|
|
"loss": 0.1745,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10627452284097672,
|
|
"step": 4330,
|
|
"valid_targets_mean": 5654.2,
|
|
"valid_targets_min": 4108
|
|
},
|
|
{
|
|
"epoch": 2.270822420115244,
|
|
"grad_norm": 0.3914400928933456,
|
|
"learning_rate": 3.417718881782315e-05,
|
|
"loss": 0.1831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12119898200035095,
|
|
"step": 4335,
|
|
"valid_targets_mean": 7076.8,
|
|
"valid_targets_min": 4527
|
|
},
|
|
{
|
|
"epoch": 2.2734415924567837,
|
|
"grad_norm": 0.5025289124302289,
|
|
"learning_rate": 3.41587506389033e-05,
|
|
"loss": 0.1982,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11772619932889938,
|
|
"step": 4340,
|
|
"valid_targets_mean": 2271.2,
|
|
"valid_targets_min": 1783
|
|
},
|
|
{
|
|
"epoch": 2.2760607647983235,
|
|
"grad_norm": 0.37806515377016836,
|
|
"learning_rate": 3.4140288304138885e-05,
|
|
"loss": 0.1659,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0790642648935318,
|
|
"step": 4345,
|
|
"valid_targets_mean": 6375.6,
|
|
"valid_targets_min": 5316
|
|
},
|
|
{
|
|
"epoch": 2.278679937139864,
|
|
"grad_norm": 0.3847792712836001,
|
|
"learning_rate": 3.4121801845027965e-05,
|
|
"loss": 0.1722,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09972573816776276,
|
|
"step": 4350,
|
|
"valid_targets_mean": 6093.6,
|
|
"valid_targets_min": 5056
|
|
},
|
|
{
|
|
"epoch": 2.2812991094814037,
|
|
"grad_norm": 0.35680978232665095,
|
|
"learning_rate": 3.410329129310977e-05,
|
|
"loss": 0.1717,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09506428241729736,
|
|
"step": 4355,
|
|
"valid_targets_mean": 7712.4,
|
|
"valid_targets_min": 2834
|
|
},
|
|
{
|
|
"epoch": 2.283918281822944,
|
|
"grad_norm": 0.37489894585228845,
|
|
"learning_rate": 3.4084756679964645e-05,
|
|
"loss": 0.2041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11063200235366821,
|
|
"step": 4360,
|
|
"valid_targets_mean": 7032.0,
|
|
"valid_targets_min": 4181
|
|
},
|
|
{
|
|
"epoch": 2.286537454164484,
|
|
"grad_norm": 0.34295699089468396,
|
|
"learning_rate": 3.4066198037213956e-05,
|
|
"loss": 0.1834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08322663605213165,
|
|
"step": 4365,
|
|
"valid_targets_mean": 7101.1,
|
|
"valid_targets_min": 4689
|
|
},
|
|
{
|
|
"epoch": 2.289156626506024,
|
|
"grad_norm": 0.3670372159304986,
|
|
"learning_rate": 3.4047615396520073e-05,
|
|
"loss": 0.1805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08686234056949615,
|
|
"step": 4370,
|
|
"valid_targets_mean": 6668.0,
|
|
"valid_targets_min": 4670
|
|
},
|
|
{
|
|
"epoch": 2.291775798847564,
|
|
"grad_norm": 0.37134005811384047,
|
|
"learning_rate": 3.402900878958633e-05,
|
|
"loss": 0.1917,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11375340819358826,
|
|
"step": 4375,
|
|
"valid_targets_mean": 7794.4,
|
|
"valid_targets_min": 4501
|
|
},
|
|
{
|
|
"epoch": 2.2943949711891043,
|
|
"grad_norm": 0.4295427738825388,
|
|
"learning_rate": 3.4010378248156925e-05,
|
|
"loss": 0.2027,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09965578466653824,
|
|
"step": 4380,
|
|
"valid_targets_mean": 6428.1,
|
|
"valid_targets_min": 4998
|
|
},
|
|
{
|
|
"epoch": 2.297014143530644,
|
|
"grad_norm": 0.389803008593126,
|
|
"learning_rate": 3.39917238040169e-05,
|
|
"loss": 0.2194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09798961132764816,
|
|
"step": 4385,
|
|
"valid_targets_mean": 6159.2,
|
|
"valid_targets_min": 4640
|
|
},
|
|
{
|
|
"epoch": 2.2996333158721844,
|
|
"grad_norm": 0.3775684215634937,
|
|
"learning_rate": 3.397304548899207e-05,
|
|
"loss": 0.2003,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0962294191122055,
|
|
"step": 4390,
|
|
"valid_targets_mean": 6138.4,
|
|
"valid_targets_min": 4800
|
|
},
|
|
{
|
|
"epoch": 2.3022524882137243,
|
|
"grad_norm": 0.4164029673006619,
|
|
"learning_rate": 3.395434333494899e-05,
|
|
"loss": 0.1918,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11183840781450272,
|
|
"step": 4395,
|
|
"valid_targets_mean": 6040.1,
|
|
"valid_targets_min": 4455
|
|
},
|
|
{
|
|
"epoch": 2.3048716605552646,
|
|
"grad_norm": 0.38789717205487884,
|
|
"learning_rate": 3.3935617373794875e-05,
|
|
"loss": 0.185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09181490540504456,
|
|
"step": 4400,
|
|
"valid_targets_mean": 6714.5,
|
|
"valid_targets_min": 4891
|
|
},
|
|
{
|
|
"epoch": 2.3074908328968045,
|
|
"grad_norm": 0.40947172820355565,
|
|
"learning_rate": 3.391686763747755e-05,
|
|
"loss": 0.1972,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1077517420053482,
|
|
"step": 4405,
|
|
"valid_targets_mean": 5823.5,
|
|
"valid_targets_min": 4710
|
|
},
|
|
{
|
|
"epoch": 2.3101100052383448,
|
|
"grad_norm": 0.3550831915380223,
|
|
"learning_rate": 3.3898094157985424e-05,
|
|
"loss": 0.1871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08461962640285492,
|
|
"step": 4410,
|
|
"valid_targets_mean": 6401.5,
|
|
"valid_targets_min": 5370
|
|
},
|
|
{
|
|
"epoch": 2.3127291775798846,
|
|
"grad_norm": 0.39518422640270573,
|
|
"learning_rate": 3.3879296967347386e-05,
|
|
"loss": 0.1934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09296444058418274,
|
|
"step": 4415,
|
|
"valid_targets_mean": 6424.5,
|
|
"valid_targets_min": 4608
|
|
},
|
|
{
|
|
"epoch": 2.315348349921425,
|
|
"grad_norm": 0.38336885075106303,
|
|
"learning_rate": 3.38604760976328e-05,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08820196986198425,
|
|
"step": 4420,
|
|
"valid_targets_mean": 5586.1,
|
|
"valid_targets_min": 4580
|
|
},
|
|
{
|
|
"epoch": 2.3179675222629648,
|
|
"grad_norm": 0.539587906084315,
|
|
"learning_rate": 3.3841631580951414e-05,
|
|
"loss": 0.1914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09103749692440033,
|
|
"step": 4425,
|
|
"valid_targets_mean": 6329.4,
|
|
"valid_targets_min": 4754
|
|
},
|
|
{
|
|
"epoch": 2.320586694604505,
|
|
"grad_norm": 0.3132796125370284,
|
|
"learning_rate": 3.382276344945334e-05,
|
|
"loss": 0.1655,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07120954990386963,
|
|
"step": 4430,
|
|
"valid_targets_mean": 7311.5,
|
|
"valid_targets_min": 6287
|
|
},
|
|
{
|
|
"epoch": 2.323205866946045,
|
|
"grad_norm": 0.3786549569109157,
|
|
"learning_rate": 3.380387173532895e-05,
|
|
"loss": 0.179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10338225960731506,
|
|
"step": 4435,
|
|
"valid_targets_mean": 7010.4,
|
|
"valid_targets_min": 3468
|
|
},
|
|
{
|
|
"epoch": 2.325825039287585,
|
|
"grad_norm": 0.354358681080531,
|
|
"learning_rate": 3.378495647080886e-05,
|
|
"loss": 0.1925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09100731462240219,
|
|
"step": 4440,
|
|
"valid_targets_mean": 6255.4,
|
|
"valid_targets_min": 5221
|
|
},
|
|
{
|
|
"epoch": 2.328444211629125,
|
|
"grad_norm": 0.36484120389471236,
|
|
"learning_rate": 3.376601768816388e-05,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08618143945932388,
|
|
"step": 4445,
|
|
"valid_targets_mean": 6060.5,
|
|
"valid_targets_min": 4438
|
|
},
|
|
{
|
|
"epoch": 2.3310633839706654,
|
|
"grad_norm": 0.38168634070322494,
|
|
"learning_rate": 3.374705541970494e-05,
|
|
"loss": 0.1855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09062792360782623,
|
|
"step": 4450,
|
|
"valid_targets_mean": 5617.6,
|
|
"valid_targets_min": 3915
|
|
},
|
|
{
|
|
"epoch": 2.3336825563122052,
|
|
"grad_norm": 0.3182791292068853,
|
|
"learning_rate": 3.372806969778302e-05,
|
|
"loss": 0.1773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06596499681472778,
|
|
"step": 4455,
|
|
"valid_targets_mean": 5274.1,
|
|
"valid_targets_min": 4135
|
|
},
|
|
{
|
|
"epoch": 2.3363017286537455,
|
|
"grad_norm": 0.3980520306984013,
|
|
"learning_rate": 3.3709060554789114e-05,
|
|
"loss": 0.1854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11008140444755554,
|
|
"step": 4460,
|
|
"valid_targets_mean": 6510.1,
|
|
"valid_targets_min": 4295
|
|
},
|
|
{
|
|
"epoch": 2.3389209009952854,
|
|
"grad_norm": 0.399551997504261,
|
|
"learning_rate": 3.36900280231542e-05,
|
|
"loss": 0.1856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.110128253698349,
|
|
"step": 4465,
|
|
"valid_targets_mean": 6065.0,
|
|
"valid_targets_min": 4865
|
|
},
|
|
{
|
|
"epoch": 2.3415400733368257,
|
|
"grad_norm": 0.38392949159985834,
|
|
"learning_rate": 3.3670972135349136e-05,
|
|
"loss": 0.1854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09090718626976013,
|
|
"step": 4470,
|
|
"valid_targets_mean": 6150.2,
|
|
"valid_targets_min": 4643
|
|
},
|
|
{
|
|
"epoch": 2.3441592456783655,
|
|
"grad_norm": 0.35208214019483036,
|
|
"learning_rate": 3.365189292388463e-05,
|
|
"loss": 0.1763,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09005999565124512,
|
|
"step": 4475,
|
|
"valid_targets_mean": 6384.5,
|
|
"valid_targets_min": 4705
|
|
},
|
|
{
|
|
"epoch": 2.346778418019906,
|
|
"grad_norm": 0.3846867579860675,
|
|
"learning_rate": 3.36327904213112e-05,
|
|
"loss": 0.2026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09906090795993805,
|
|
"step": 4480,
|
|
"valid_targets_mean": 5637.6,
|
|
"valid_targets_min": 4652
|
|
},
|
|
{
|
|
"epoch": 2.3493975903614457,
|
|
"grad_norm": 0.36982382045858164,
|
|
"learning_rate": 3.361366466021907e-05,
|
|
"loss": 0.1652,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07524531334638596,
|
|
"step": 4485,
|
|
"valid_targets_mean": 5898.2,
|
|
"valid_targets_min": 5007
|
|
},
|
|
{
|
|
"epoch": 2.352016762702986,
|
|
"grad_norm": 0.3586455082897619,
|
|
"learning_rate": 3.3594515673238164e-05,
|
|
"loss": 0.1613,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08170197904109955,
|
|
"step": 4490,
|
|
"valid_targets_mean": 5927.9,
|
|
"valid_targets_min": 4546
|
|
},
|
|
{
|
|
"epoch": 2.354635935044526,
|
|
"grad_norm": 0.3627252830677402,
|
|
"learning_rate": 3.357534349303804e-05,
|
|
"loss": 0.1652,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08540515601634979,
|
|
"step": 4495,
|
|
"valid_targets_mean": 5954.4,
|
|
"valid_targets_min": 3200
|
|
},
|
|
{
|
|
"epoch": 2.357255107386066,
|
|
"grad_norm": 0.35767876753365935,
|
|
"learning_rate": 3.3556148152327796e-05,
|
|
"loss": 0.1716,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08642897009849548,
|
|
"step": 4500,
|
|
"valid_targets_mean": 6702.5,
|
|
"valid_targets_min": 5010
|
|
},
|
|
{
|
|
"epoch": 2.359874279727606,
|
|
"grad_norm": 0.38355130876545307,
|
|
"learning_rate": 3.3536929683856075e-05,
|
|
"loss": 0.193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0961657464504242,
|
|
"step": 4505,
|
|
"valid_targets_mean": 6308.1,
|
|
"valid_targets_min": 4904
|
|
},
|
|
{
|
|
"epoch": 2.3624934520691463,
|
|
"grad_norm": 0.39846948733145526,
|
|
"learning_rate": 3.351768812041095e-05,
|
|
"loss": 0.2206,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11881693452596664,
|
|
"step": 4510,
|
|
"valid_targets_mean": 6365.5,
|
|
"valid_targets_min": 5081
|
|
},
|
|
{
|
|
"epoch": 2.365112624410686,
|
|
"grad_norm": 0.4071323760198838,
|
|
"learning_rate": 3.349842349481992e-05,
|
|
"loss": 0.1888,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07891501486301422,
|
|
"step": 4515,
|
|
"valid_targets_mean": 4943.8,
|
|
"valid_targets_min": 3629
|
|
},
|
|
{
|
|
"epoch": 2.3677317967522264,
|
|
"grad_norm": 0.3910265333978852,
|
|
"learning_rate": 3.347913583994982e-05,
|
|
"loss": 0.1807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09046255052089691,
|
|
"step": 4520,
|
|
"valid_targets_mean": 6236.2,
|
|
"valid_targets_min": 4324
|
|
},
|
|
{
|
|
"epoch": 2.3703509690937663,
|
|
"grad_norm": 0.35925362948621203,
|
|
"learning_rate": 3.345982518870676e-05,
|
|
"loss": 0.1842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07996508479118347,
|
|
"step": 4525,
|
|
"valid_targets_mean": 7204.9,
|
|
"valid_targets_min": 4920
|
|
},
|
|
{
|
|
"epoch": 2.3729701414353066,
|
|
"grad_norm": 0.3790145575135124,
|
|
"learning_rate": 3.344049157403611e-05,
|
|
"loss": 0.1932,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0945851132273674,
|
|
"step": 4530,
|
|
"valid_targets_mean": 6281.5,
|
|
"valid_targets_min": 4882
|
|
},
|
|
{
|
|
"epoch": 2.3755893137768465,
|
|
"grad_norm": 0.3562095630949342,
|
|
"learning_rate": 3.342113502892239e-05,
|
|
"loss": 0.1976,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09100031852722168,
|
|
"step": 4535,
|
|
"valid_targets_mean": 7296.2,
|
|
"valid_targets_min": 3943
|
|
},
|
|
{
|
|
"epoch": 2.3782084861183868,
|
|
"grad_norm": 0.3889102556751455,
|
|
"learning_rate": 3.340175558638927e-05,
|
|
"loss": 0.1876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10129101574420929,
|
|
"step": 4540,
|
|
"valid_targets_mean": 7207.0,
|
|
"valid_targets_min": 5382
|
|
},
|
|
{
|
|
"epoch": 2.3808276584599266,
|
|
"grad_norm": 0.3616664728118041,
|
|
"learning_rate": 3.338235327949946e-05,
|
|
"loss": 0.1862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09182766079902649,
|
|
"step": 4545,
|
|
"valid_targets_mean": 6385.6,
|
|
"valid_targets_min": 4461
|
|
},
|
|
{
|
|
"epoch": 2.383446830801467,
|
|
"grad_norm": 0.35250339413631243,
|
|
"learning_rate": 3.3362928141354697e-05,
|
|
"loss": 0.1762,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08577057719230652,
|
|
"step": 4550,
|
|
"valid_targets_mean": 7383.1,
|
|
"valid_targets_min": 5364
|
|
},
|
|
{
|
|
"epoch": 2.3860660031430068,
|
|
"grad_norm": 0.3613497836153022,
|
|
"learning_rate": 3.334348020509565e-05,
|
|
"loss": 0.1753,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0901482105255127,
|
|
"step": 4555,
|
|
"valid_targets_mean": 5899.0,
|
|
"valid_targets_min": 4444
|
|
},
|
|
{
|
|
"epoch": 2.388685175484547,
|
|
"grad_norm": 0.3732680956771438,
|
|
"learning_rate": 3.332400950390189e-05,
|
|
"loss": 0.2313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09998985379934311,
|
|
"step": 4560,
|
|
"valid_targets_mean": 6466.9,
|
|
"valid_targets_min": 4474
|
|
},
|
|
{
|
|
"epoch": 2.391304347826087,
|
|
"grad_norm": 0.3215897801694627,
|
|
"learning_rate": 3.330451607099185e-05,
|
|
"loss": 0.1763,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09537540376186371,
|
|
"step": 4565,
|
|
"valid_targets_mean": 7983.6,
|
|
"valid_targets_min": 6076
|
|
},
|
|
{
|
|
"epoch": 2.393923520167627,
|
|
"grad_norm": 0.359957824578163,
|
|
"learning_rate": 3.328499993962271e-05,
|
|
"loss": 0.1772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09370891749858856,
|
|
"step": 4570,
|
|
"valid_targets_mean": 6965.8,
|
|
"valid_targets_min": 4718
|
|
},
|
|
{
|
|
"epoch": 2.396542692509167,
|
|
"grad_norm": 0.361204202925108,
|
|
"learning_rate": 3.326546114309039e-05,
|
|
"loss": 0.1899,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09906766563653946,
|
|
"step": 4575,
|
|
"valid_targets_mean": 6808.2,
|
|
"valid_targets_min": 4714
|
|
},
|
|
{
|
|
"epoch": 2.3991618648507074,
|
|
"grad_norm": 0.3826464800023626,
|
|
"learning_rate": 3.324589971472949e-05,
|
|
"loss": 0.1861,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08802899718284607,
|
|
"step": 4580,
|
|
"valid_targets_mean": 5286.5,
|
|
"valid_targets_min": 3828
|
|
},
|
|
{
|
|
"epoch": 2.4017810371922472,
|
|
"grad_norm": 0.35754151744307855,
|
|
"learning_rate": 3.32263156879132e-05,
|
|
"loss": 0.1622,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07199066132307053,
|
|
"step": 4585,
|
|
"valid_targets_mean": 6731.1,
|
|
"valid_targets_min": 5089
|
|
},
|
|
{
|
|
"epoch": 2.4044002095337875,
|
|
"grad_norm": 0.3830574411711552,
|
|
"learning_rate": 3.320670909605328e-05,
|
|
"loss": 0.1822,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08338411152362823,
|
|
"step": 4590,
|
|
"valid_targets_mean": 6523.9,
|
|
"valid_targets_min": 4509
|
|
},
|
|
{
|
|
"epoch": 2.4070193818753274,
|
|
"grad_norm": 0.3766898876673147,
|
|
"learning_rate": 3.318707997259998e-05,
|
|
"loss": 0.1688,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10137352347373962,
|
|
"step": 4595,
|
|
"valid_targets_mean": 7736.0,
|
|
"valid_targets_min": 5193
|
|
},
|
|
{
|
|
"epoch": 2.4096385542168672,
|
|
"grad_norm": 0.35915128351202785,
|
|
"learning_rate": 3.3167428351042e-05,
|
|
"loss": 0.1706,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0861617922782898,
|
|
"step": 4600,
|
|
"valid_targets_mean": 6165.9,
|
|
"valid_targets_min": 4343
|
|
},
|
|
{
|
|
"epoch": 2.4122577265584075,
|
|
"grad_norm": 0.36197540903740943,
|
|
"learning_rate": 3.31477542649064e-05,
|
|
"loss": 0.1758,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0850772112607956,
|
|
"step": 4605,
|
|
"valid_targets_mean": 6185.9,
|
|
"valid_targets_min": 4815
|
|
},
|
|
{
|
|
"epoch": 2.414876898899948,
|
|
"grad_norm": 0.33820687033900637,
|
|
"learning_rate": 3.31280577477586e-05,
|
|
"loss": 0.1796,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07993307709693909,
|
|
"step": 4610,
|
|
"valid_targets_mean": 7682.5,
|
|
"valid_targets_min": 5751
|
|
},
|
|
{
|
|
"epoch": 2.4174960712414877,
|
|
"grad_norm": 0.34784080140637824,
|
|
"learning_rate": 3.310833883320227e-05,
|
|
"loss": 0.1711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08610358089208603,
|
|
"step": 4615,
|
|
"valid_targets_mean": 6375.6,
|
|
"valid_targets_min": 3977
|
|
},
|
|
{
|
|
"epoch": 2.4201152435830275,
|
|
"grad_norm": 0.39648090166834754,
|
|
"learning_rate": 3.3088597554879275e-05,
|
|
"loss": 0.1823,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09887340664863586,
|
|
"step": 4620,
|
|
"valid_targets_mean": 6607.8,
|
|
"valid_targets_min": 5098
|
|
},
|
|
{
|
|
"epoch": 2.422734415924568,
|
|
"grad_norm": 0.44055110866968583,
|
|
"learning_rate": 3.3068833946469677e-05,
|
|
"loss": 0.1876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10944779962301254,
|
|
"step": 4625,
|
|
"valid_targets_mean": 5749.6,
|
|
"valid_targets_min": 4550
|
|
},
|
|
{
|
|
"epoch": 2.425353588266108,
|
|
"grad_norm": 0.4129639305919711,
|
|
"learning_rate": 3.30490480416916e-05,
|
|
"loss": 0.189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09889828413724899,
|
|
"step": 4630,
|
|
"valid_targets_mean": 5552.6,
|
|
"valid_targets_min": 3619
|
|
},
|
|
{
|
|
"epoch": 2.427972760607648,
|
|
"grad_norm": 0.6992836949420002,
|
|
"learning_rate": 3.302923987430121e-05,
|
|
"loss": 0.2792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1501704752445221,
|
|
"step": 4635,
|
|
"valid_targets_mean": 3486.5,
|
|
"valid_targets_min": 770
|
|
},
|
|
{
|
|
"epoch": 2.430591932949188,
|
|
"grad_norm": 0.49681113723633175,
|
|
"learning_rate": 3.3009409478092664e-05,
|
|
"loss": 0.2761,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12776651978492737,
|
|
"step": 4640,
|
|
"valid_targets_mean": 4703.8,
|
|
"valid_targets_min": 1799
|
|
},
|
|
{
|
|
"epoch": 2.433211105290728,
|
|
"grad_norm": 0.483795517262897,
|
|
"learning_rate": 3.298955688689805e-05,
|
|
"loss": 0.2688,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13452821969985962,
|
|
"step": 4645,
|
|
"valid_targets_mean": 5730.6,
|
|
"valid_targets_min": 3015
|
|
},
|
|
{
|
|
"epoch": 2.4358302776322684,
|
|
"grad_norm": 0.4451395934905341,
|
|
"learning_rate": 3.296968213458731e-05,
|
|
"loss": 0.2657,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12698446214199066,
|
|
"step": 4650,
|
|
"valid_targets_mean": 5475.6,
|
|
"valid_targets_min": 1841
|
|
},
|
|
{
|
|
"epoch": 2.4384494499738083,
|
|
"grad_norm": 0.568085783596915,
|
|
"learning_rate": 3.294978525506819e-05,
|
|
"loss": 0.2793,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16987842321395874,
|
|
"step": 4655,
|
|
"valid_targets_mean": 4238.2,
|
|
"valid_targets_min": 1374
|
|
},
|
|
{
|
|
"epoch": 2.441068622315348,
|
|
"grad_norm": 0.5399690888229732,
|
|
"learning_rate": 3.29298662822862e-05,
|
|
"loss": 0.275,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14218473434448242,
|
|
"step": 4660,
|
|
"valid_targets_mean": 4297.9,
|
|
"valid_targets_min": 1951
|
|
},
|
|
{
|
|
"epoch": 2.4436877946568885,
|
|
"grad_norm": 0.4665755676585302,
|
|
"learning_rate": 3.290992525022455e-05,
|
|
"loss": 0.2651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10295194387435913,
|
|
"step": 4665,
|
|
"valid_targets_mean": 3986.4,
|
|
"valid_targets_min": 967
|
|
},
|
|
{
|
|
"epoch": 2.4463069669984283,
|
|
"grad_norm": 0.4810606992303234,
|
|
"learning_rate": 3.288996219290405e-05,
|
|
"loss": 0.272,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14516176283359528,
|
|
"step": 4670,
|
|
"valid_targets_mean": 5266.1,
|
|
"valid_targets_min": 1831
|
|
},
|
|
{
|
|
"epoch": 2.4489261393399686,
|
|
"grad_norm": 0.5548954720725366,
|
|
"learning_rate": 3.2869977144383114e-05,
|
|
"loss": 0.2805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13955682516098022,
|
|
"step": 4675,
|
|
"valid_targets_mean": 3626.9,
|
|
"valid_targets_min": 2077
|
|
},
|
|
{
|
|
"epoch": 2.4515453116815085,
|
|
"grad_norm": 0.4881601822284099,
|
|
"learning_rate": 3.2849970138757675e-05,
|
|
"loss": 0.2732,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12774524092674255,
|
|
"step": 4680,
|
|
"valid_targets_mean": 4431.5,
|
|
"valid_targets_min": 2834
|
|
},
|
|
{
|
|
"epoch": 2.4541644840230488,
|
|
"grad_norm": 0.5286262928112249,
|
|
"learning_rate": 3.2829941210161106e-05,
|
|
"loss": 0.2765,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14698627591133118,
|
|
"step": 4685,
|
|
"valid_targets_mean": 4406.6,
|
|
"valid_targets_min": 1873
|
|
},
|
|
{
|
|
"epoch": 2.4567836563645886,
|
|
"grad_norm": 0.4622668282561899,
|
|
"learning_rate": 3.2809890392764216e-05,
|
|
"loss": 0.2653,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12009242177009583,
|
|
"step": 4690,
|
|
"valid_targets_mean": 3873.5,
|
|
"valid_targets_min": 1013
|
|
},
|
|
{
|
|
"epoch": 2.459402828706129,
|
|
"grad_norm": 0.5705828101329861,
|
|
"learning_rate": 3.2789817720775126e-05,
|
|
"loss": 0.2747,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14647673070430756,
|
|
"step": 4695,
|
|
"valid_targets_mean": 4464.8,
|
|
"valid_targets_min": 1939
|
|
},
|
|
{
|
|
"epoch": 2.4620220010476688,
|
|
"grad_norm": 0.58339715082754,
|
|
"learning_rate": 3.2769723228439265e-05,
|
|
"loss": 0.2698,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16818226873874664,
|
|
"step": 4700,
|
|
"valid_targets_mean": 4397.6,
|
|
"valid_targets_min": 2391
|
|
},
|
|
{
|
|
"epoch": 2.464641173389209,
|
|
"grad_norm": 0.6181542558710681,
|
|
"learning_rate": 3.274960695003927e-05,
|
|
"loss": 0.2812,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18955881893634796,
|
|
"step": 4705,
|
|
"valid_targets_mean": 4258.1,
|
|
"valid_targets_min": 1589
|
|
},
|
|
{
|
|
"epoch": 2.467260345730749,
|
|
"grad_norm": 0.521980103704629,
|
|
"learning_rate": 3.272946891989495e-05,
|
|
"loss": 0.2645,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12609046697616577,
|
|
"step": 4710,
|
|
"valid_targets_mean": 4111.2,
|
|
"valid_targets_min": 1840
|
|
},
|
|
{
|
|
"epoch": 2.4698795180722892,
|
|
"grad_norm": 0.5491561012683551,
|
|
"learning_rate": 3.2709309172363236e-05,
|
|
"loss": 0.2735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13917270302772522,
|
|
"step": 4715,
|
|
"valid_targets_mean": 3903.1,
|
|
"valid_targets_min": 1269
|
|
},
|
|
{
|
|
"epoch": 2.472498690413829,
|
|
"grad_norm": 0.5919069763103967,
|
|
"learning_rate": 3.268912774183811e-05,
|
|
"loss": 0.2854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13042256236076355,
|
|
"step": 4720,
|
|
"valid_targets_mean": 3429.4,
|
|
"valid_targets_min": 1695
|
|
},
|
|
{
|
|
"epoch": 2.4751178627553694,
|
|
"grad_norm": 0.5622471486979514,
|
|
"learning_rate": 3.2668924662750535e-05,
|
|
"loss": 0.2826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16786539554595947,
|
|
"step": 4725,
|
|
"valid_targets_mean": 4396.5,
|
|
"valid_targets_min": 1542
|
|
},
|
|
{
|
|
"epoch": 2.4777370350969092,
|
|
"grad_norm": 0.599374258150151,
|
|
"learning_rate": 3.2648699969568415e-05,
|
|
"loss": 0.2677,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1258111596107483,
|
|
"step": 4730,
|
|
"valid_targets_mean": 3763.1,
|
|
"valid_targets_min": 1438
|
|
},
|
|
{
|
|
"epoch": 2.4803562074384495,
|
|
"grad_norm": 0.5574771684252116,
|
|
"learning_rate": 3.262845369679652e-05,
|
|
"loss": 0.2798,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14597754180431366,
|
|
"step": 4735,
|
|
"valid_targets_mean": 4322.0,
|
|
"valid_targets_min": 2080
|
|
},
|
|
{
|
|
"epoch": 2.4829753797799894,
|
|
"grad_norm": 0.5335917055886615,
|
|
"learning_rate": 3.260818587897646e-05,
|
|
"loss": 0.2668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11487531661987305,
|
|
"step": 4740,
|
|
"valid_targets_mean": 3656.5,
|
|
"valid_targets_min": 1026
|
|
},
|
|
{
|
|
"epoch": 2.4855945521215297,
|
|
"grad_norm": 0.6647589643530221,
|
|
"learning_rate": 3.258789655068657e-05,
|
|
"loss": 0.2549,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14061737060546875,
|
|
"step": 4745,
|
|
"valid_targets_mean": 3275.9,
|
|
"valid_targets_min": 1772
|
|
},
|
|
{
|
|
"epoch": 2.4882137244630695,
|
|
"grad_norm": 0.5828423175607111,
|
|
"learning_rate": 3.256758574654191e-05,
|
|
"loss": 0.2731,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14968352019786835,
|
|
"step": 4750,
|
|
"valid_targets_mean": 3340.6,
|
|
"valid_targets_min": 1588
|
|
},
|
|
{
|
|
"epoch": 2.49083289680461,
|
|
"grad_norm": 0.6068833100306307,
|
|
"learning_rate": 3.254725350119416e-05,
|
|
"loss": 0.2738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12587854266166687,
|
|
"step": 4755,
|
|
"valid_targets_mean": 3175.5,
|
|
"valid_targets_min": 1276
|
|
},
|
|
{
|
|
"epoch": 2.4934520691461497,
|
|
"grad_norm": 0.4952352072659673,
|
|
"learning_rate": 3.2526899849331604e-05,
|
|
"loss": 0.2669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11407821625471115,
|
|
"step": 4760,
|
|
"valid_targets_mean": 4294.9,
|
|
"valid_targets_min": 1631
|
|
},
|
|
{
|
|
"epoch": 2.49607124148769,
|
|
"grad_norm": 0.5488663596585718,
|
|
"learning_rate": 3.250652482567902e-05,
|
|
"loss": 0.2759,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16853857040405273,
|
|
"step": 4765,
|
|
"valid_targets_mean": 5289.8,
|
|
"valid_targets_min": 3090
|
|
},
|
|
{
|
|
"epoch": 2.49869041382923,
|
|
"grad_norm": 0.6661768035799297,
|
|
"learning_rate": 3.248612846499767e-05,
|
|
"loss": 0.2595,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13370314240455627,
|
|
"step": 4770,
|
|
"valid_targets_mean": 3234.0,
|
|
"valid_targets_min": 1446
|
|
},
|
|
{
|
|
"epoch": 2.50130958617077,
|
|
"grad_norm": 0.6239150762558955,
|
|
"learning_rate": 3.24657108020852e-05,
|
|
"loss": 0.2749,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13287469744682312,
|
|
"step": 4775,
|
|
"valid_targets_mean": 3117.9,
|
|
"valid_targets_min": 1685
|
|
},
|
|
{
|
|
"epoch": 2.50392875851231,
|
|
"grad_norm": 0.5235567064263872,
|
|
"learning_rate": 3.244527187177561e-05,
|
|
"loss": 0.2598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12638260424137115,
|
|
"step": 4780,
|
|
"valid_targets_mean": 3541.6,
|
|
"valid_targets_min": 1596
|
|
},
|
|
{
|
|
"epoch": 2.5065479308538503,
|
|
"grad_norm": 0.6633883278924252,
|
|
"learning_rate": 3.2424811708939186e-05,
|
|
"loss": 0.2637,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14612677693367004,
|
|
"step": 4785,
|
|
"valid_targets_mean": 5101.5,
|
|
"valid_targets_min": 2251
|
|
},
|
|
{
|
|
"epoch": 2.50916710319539,
|
|
"grad_norm": 0.650243258812906,
|
|
"learning_rate": 3.240433034848243e-05,
|
|
"loss": 0.2641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1239681988954544,
|
|
"step": 4790,
|
|
"valid_targets_mean": 2717.9,
|
|
"valid_targets_min": 1458
|
|
},
|
|
{
|
|
"epoch": 2.5117862755369305,
|
|
"grad_norm": 0.6162560405529081,
|
|
"learning_rate": 3.238382782534802e-05,
|
|
"loss": 0.2714,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11557632684707642,
|
|
"step": 4795,
|
|
"valid_targets_mean": 3202.1,
|
|
"valid_targets_min": 1162
|
|
},
|
|
{
|
|
"epoch": 2.5144054478784703,
|
|
"grad_norm": 0.6153507368134533,
|
|
"learning_rate": 3.236330417451472e-05,
|
|
"loss": 0.2622,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1267622411251068,
|
|
"step": 4800,
|
|
"valid_targets_mean": 3012.8,
|
|
"valid_targets_min": 1480
|
|
},
|
|
{
|
|
"epoch": 2.5170246202200106,
|
|
"grad_norm": 0.5956133163742262,
|
|
"learning_rate": 3.234275943099736e-05,
|
|
"loss": 0.2721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14925609529018402,
|
|
"step": 4805,
|
|
"valid_targets_mean": 3682.4,
|
|
"valid_targets_min": 1229
|
|
},
|
|
{
|
|
"epoch": 2.5196437925615505,
|
|
"grad_norm": 0.6351685164908751,
|
|
"learning_rate": 3.232219362984674e-05,
|
|
"loss": 0.2691,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14006267488002777,
|
|
"step": 4810,
|
|
"valid_targets_mean": 3763.1,
|
|
"valid_targets_min": 1276
|
|
},
|
|
{
|
|
"epoch": 2.5222629649030908,
|
|
"grad_norm": 0.6162058804285464,
|
|
"learning_rate": 3.23016068061496e-05,
|
|
"loss": 0.2617,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11890268325805664,
|
|
"step": 4815,
|
|
"valid_targets_mean": 3172.8,
|
|
"valid_targets_min": 1208
|
|
},
|
|
{
|
|
"epoch": 2.5248821372446306,
|
|
"grad_norm": 0.5749928458656193,
|
|
"learning_rate": 3.228099899502853e-05,
|
|
"loss": 0.2603,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12680307030677795,
|
|
"step": 4820,
|
|
"valid_targets_mean": 3354.0,
|
|
"valid_targets_min": 924
|
|
},
|
|
{
|
|
"epoch": 2.527501309586171,
|
|
"grad_norm": 0.5724131019946627,
|
|
"learning_rate": 3.226037023164194e-05,
|
|
"loss": 0.267,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13974511623382568,
|
|
"step": 4825,
|
|
"valid_targets_mean": 4218.9,
|
|
"valid_targets_min": 1531
|
|
},
|
|
{
|
|
"epoch": 2.5301204819277108,
|
|
"grad_norm": 0.5910025466904364,
|
|
"learning_rate": 3.223972055118398e-05,
|
|
"loss": 0.2672,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13048329949378967,
|
|
"step": 4830,
|
|
"valid_targets_mean": 3040.9,
|
|
"valid_targets_min": 1643
|
|
},
|
|
{
|
|
"epoch": 2.532739654269251,
|
|
"grad_norm": 0.567667815004245,
|
|
"learning_rate": 3.221904998888448e-05,
|
|
"loss": 0.2596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15320153534412384,
|
|
"step": 4835,
|
|
"valid_targets_mean": 4129.6,
|
|
"valid_targets_min": 1244
|
|
},
|
|
{
|
|
"epoch": 2.535358826610791,
|
|
"grad_norm": 0.5616875504141101,
|
|
"learning_rate": 3.219835858000892e-05,
|
|
"loss": 0.2719,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11181098222732544,
|
|
"step": 4840,
|
|
"valid_targets_mean": 3491.1,
|
|
"valid_targets_min": 730
|
|
},
|
|
{
|
|
"epoch": 2.5379779989523312,
|
|
"grad_norm": 0.6548795772657205,
|
|
"learning_rate": 3.217764635985831e-05,
|
|
"loss": 0.2657,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11994566023349762,
|
|
"step": 4845,
|
|
"valid_targets_mean": 3085.4,
|
|
"valid_targets_min": 977
|
|
},
|
|
{
|
|
"epoch": 2.540597171293871,
|
|
"grad_norm": 0.5326832519042694,
|
|
"learning_rate": 3.215691336376919e-05,
|
|
"loss": 0.2576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13010746240615845,
|
|
"step": 4850,
|
|
"valid_targets_mean": 4896.9,
|
|
"valid_targets_min": 2091
|
|
},
|
|
{
|
|
"epoch": 2.543216343635411,
|
|
"grad_norm": 0.5801183408858036,
|
|
"learning_rate": 3.213615962711354e-05,
|
|
"loss": 0.2525,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1055474579334259,
|
|
"step": 4855,
|
|
"valid_targets_mean": 3125.9,
|
|
"valid_targets_min": 1654
|
|
},
|
|
{
|
|
"epoch": 2.5458355159769512,
|
|
"grad_norm": 0.5219430315952756,
|
|
"learning_rate": 3.211538518529872e-05,
|
|
"loss": 0.2731,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12499532848596573,
|
|
"step": 4860,
|
|
"valid_targets_mean": 4228.1,
|
|
"valid_targets_min": 1432
|
|
},
|
|
{
|
|
"epoch": 2.5484546883184915,
|
|
"grad_norm": 0.5512349112170226,
|
|
"learning_rate": 3.209459007376744e-05,
|
|
"loss": 0.2596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1550905406475067,
|
|
"step": 4865,
|
|
"valid_targets_mean": 4512.6,
|
|
"valid_targets_min": 2422
|
|
},
|
|
{
|
|
"epoch": 2.5510738606600314,
|
|
"grad_norm": 0.6283528677051325,
|
|
"learning_rate": 3.207377432799762e-05,
|
|
"loss": 0.2697,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1240365281701088,
|
|
"step": 4870,
|
|
"valid_targets_mean": 3584.2,
|
|
"valid_targets_min": 1124
|
|
},
|
|
{
|
|
"epoch": 2.5536930330015712,
|
|
"grad_norm": 0.5565275234233498,
|
|
"learning_rate": 3.205293798350243e-05,
|
|
"loss": 0.2647,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13041697442531586,
|
|
"step": 4875,
|
|
"valid_targets_mean": 3950.6,
|
|
"valid_targets_min": 1219
|
|
},
|
|
{
|
|
"epoch": 2.5563122053431115,
|
|
"grad_norm": 0.5639979123752346,
|
|
"learning_rate": 3.203208107583017e-05,
|
|
"loss": 0.2721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14429600536823273,
|
|
"step": 4880,
|
|
"valid_targets_mean": 3843.0,
|
|
"valid_targets_min": 1328
|
|
},
|
|
{
|
|
"epoch": 2.558931377684652,
|
|
"grad_norm": 0.6039266296646495,
|
|
"learning_rate": 3.201120364056422e-05,
|
|
"loss": 0.2547,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11220163851976395,
|
|
"step": 4885,
|
|
"valid_targets_mean": 2625.0,
|
|
"valid_targets_min": 916
|
|
},
|
|
{
|
|
"epoch": 2.5615505500261917,
|
|
"grad_norm": 0.5575623583731243,
|
|
"learning_rate": 3.199030571332299e-05,
|
|
"loss": 0.2734,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08969323337078094,
|
|
"step": 4890,
|
|
"valid_targets_mean": 3081.5,
|
|
"valid_targets_min": 1302
|
|
},
|
|
{
|
|
"epoch": 2.5641697223677316,
|
|
"grad_norm": 0.5863420722239778,
|
|
"learning_rate": 3.1969387329759823e-05,
|
|
"loss": 0.2594,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.172468364238739,
|
|
"step": 4895,
|
|
"valid_targets_mean": 4390.5,
|
|
"valid_targets_min": 1500
|
|
},
|
|
{
|
|
"epoch": 2.566788894709272,
|
|
"grad_norm": 0.5419491414286708,
|
|
"learning_rate": 3.194844852556301e-05,
|
|
"loss": 0.249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14079037308692932,
|
|
"step": 4900,
|
|
"valid_targets_mean": 4516.0,
|
|
"valid_targets_min": 1909
|
|
},
|
|
{
|
|
"epoch": 2.569408067050812,
|
|
"grad_norm": 0.5742549479956544,
|
|
"learning_rate": 3.1927489336455634e-05,
|
|
"loss": 0.2446,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11546041071414948,
|
|
"step": 4905,
|
|
"valid_targets_mean": 3179.0,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 2.572027239392352,
|
|
"grad_norm": 0.6280354658180206,
|
|
"learning_rate": 3.190650979819558e-05,
|
|
"loss": 0.2755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13638462126255035,
|
|
"step": 4910,
|
|
"valid_targets_mean": 3833.9,
|
|
"valid_targets_min": 2662
|
|
},
|
|
{
|
|
"epoch": 2.574646411733892,
|
|
"grad_norm": 0.518374228157193,
|
|
"learning_rate": 3.188550994657544e-05,
|
|
"loss": 0.2596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1271933913230896,
|
|
"step": 4915,
|
|
"valid_targets_mean": 4343.4,
|
|
"valid_targets_min": 1460
|
|
},
|
|
{
|
|
"epoch": 2.577265584075432,
|
|
"grad_norm": 0.5180685671704083,
|
|
"learning_rate": 3.186448981742249e-05,
|
|
"loss": 0.263,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14760158956050873,
|
|
"step": 4920,
|
|
"valid_targets_mean": 5864.5,
|
|
"valid_targets_min": 3945
|
|
},
|
|
{
|
|
"epoch": 2.5798847564169725,
|
|
"grad_norm": 0.574567284867831,
|
|
"learning_rate": 3.184344944659855e-05,
|
|
"loss": 0.2623,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.156520813703537,
|
|
"step": 4925,
|
|
"valid_targets_mean": 4052.5,
|
|
"valid_targets_min": 2504
|
|
},
|
|
{
|
|
"epoch": 2.5825039287585123,
|
|
"grad_norm": 0.5578714284179954,
|
|
"learning_rate": 3.182238887000002e-05,
|
|
"loss": 0.2681,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1554621011018753,
|
|
"step": 4930,
|
|
"valid_targets_mean": 5593.9,
|
|
"valid_targets_min": 3596
|
|
},
|
|
{
|
|
"epoch": 2.585123101100052,
|
|
"grad_norm": 0.6046078395960688,
|
|
"learning_rate": 3.180130812355775e-05,
|
|
"loss": 0.2702,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12882524728775024,
|
|
"step": 4935,
|
|
"valid_targets_mean": 3224.1,
|
|
"valid_targets_min": 1352
|
|
},
|
|
{
|
|
"epoch": 2.5877422734415925,
|
|
"grad_norm": 0.5122085684412675,
|
|
"learning_rate": 3.1780207243237e-05,
|
|
"loss": 0.2553,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12836720049381256,
|
|
"step": 4940,
|
|
"valid_targets_mean": 5189.8,
|
|
"valid_targets_min": 1722
|
|
},
|
|
{
|
|
"epoch": 2.5903614457831328,
|
|
"grad_norm": 0.5459458711195225,
|
|
"learning_rate": 3.17590862650374e-05,
|
|
"loss": 0.2531,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1324058324098587,
|
|
"step": 4945,
|
|
"valid_targets_mean": 4270.8,
|
|
"valid_targets_min": 2090
|
|
},
|
|
{
|
|
"epoch": 2.5929806181246726,
|
|
"grad_norm": 0.6004641655982031,
|
|
"learning_rate": 3.173794522499284e-05,
|
|
"loss": 0.2529,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13428449630737305,
|
|
"step": 4950,
|
|
"valid_targets_mean": 3596.2,
|
|
"valid_targets_min": 1731
|
|
},
|
|
{
|
|
"epoch": 2.5955997904662125,
|
|
"grad_norm": 0.6172992051651541,
|
|
"learning_rate": 3.171678415917146e-05,
|
|
"loss": 0.2545,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13604748249053955,
|
|
"step": 4955,
|
|
"valid_targets_mean": 3311.6,
|
|
"valid_targets_min": 1460
|
|
},
|
|
{
|
|
"epoch": 2.5982189628077528,
|
|
"grad_norm": 0.572741597250024,
|
|
"learning_rate": 3.169560310367556e-05,
|
|
"loss": 0.2431,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1529918611049652,
|
|
"step": 4960,
|
|
"valid_targets_mean": 4707.2,
|
|
"valid_targets_min": 2357
|
|
},
|
|
{
|
|
"epoch": 2.600838135149293,
|
|
"grad_norm": 0.5831300386644992,
|
|
"learning_rate": 3.1674402094641534e-05,
|
|
"loss": 0.2444,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12589526176452637,
|
|
"step": 4965,
|
|
"valid_targets_mean": 3636.2,
|
|
"valid_targets_min": 1511
|
|
},
|
|
{
|
|
"epoch": 2.603457307490833,
|
|
"grad_norm": 0.5927490633887448,
|
|
"learning_rate": 3.1653181168239824e-05,
|
|
"loss": 0.2478,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14207634329795837,
|
|
"step": 4970,
|
|
"valid_targets_mean": 4583.0,
|
|
"valid_targets_min": 1764
|
|
},
|
|
{
|
|
"epoch": 2.606076479832373,
|
|
"grad_norm": 0.6305783971358412,
|
|
"learning_rate": 3.163194036067485e-05,
|
|
"loss": 0.2689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14542587101459503,
|
|
"step": 4975,
|
|
"valid_targets_mean": 3437.5,
|
|
"valid_targets_min": 1175
|
|
},
|
|
{
|
|
"epoch": 2.608695652173913,
|
|
"grad_norm": 0.4970617621017961,
|
|
"learning_rate": 3.161067970818496e-05,
|
|
"loss": 0.2616,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09945046156644821,
|
|
"step": 4980,
|
|
"valid_targets_mean": 3982.1,
|
|
"valid_targets_min": 2546
|
|
},
|
|
{
|
|
"epoch": 2.6113148245154534,
|
|
"grad_norm": 0.5949517503873936,
|
|
"learning_rate": 3.158939924704236e-05,
|
|
"loss": 0.2625,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.146941676735878,
|
|
"step": 4985,
|
|
"valid_targets_mean": 4124.0,
|
|
"valid_targets_min": 1244
|
|
},
|
|
{
|
|
"epoch": 2.6139339968569932,
|
|
"grad_norm": 0.525319801845006,
|
|
"learning_rate": 3.156809901355303e-05,
|
|
"loss": 0.2543,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13219237327575684,
|
|
"step": 4990,
|
|
"valid_targets_mean": 5047.1,
|
|
"valid_targets_min": 1856
|
|
},
|
|
{
|
|
"epoch": 2.616553169198533,
|
|
"grad_norm": 0.5590760328022601,
|
|
"learning_rate": 3.154677904405671e-05,
|
|
"loss": 0.2533,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14278973639011383,
|
|
"step": 4995,
|
|
"valid_targets_mean": 3927.4,
|
|
"valid_targets_min": 1948
|
|
},
|
|
{
|
|
"epoch": 2.6191723415400734,
|
|
"grad_norm": 0.5511814734502197,
|
|
"learning_rate": 3.152543937492678e-05,
|
|
"loss": 0.2481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12088270485401154,
|
|
"step": 5000,
|
|
"valid_targets_mean": 3516.5,
|
|
"valid_targets_min": 1248
|
|
},
|
|
{
|
|
"epoch": 2.6217915138816132,
|
|
"grad_norm": 0.5009950800692684,
|
|
"learning_rate": 3.150408004257026e-05,
|
|
"loss": 0.2514,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12171660363674164,
|
|
"step": 5005,
|
|
"valid_targets_mean": 4611.0,
|
|
"valid_targets_min": 1145
|
|
},
|
|
{
|
|
"epoch": 2.6244106862231535,
|
|
"grad_norm": 0.6747271785946586,
|
|
"learning_rate": 3.1482701083427696e-05,
|
|
"loss": 0.2541,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1250152587890625,
|
|
"step": 5010,
|
|
"valid_targets_mean": 3262.6,
|
|
"valid_targets_min": 2058
|
|
},
|
|
{
|
|
"epoch": 2.6270298585646934,
|
|
"grad_norm": 0.5540119181488308,
|
|
"learning_rate": 3.1461302533973136e-05,
|
|
"loss": 0.2537,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1475522816181183,
|
|
"step": 5015,
|
|
"valid_targets_mean": 4546.5,
|
|
"valid_targets_min": 1708
|
|
},
|
|
{
|
|
"epoch": 2.6296490309062337,
|
|
"grad_norm": 0.581373530694527,
|
|
"learning_rate": 3.143988443071403e-05,
|
|
"loss": 0.2536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11121198534965515,
|
|
"step": 5020,
|
|
"valid_targets_mean": 3555.8,
|
|
"valid_targets_min": 1301
|
|
},
|
|
{
|
|
"epoch": 2.6322682032477736,
|
|
"grad_norm": 0.5712205243857897,
|
|
"learning_rate": 3.141844681019121e-05,
|
|
"loss": 0.2699,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09923257678747177,
|
|
"step": 5025,
|
|
"valid_targets_mean": 3068.1,
|
|
"valid_targets_min": 1971
|
|
},
|
|
{
|
|
"epoch": 2.634887375589314,
|
|
"grad_norm": 0.5980222178924863,
|
|
"learning_rate": 3.139698970897879e-05,
|
|
"loss": 0.2651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15571749210357666,
|
|
"step": 5030,
|
|
"valid_targets_mean": 5172.4,
|
|
"valid_targets_min": 1054
|
|
},
|
|
{
|
|
"epoch": 2.6375065479308537,
|
|
"grad_norm": 0.609268242830208,
|
|
"learning_rate": 3.1375513163684116e-05,
|
|
"loss": 0.2603,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14475512504577637,
|
|
"step": 5035,
|
|
"valid_targets_mean": 3143.4,
|
|
"valid_targets_min": 1612
|
|
},
|
|
{
|
|
"epoch": 2.640125720272394,
|
|
"grad_norm": 0.573022933412219,
|
|
"learning_rate": 3.135401721094773e-05,
|
|
"loss": 0.2704,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12014967203140259,
|
|
"step": 5040,
|
|
"valid_targets_mean": 3499.6,
|
|
"valid_targets_min": 1274
|
|
},
|
|
{
|
|
"epoch": 2.642744892613934,
|
|
"grad_norm": 0.6055266043189158,
|
|
"learning_rate": 3.1332501887443264e-05,
|
|
"loss": 0.2611,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13370947539806366,
|
|
"step": 5045,
|
|
"valid_targets_mean": 3628.4,
|
|
"valid_targets_min": 1625
|
|
},
|
|
{
|
|
"epoch": 2.645364064955474,
|
|
"grad_norm": 0.5638328015687015,
|
|
"learning_rate": 3.13109672298774e-05,
|
|
"loss": 0.2685,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14932434260845184,
|
|
"step": 5050,
|
|
"valid_targets_mean": 4642.5,
|
|
"valid_targets_min": 1381
|
|
},
|
|
{
|
|
"epoch": 2.647983237297014,
|
|
"grad_norm": 0.6375225173023469,
|
|
"learning_rate": 3.128941327498982e-05,
|
|
"loss": 0.2739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15590864419937134,
|
|
"step": 5055,
|
|
"valid_targets_mean": 3540.1,
|
|
"valid_targets_min": 1591
|
|
},
|
|
{
|
|
"epoch": 2.6506024096385543,
|
|
"grad_norm": 0.5701764304397182,
|
|
"learning_rate": 3.12678400595531e-05,
|
|
"loss": 0.2544,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11090701818466187,
|
|
"step": 5060,
|
|
"valid_targets_mean": 3714.9,
|
|
"valid_targets_min": 1227
|
|
},
|
|
{
|
|
"epoch": 2.653221581980094,
|
|
"grad_norm": 0.5933349445025315,
|
|
"learning_rate": 3.1246247620372705e-05,
|
|
"loss": 0.2598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1708788275718689,
|
|
"step": 5065,
|
|
"valid_targets_mean": 4004.1,
|
|
"valid_targets_min": 1409
|
|
},
|
|
{
|
|
"epoch": 2.6558407543216345,
|
|
"grad_norm": 0.6464940498552127,
|
|
"learning_rate": 3.122463599428688e-05,
|
|
"loss": 0.2613,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1424185037612915,
|
|
"step": 5070,
|
|
"valid_targets_mean": 2940.8,
|
|
"valid_targets_min": 885
|
|
},
|
|
{
|
|
"epoch": 2.6584599266631743,
|
|
"grad_norm": 0.6196882426891219,
|
|
"learning_rate": 3.120300521816662e-05,
|
|
"loss": 0.2632,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12973468005657196,
|
|
"step": 5075,
|
|
"valid_targets_mean": 3441.9,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 2.6610790990047146,
|
|
"grad_norm": 0.5928158810880748,
|
|
"learning_rate": 3.118135532891557e-05,
|
|
"loss": 0.2605,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12601789832115173,
|
|
"step": 5080,
|
|
"valid_targets_mean": 4079.0,
|
|
"valid_targets_min": 2331
|
|
},
|
|
{
|
|
"epoch": 2.6636982713462545,
|
|
"grad_norm": 0.5101825477237227,
|
|
"learning_rate": 3.1159686363470004e-05,
|
|
"loss": 0.2421,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13512998819351196,
|
|
"step": 5085,
|
|
"valid_targets_mean": 5371.5,
|
|
"valid_targets_min": 2618
|
|
},
|
|
{
|
|
"epoch": 2.6663174436877948,
|
|
"grad_norm": 0.607352602278322,
|
|
"learning_rate": 3.1137998358798725e-05,
|
|
"loss": 0.2551,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12532103061676025,
|
|
"step": 5090,
|
|
"valid_targets_mean": 3675.1,
|
|
"valid_targets_min": 1434
|
|
},
|
|
{
|
|
"epoch": 2.6689366160293346,
|
|
"grad_norm": 0.627662258246156,
|
|
"learning_rate": 3.1116291351903024e-05,
|
|
"loss": 0.2482,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13801538944244385,
|
|
"step": 5095,
|
|
"valid_targets_mean": 3813.4,
|
|
"valid_targets_min": 1074
|
|
},
|
|
{
|
|
"epoch": 2.671555788370875,
|
|
"grad_norm": 0.6024155475347305,
|
|
"learning_rate": 3.109456537981662e-05,
|
|
"loss": 0.26,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12260854989290237,
|
|
"step": 5100,
|
|
"valid_targets_mean": 3690.8,
|
|
"valid_targets_min": 980
|
|
},
|
|
{
|
|
"epoch": 2.674174960712415,
|
|
"grad_norm": 0.604902604391588,
|
|
"learning_rate": 3.107282047960557e-05,
|
|
"loss": 0.2535,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14585041999816895,
|
|
"step": 5105,
|
|
"valid_targets_mean": 4903.5,
|
|
"valid_targets_min": 2366
|
|
},
|
|
{
|
|
"epoch": 2.676794133053955,
|
|
"grad_norm": 0.5585577080282759,
|
|
"learning_rate": 3.1051056688368245e-05,
|
|
"loss": 0.2641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1398429274559021,
|
|
"step": 5110,
|
|
"valid_targets_mean": 4654.6,
|
|
"valid_targets_min": 1460
|
|
},
|
|
{
|
|
"epoch": 2.679413305395495,
|
|
"grad_norm": 0.6854350057724503,
|
|
"learning_rate": 3.102927404323524e-05,
|
|
"loss": 0.2512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13275103271007538,
|
|
"step": 5115,
|
|
"valid_targets_mean": 3612.0,
|
|
"valid_targets_min": 1351
|
|
},
|
|
{
|
|
"epoch": 2.6820324777370352,
|
|
"grad_norm": 0.5979776717379521,
|
|
"learning_rate": 3.1007472581369284e-05,
|
|
"loss": 0.2538,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13742196559906006,
|
|
"step": 5120,
|
|
"valid_targets_mean": 3632.5,
|
|
"valid_targets_min": 2014
|
|
},
|
|
{
|
|
"epoch": 2.684651650078575,
|
|
"grad_norm": 0.5899005881509998,
|
|
"learning_rate": 3.098565233996525e-05,
|
|
"loss": 0.2575,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11421630531549454,
|
|
"step": 5125,
|
|
"valid_targets_mean": 3135.9,
|
|
"valid_targets_min": 1084
|
|
},
|
|
{
|
|
"epoch": 2.6872708224201154,
|
|
"grad_norm": 0.6249029388850529,
|
|
"learning_rate": 3.096381335625005e-05,
|
|
"loss": 0.2455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13513506948947906,
|
|
"step": 5130,
|
|
"valid_targets_mean": 3479.2,
|
|
"valid_targets_min": 1618
|
|
},
|
|
{
|
|
"epoch": 2.6898899947616552,
|
|
"grad_norm": 0.6100810280728409,
|
|
"learning_rate": 3.094195566748254e-05,
|
|
"loss": 0.2398,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13206806778907776,
|
|
"step": 5135,
|
|
"valid_targets_mean": 3399.5,
|
|
"valid_targets_min": 1747
|
|
},
|
|
{
|
|
"epoch": 2.6925091671031955,
|
|
"grad_norm": 0.5283575842317508,
|
|
"learning_rate": 3.092007931095351e-05,
|
|
"loss": 0.2493,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14287996292114258,
|
|
"step": 5140,
|
|
"valid_targets_mean": 5037.9,
|
|
"valid_targets_min": 2092
|
|
},
|
|
{
|
|
"epoch": 2.6951283394447354,
|
|
"grad_norm": 0.5752611882581933,
|
|
"learning_rate": 3.0898184323985594e-05,
|
|
"loss": 0.2483,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13903933763504028,
|
|
"step": 5145,
|
|
"valid_targets_mean": 3843.4,
|
|
"valid_targets_min": 2786
|
|
},
|
|
{
|
|
"epoch": 2.6977475117862757,
|
|
"grad_norm": 0.535866848835118,
|
|
"learning_rate": 3.0876270743933214e-05,
|
|
"loss": 0.2368,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10054393112659454,
|
|
"step": 5150,
|
|
"valid_targets_mean": 4153.8,
|
|
"valid_targets_min": 1246
|
|
},
|
|
{
|
|
"epoch": 2.7003666841278156,
|
|
"grad_norm": 0.5746289328777828,
|
|
"learning_rate": 3.08543386081825e-05,
|
|
"loss": 0.2421,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10839236527681351,
|
|
"step": 5155,
|
|
"valid_targets_mean": 3389.8,
|
|
"valid_targets_min": 1484
|
|
},
|
|
{
|
|
"epoch": 2.702985856469356,
|
|
"grad_norm": 0.578449089828753,
|
|
"learning_rate": 3.083238795415126e-05,
|
|
"loss": 0.2481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1460162103176117,
|
|
"step": 5160,
|
|
"valid_targets_mean": 4818.0,
|
|
"valid_targets_min": 3185
|
|
},
|
|
{
|
|
"epoch": 2.7056050288108957,
|
|
"grad_norm": 0.6219474109360055,
|
|
"learning_rate": 3.0810418819288886e-05,
|
|
"loss": 0.2471,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11964847147464752,
|
|
"step": 5165,
|
|
"valid_targets_mean": 3818.9,
|
|
"valid_targets_min": 1530
|
|
},
|
|
{
|
|
"epoch": 2.7082242011524356,
|
|
"grad_norm": 0.6404418781897823,
|
|
"learning_rate": 3.0788431241076286e-05,
|
|
"loss": 0.2719,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.148534893989563,
|
|
"step": 5170,
|
|
"valid_targets_mean": 3010.2,
|
|
"valid_targets_min": 887
|
|
},
|
|
{
|
|
"epoch": 2.710843373493976,
|
|
"grad_norm": 0.5670251738122322,
|
|
"learning_rate": 3.076642525702586e-05,
|
|
"loss": 0.2432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12288151681423187,
|
|
"step": 5175,
|
|
"valid_targets_mean": 4294.1,
|
|
"valid_targets_min": 2525
|
|
},
|
|
{
|
|
"epoch": 2.713462545835516,
|
|
"grad_norm": 0.5556573831730579,
|
|
"learning_rate": 3.074440090468138e-05,
|
|
"loss": 0.2472,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1274084895849228,
|
|
"step": 5180,
|
|
"valid_targets_mean": 4256.4,
|
|
"valid_targets_min": 1759
|
|
},
|
|
{
|
|
"epoch": 2.716081718177056,
|
|
"grad_norm": 0.5237360238484792,
|
|
"learning_rate": 3.072235822161799e-05,
|
|
"loss": 0.2408,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12600873410701752,
|
|
"step": 5185,
|
|
"valid_targets_mean": 4883.9,
|
|
"valid_targets_min": 2155
|
|
},
|
|
{
|
|
"epoch": 2.718700890518596,
|
|
"grad_norm": 0.6052618076285045,
|
|
"learning_rate": 3.070029724544208e-05,
|
|
"loss": 0.2558,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12613743543624878,
|
|
"step": 5190,
|
|
"valid_targets_mean": 3758.9,
|
|
"valid_targets_min": 2048
|
|
},
|
|
{
|
|
"epoch": 2.721320062860136,
|
|
"grad_norm": 0.6131391571695859,
|
|
"learning_rate": 3.067821801379125e-05,
|
|
"loss": 0.2521,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10715790092945099,
|
|
"step": 5195,
|
|
"valid_targets_mean": 2550.0,
|
|
"valid_targets_min": 1233
|
|
},
|
|
{
|
|
"epoch": 2.7239392352016765,
|
|
"grad_norm": 0.5511481275343706,
|
|
"learning_rate": 3.0656120564334265e-05,
|
|
"loss": 0.2559,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09842834621667862,
|
|
"step": 5200,
|
|
"valid_targets_mean": 2796.5,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 2.7265584075432163,
|
|
"grad_norm": 0.5393218879075641,
|
|
"learning_rate": 3.063400493477097e-05,
|
|
"loss": 0.251,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1217995434999466,
|
|
"step": 5205,
|
|
"valid_targets_mean": 4085.5,
|
|
"valid_targets_min": 1800
|
|
},
|
|
{
|
|
"epoch": 2.729177579884756,
|
|
"grad_norm": 0.5974752377112322,
|
|
"learning_rate": 3.061187116283219e-05,
|
|
"loss": 0.2528,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12144285440444946,
|
|
"step": 5210,
|
|
"valid_targets_mean": 3429.1,
|
|
"valid_targets_min": 2327
|
|
},
|
|
{
|
|
"epoch": 2.7317967522262965,
|
|
"grad_norm": 0.5971786736186465,
|
|
"learning_rate": 3.0589719286279736e-05,
|
|
"loss": 0.2499,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12138991802930832,
|
|
"step": 5215,
|
|
"valid_targets_mean": 3262.4,
|
|
"valid_targets_min": 1597
|
|
},
|
|
{
|
|
"epoch": 2.7344159245678368,
|
|
"grad_norm": 0.43442958150724836,
|
|
"learning_rate": 3.056754934290631e-05,
|
|
"loss": 0.168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07654788345098495,
|
|
"step": 5220,
|
|
"valid_targets_mean": 3631.8,
|
|
"valid_targets_min": 2342
|
|
},
|
|
{
|
|
"epoch": 2.7370350969093766,
|
|
"grad_norm": 0.4518304418024818,
|
|
"learning_rate": 3.054536137053542e-05,
|
|
"loss": 0.1062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039826858788728714,
|
|
"step": 5225,
|
|
"valid_targets_mean": 3542.6,
|
|
"valid_targets_min": 2731
|
|
},
|
|
{
|
|
"epoch": 2.7396542692509165,
|
|
"grad_norm": 0.3723356923704959,
|
|
"learning_rate": 3.0523155407021343e-05,
|
|
"loss": 0.0993,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03338192030787468,
|
|
"step": 5230,
|
|
"valid_targets_mean": 3569.4,
|
|
"valid_targets_min": 2927
|
|
},
|
|
{
|
|
"epoch": 2.742273441592457,
|
|
"grad_norm": 0.4012919897718542,
|
|
"learning_rate": 3.0500931490249032e-05,
|
|
"loss": 0.1141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06943926960229874,
|
|
"step": 5235,
|
|
"valid_targets_mean": 3426.9,
|
|
"valid_targets_min": 2213
|
|
},
|
|
{
|
|
"epoch": 2.744892613933997,
|
|
"grad_norm": 0.46765799501116584,
|
|
"learning_rate": 3.04786896581341e-05,
|
|
"loss": 0.0949,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055836021900177,
|
|
"step": 5240,
|
|
"valid_targets_mean": 2539.8,
|
|
"valid_targets_min": 560
|
|
},
|
|
{
|
|
"epoch": 2.747511786275537,
|
|
"grad_norm": 0.4050166095909045,
|
|
"learning_rate": 3.0456429948622702e-05,
|
|
"loss": 0.1012,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05261792242527008,
|
|
"step": 5245,
|
|
"valid_targets_mean": 2907.5,
|
|
"valid_targets_min": 1002
|
|
},
|
|
{
|
|
"epoch": 2.750130958617077,
|
|
"grad_norm": 0.6799795998400201,
|
|
"learning_rate": 3.04341523996915e-05,
|
|
"loss": 0.1633,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08086485415697098,
|
|
"step": 5250,
|
|
"valid_targets_mean": 1458.6,
|
|
"valid_targets_min": 650
|
|
},
|
|
{
|
|
"epoch": 2.752750130958617,
|
|
"grad_norm": 0.37334798666819397,
|
|
"learning_rate": 3.041185704934759e-05,
|
|
"loss": 0.0989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04849667102098465,
|
|
"step": 5255,
|
|
"valid_targets_mean": 3396.1,
|
|
"valid_targets_min": 1037
|
|
},
|
|
{
|
|
"epoch": 2.7553693033001574,
|
|
"grad_norm": 0.5441076998623068,
|
|
"learning_rate": 3.0389543935628438e-05,
|
|
"loss": 0.1245,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1124047264456749,
|
|
"step": 5260,
|
|
"valid_targets_mean": 2438.9,
|
|
"valid_targets_min": 883
|
|
},
|
|
{
|
|
"epoch": 2.7579884756416972,
|
|
"grad_norm": 0.24267328190315704,
|
|
"learning_rate": 3.0367213096601822e-05,
|
|
"loss": 0.1197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03160983324050903,
|
|
"step": 5265,
|
|
"valid_targets_mean": 4958.4,
|
|
"valid_targets_min": 3285
|
|
},
|
|
{
|
|
"epoch": 2.760607647983237,
|
|
"grad_norm": 0.3283148739817792,
|
|
"learning_rate": 3.0344864570365752e-05,
|
|
"loss": 0.0927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039658188819885254,
|
|
"step": 5270,
|
|
"valid_targets_mean": 4092.4,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 2.7632268203247774,
|
|
"grad_norm": 0.3607218133338841,
|
|
"learning_rate": 3.0322498395048422e-05,
|
|
"loss": 0.0983,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050965845584869385,
|
|
"step": 5275,
|
|
"valid_targets_mean": 3804.9,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 2.7658459926663177,
|
|
"grad_norm": 0.4909546931767392,
|
|
"learning_rate": 3.030011460880812e-05,
|
|
"loss": 0.0981,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06120920181274414,
|
|
"step": 5280,
|
|
"valid_targets_mean": 2547.8,
|
|
"valid_targets_min": 729
|
|
},
|
|
{
|
|
"epoch": 2.7684651650078576,
|
|
"grad_norm": 0.3042317920249153,
|
|
"learning_rate": 3.027771324983321e-05,
|
|
"loss": 0.137,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043728526681661606,
|
|
"step": 5285,
|
|
"valid_targets_mean": 4258.8,
|
|
"valid_targets_min": 1961
|
|
},
|
|
{
|
|
"epoch": 2.7710843373493974,
|
|
"grad_norm": 0.31391062046456664,
|
|
"learning_rate": 3.0255294356342e-05,
|
|
"loss": 0.0924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04248586297035217,
|
|
"step": 5290,
|
|
"valid_targets_mean": 3557.5,
|
|
"valid_targets_min": 1822
|
|
},
|
|
{
|
|
"epoch": 2.7737035096909377,
|
|
"grad_norm": 0.4130453023231338,
|
|
"learning_rate": 3.0232857966582746e-05,
|
|
"loss": 0.1085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05159539356827736,
|
|
"step": 5295,
|
|
"valid_targets_mean": 2159.6,
|
|
"valid_targets_min": 841
|
|
},
|
|
{
|
|
"epoch": 2.776322682032478,
|
|
"grad_norm": 0.37793693540997453,
|
|
"learning_rate": 3.0210404118833527e-05,
|
|
"loss": 0.085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045104995369911194,
|
|
"step": 5300,
|
|
"valid_targets_mean": 2735.1,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 2.778941854374018,
|
|
"grad_norm": 0.7126770459670307,
|
|
"learning_rate": 3.0187932851402225e-05,
|
|
"loss": 0.1075,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0645991638302803,
|
|
"step": 5305,
|
|
"valid_targets_mean": 1669.0,
|
|
"valid_targets_min": 770
|
|
},
|
|
{
|
|
"epoch": 2.7815610267155577,
|
|
"grad_norm": 0.39540853200518833,
|
|
"learning_rate": 3.0165444202626433e-05,
|
|
"loss": 0.1099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041591890156269073,
|
|
"step": 5310,
|
|
"valid_targets_mean": 2415.0,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 2.784180199057098,
|
|
"grad_norm": 0.43367018565424464,
|
|
"learning_rate": 3.01429382108734e-05,
|
|
"loss": 0.1149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05857367068529129,
|
|
"step": 5315,
|
|
"valid_targets_mean": 3279.4,
|
|
"valid_targets_min": 2224
|
|
},
|
|
{
|
|
"epoch": 2.786799371398638,
|
|
"grad_norm": 0.4122107227346459,
|
|
"learning_rate": 3.012041491453996e-05,
|
|
"loss": 0.1146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05071114003658295,
|
|
"step": 5320,
|
|
"valid_targets_mean": 2944.5,
|
|
"valid_targets_min": 1341
|
|
},
|
|
{
|
|
"epoch": 2.789418543740178,
|
|
"grad_norm": 0.3863943077409809,
|
|
"learning_rate": 3.0097874352052477e-05,
|
|
"loss": 0.1475,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05821670591831207,
|
|
"step": 5325,
|
|
"valid_targets_mean": 3912.9,
|
|
"valid_targets_min": 3005
|
|
},
|
|
{
|
|
"epoch": 2.792037716081718,
|
|
"grad_norm": 0.363574384955353,
|
|
"learning_rate": 3.0075316561866766e-05,
|
|
"loss": 0.0925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04231640696525574,
|
|
"step": 5330,
|
|
"valid_targets_mean": 2799.0,
|
|
"valid_targets_min": 669
|
|
},
|
|
{
|
|
"epoch": 2.7946568884232583,
|
|
"grad_norm": 0.6689930239576413,
|
|
"learning_rate": 3.0052741582468034e-05,
|
|
"loss": 0.1149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07634914666414261,
|
|
"step": 5335,
|
|
"valid_targets_mean": 1381.5,
|
|
"valid_targets_min": 702
|
|
},
|
|
{
|
|
"epoch": 2.797276060764798,
|
|
"grad_norm": 0.6259194767422449,
|
|
"learning_rate": 3.003014945237082e-05,
|
|
"loss": 0.1297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04922054335474968,
|
|
"step": 5340,
|
|
"valid_targets_mean": 869.9,
|
|
"valid_targets_min": 15
|
|
},
|
|
{
|
|
"epoch": 2.7998952331063385,
|
|
"grad_norm": 0.35584135567152586,
|
|
"learning_rate": 3.0007540210118916e-05,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05319155752658844,
|
|
"step": 5345,
|
|
"valid_targets_mean": 3761.5,
|
|
"valid_targets_min": 2827
|
|
},
|
|
{
|
|
"epoch": 2.8025144054478783,
|
|
"grad_norm": 0.46814357476427576,
|
|
"learning_rate": 2.998491389428531e-05,
|
|
"loss": 0.1648,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10782517492771149,
|
|
"step": 5350,
|
|
"valid_targets_mean": 3383.4,
|
|
"valid_targets_min": 2079
|
|
},
|
|
{
|
|
"epoch": 2.8051335777894186,
|
|
"grad_norm": 0.3933848920767778,
|
|
"learning_rate": 2.9962270543472125e-05,
|
|
"loss": 0.2572,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051113810390233994,
|
|
"step": 5355,
|
|
"valid_targets_mean": 3586.4,
|
|
"valid_targets_min": 1932
|
|
},
|
|
{
|
|
"epoch": 2.8077527501309585,
|
|
"grad_norm": 0.5260067132659276,
|
|
"learning_rate": 2.993961019631054e-05,
|
|
"loss": 0.1201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05595569685101509,
|
|
"step": 5360,
|
|
"valid_targets_mean": 1462.1,
|
|
"valid_targets_min": 820
|
|
},
|
|
{
|
|
"epoch": 2.810371922472499,
|
|
"grad_norm": 0.4903945335004159,
|
|
"learning_rate": 2.9916932891460732e-05,
|
|
"loss": 0.1173,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10142762213945389,
|
|
"step": 5365,
|
|
"valid_targets_mean": 2709.1,
|
|
"valid_targets_min": 965
|
|
},
|
|
{
|
|
"epoch": 2.8129910948140386,
|
|
"grad_norm": 0.3286219206131438,
|
|
"learning_rate": 2.9894238667611816e-05,
|
|
"loss": 0.1924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04985406622290611,
|
|
"step": 5370,
|
|
"valid_targets_mean": 4498.0,
|
|
"valid_targets_min": 2262
|
|
},
|
|
{
|
|
"epoch": 2.815610267155579,
|
|
"grad_norm": 0.413834999480805,
|
|
"learning_rate": 2.987152756348176e-05,
|
|
"loss": 0.0962,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08123902976512909,
|
|
"step": 5375,
|
|
"valid_targets_mean": 4270.9,
|
|
"valid_targets_min": 2733
|
|
},
|
|
{
|
|
"epoch": 2.818229439497119,
|
|
"grad_norm": 0.38853493422731167,
|
|
"learning_rate": 2.9848799617817336e-05,
|
|
"loss": 0.1036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07313060760498047,
|
|
"step": 5380,
|
|
"valid_targets_mean": 4916.5,
|
|
"valid_targets_min": 3028
|
|
},
|
|
{
|
|
"epoch": 2.820848611838659,
|
|
"grad_norm": 0.3180382314400335,
|
|
"learning_rate": 2.9826054869394054e-05,
|
|
"loss": 0.1041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03528398275375366,
|
|
"step": 5385,
|
|
"valid_targets_mean": 4158.5,
|
|
"valid_targets_min": 3317
|
|
},
|
|
{
|
|
"epoch": 2.823467784180199,
|
|
"grad_norm": 0.3309466292419488,
|
|
"learning_rate": 2.9803293357016087e-05,
|
|
"loss": 0.0954,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041354067623615265,
|
|
"step": 5390,
|
|
"valid_targets_mean": 4017.0,
|
|
"valid_targets_min": 748
|
|
},
|
|
{
|
|
"epoch": 2.8260869565217392,
|
|
"grad_norm": 0.39593819578991357,
|
|
"learning_rate": 2.9780515119516208e-05,
|
|
"loss": 0.0951,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059333547949790955,
|
|
"step": 5395,
|
|
"valid_targets_mean": 3627.6,
|
|
"valid_targets_min": 2904
|
|
},
|
|
{
|
|
"epoch": 2.828706128863279,
|
|
"grad_norm": 0.4648870637564455,
|
|
"learning_rate": 2.975772019575572e-05,
|
|
"loss": 0.1002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05297884717583656,
|
|
"step": 5400,
|
|
"valid_targets_mean": 2574.8,
|
|
"valid_targets_min": 1939
|
|
},
|
|
{
|
|
"epoch": 2.8313253012048194,
|
|
"grad_norm": 0.402889325989699,
|
|
"learning_rate": 2.9734908624624407e-05,
|
|
"loss": 0.084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04135642200708389,
|
|
"step": 5405,
|
|
"valid_targets_mean": 2599.0,
|
|
"valid_targets_min": 772
|
|
},
|
|
{
|
|
"epoch": 2.8339444735463593,
|
|
"grad_norm": 0.5410229605482512,
|
|
"learning_rate": 2.971208044504044e-05,
|
|
"loss": 0.1508,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09342281520366669,
|
|
"step": 5410,
|
|
"valid_targets_mean": 2805.9,
|
|
"valid_targets_min": 1289
|
|
},
|
|
{
|
|
"epoch": 2.8365636458878996,
|
|
"grad_norm": 0.3768403055787591,
|
|
"learning_rate": 2.968923569595034e-05,
|
|
"loss": 0.0774,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04221757873892784,
|
|
"step": 5415,
|
|
"valid_targets_mean": 3667.9,
|
|
"valid_targets_min": 2730
|
|
},
|
|
{
|
|
"epoch": 2.8391828182294394,
|
|
"grad_norm": 0.38783078033593726,
|
|
"learning_rate": 2.9666374416328874e-05,
|
|
"loss": 0.0912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0400664359331131,
|
|
"step": 5420,
|
|
"valid_targets_mean": 2270.9,
|
|
"valid_targets_min": 616
|
|
},
|
|
{
|
|
"epoch": 2.8418019905709797,
|
|
"grad_norm": 0.33047465013092353,
|
|
"learning_rate": 2.9643496645179034e-05,
|
|
"loss": 0.0862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04919986054301262,
|
|
"step": 5425,
|
|
"valid_targets_mean": 4381.4,
|
|
"valid_targets_min": 3979
|
|
},
|
|
{
|
|
"epoch": 2.8444211629125196,
|
|
"grad_norm": 0.4061462618944902,
|
|
"learning_rate": 2.9620602421531944e-05,
|
|
"loss": 0.1029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0646735206246376,
|
|
"step": 5430,
|
|
"valid_targets_mean": 2635.8,
|
|
"valid_targets_min": 795
|
|
},
|
|
{
|
|
"epoch": 2.84704033525406,
|
|
"grad_norm": 0.3296198223401941,
|
|
"learning_rate": 2.959769178444678e-05,
|
|
"loss": 0.1302,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04333513230085373,
|
|
"step": 5435,
|
|
"valid_targets_mean": 3182.4,
|
|
"valid_targets_min": 729
|
|
},
|
|
{
|
|
"epoch": 2.8496595075955997,
|
|
"grad_norm": 0.5134410202466364,
|
|
"learning_rate": 2.9574764773010743e-05,
|
|
"loss": 0.1258,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05219130218029022,
|
|
"step": 5440,
|
|
"valid_targets_mean": 1878.5,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 2.85227867993714,
|
|
"grad_norm": 0.603936272047619,
|
|
"learning_rate": 2.9551821426338958e-05,
|
|
"loss": 0.1134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09199359267950058,
|
|
"step": 5445,
|
|
"valid_targets_mean": 2249.8,
|
|
"valid_targets_min": 664
|
|
},
|
|
{
|
|
"epoch": 2.85489785227868,
|
|
"grad_norm": 0.4318924619202293,
|
|
"learning_rate": 2.9528861783574416e-05,
|
|
"loss": 0.1215,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043887414038181305,
|
|
"step": 5450,
|
|
"valid_targets_mean": 2284.8,
|
|
"valid_targets_min": 890
|
|
},
|
|
{
|
|
"epoch": 2.85751702462022,
|
|
"grad_norm": 0.34140930648182255,
|
|
"learning_rate": 2.950588588388792e-05,
|
|
"loss": 0.1407,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045628830790519714,
|
|
"step": 5455,
|
|
"valid_targets_mean": 3838.2,
|
|
"valid_targets_min": 2440
|
|
},
|
|
{
|
|
"epoch": 2.86013619696176,
|
|
"grad_norm": 0.400531024076249,
|
|
"learning_rate": 2.9482893766478e-05,
|
|
"loss": 0.1043,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041745901107788086,
|
|
"step": 5460,
|
|
"valid_targets_mean": 3074.5,
|
|
"valid_targets_min": 840
|
|
},
|
|
{
|
|
"epoch": 2.8627553693033003,
|
|
"grad_norm": 0.4452066763835879,
|
|
"learning_rate": 2.9459885470570856e-05,
|
|
"loss": 0.1263,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05026886984705925,
|
|
"step": 5465,
|
|
"valid_targets_mean": 3609.8,
|
|
"valid_targets_min": 1014
|
|
},
|
|
{
|
|
"epoch": 2.86537454164484,
|
|
"grad_norm": 0.42926121252820065,
|
|
"learning_rate": 2.9436861035420298e-05,
|
|
"loss": 0.1082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044806621968746185,
|
|
"step": 5470,
|
|
"valid_targets_mean": 1142.9,
|
|
"valid_targets_min": 485
|
|
},
|
|
{
|
|
"epoch": 2.8679937139863805,
|
|
"grad_norm": 0.4386900241651716,
|
|
"learning_rate": 2.9413820500307665e-05,
|
|
"loss": 0.0971,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04453166574239731,
|
|
"step": 5475,
|
|
"valid_targets_mean": 2665.6,
|
|
"valid_targets_min": 605
|
|
},
|
|
{
|
|
"epoch": 2.8706128863279203,
|
|
"grad_norm": 0.28254308805117223,
|
|
"learning_rate": 2.939076390454176e-05,
|
|
"loss": 0.0934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03537259250879288,
|
|
"step": 5480,
|
|
"valid_targets_mean": 4443.1,
|
|
"valid_targets_min": 848
|
|
},
|
|
{
|
|
"epoch": 2.87323205866946,
|
|
"grad_norm": 0.3871232976277066,
|
|
"learning_rate": 2.9367691287458786e-05,
|
|
"loss": 0.093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05675549805164337,
|
|
"step": 5485,
|
|
"valid_targets_mean": 4255.2,
|
|
"valid_targets_min": 884
|
|
},
|
|
{
|
|
"epoch": 2.8758512310110005,
|
|
"grad_norm": 0.3994867830249035,
|
|
"learning_rate": 2.9344602688422286e-05,
|
|
"loss": 0.0936,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.048137348145246506,
|
|
"step": 5490,
|
|
"valid_targets_mean": 3826.2,
|
|
"valid_targets_min": 1992
|
|
},
|
|
{
|
|
"epoch": 2.878470403352541,
|
|
"grad_norm": 0.37453332633288844,
|
|
"learning_rate": 2.932149814682308e-05,
|
|
"loss": 0.107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06611055880784988,
|
|
"step": 5495,
|
|
"valid_targets_mean": 3189.2,
|
|
"valid_targets_min": 808
|
|
},
|
|
{
|
|
"epoch": 2.8810895756940806,
|
|
"grad_norm": 0.42821727671187865,
|
|
"learning_rate": 2.929837770207915e-05,
|
|
"loss": 0.1058,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05029541254043579,
|
|
"step": 5500,
|
|
"valid_targets_mean": 3188.5,
|
|
"valid_targets_min": 1814
|
|
},
|
|
{
|
|
"epoch": 2.8837087480356205,
|
|
"grad_norm": 0.42426030057120456,
|
|
"learning_rate": 2.927524139363565e-05,
|
|
"loss": 0.1118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05611751228570938,
|
|
"step": 5505,
|
|
"valid_targets_mean": 2962.6,
|
|
"valid_targets_min": 931
|
|
},
|
|
{
|
|
"epoch": 2.886327920377161,
|
|
"grad_norm": 0.3289453216560865,
|
|
"learning_rate": 2.925208926096478e-05,
|
|
"loss": 0.0852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04311997443437576,
|
|
"step": 5510,
|
|
"valid_targets_mean": 4098.5,
|
|
"valid_targets_min": 3848
|
|
},
|
|
{
|
|
"epoch": 2.888947092718701,
|
|
"grad_norm": 0.48169193309453806,
|
|
"learning_rate": 2.9228921343565733e-05,
|
|
"loss": 0.1014,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052452776581048965,
|
|
"step": 5515,
|
|
"valid_targets_mean": 1851.2,
|
|
"valid_targets_min": 664
|
|
},
|
|
{
|
|
"epoch": 2.891566265060241,
|
|
"grad_norm": 0.37235486925460154,
|
|
"learning_rate": 2.9205737680964645e-05,
|
|
"loss": 0.1038,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05145837739109993,
|
|
"step": 5520,
|
|
"valid_targets_mean": 3527.9,
|
|
"valid_targets_min": 2309
|
|
},
|
|
{
|
|
"epoch": 2.894185437401781,
|
|
"grad_norm": 0.39994110176146797,
|
|
"learning_rate": 2.9182538312714506e-05,
|
|
"loss": 0.0827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041609033942222595,
|
|
"step": 5525,
|
|
"valid_targets_mean": 2707.0,
|
|
"valid_targets_min": 1003
|
|
},
|
|
{
|
|
"epoch": 2.896804609743321,
|
|
"grad_norm": 0.3375712957693352,
|
|
"learning_rate": 2.91593232783951e-05,
|
|
"loss": 0.1022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03793808072805405,
|
|
"step": 5530,
|
|
"valid_targets_mean": 3107.2,
|
|
"valid_targets_min": 531
|
|
},
|
|
{
|
|
"epoch": 2.8994237820848614,
|
|
"grad_norm": 0.35215830014753513,
|
|
"learning_rate": 2.913609261761295e-05,
|
|
"loss": 0.102,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046030208468437195,
|
|
"step": 5535,
|
|
"valid_targets_mean": 2845.9,
|
|
"valid_targets_min": 1003
|
|
},
|
|
{
|
|
"epoch": 2.9020429544264013,
|
|
"grad_norm": 0.4671244222007615,
|
|
"learning_rate": 2.9112846370001223e-05,
|
|
"loss": 0.1221,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03913413733243942,
|
|
"step": 5540,
|
|
"valid_targets_mean": 1568.5,
|
|
"valid_targets_min": 597
|
|
},
|
|
{
|
|
"epoch": 2.904662126767941,
|
|
"grad_norm": 0.40153476187632986,
|
|
"learning_rate": 2.9089584575219688e-05,
|
|
"loss": 0.1006,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03844350203871727,
|
|
"step": 5545,
|
|
"valid_targets_mean": 1980.2,
|
|
"valid_targets_min": 665
|
|
},
|
|
{
|
|
"epoch": 2.9072812991094814,
|
|
"grad_norm": 0.49738924762354936,
|
|
"learning_rate": 2.906630727295463e-05,
|
|
"loss": 0.1192,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04668961465358734,
|
|
"step": 5550,
|
|
"valid_targets_mean": 1768.5,
|
|
"valid_targets_min": 853
|
|
},
|
|
{
|
|
"epoch": 2.9099004714510217,
|
|
"grad_norm": 0.44675384756678116,
|
|
"learning_rate": 2.9043014502918807e-05,
|
|
"loss": 0.1068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07055791467428207,
|
|
"step": 5555,
|
|
"valid_targets_mean": 2583.2,
|
|
"valid_targets_min": 997
|
|
},
|
|
{
|
|
"epoch": 2.9125196437925616,
|
|
"grad_norm": 0.5219035167189274,
|
|
"learning_rate": 2.9019706304851348e-05,
|
|
"loss": 0.1136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06573070585727692,
|
|
"step": 5560,
|
|
"valid_targets_mean": 2746.6,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 2.9151388161341014,
|
|
"grad_norm": 0.31231730341467273,
|
|
"learning_rate": 2.8996382718517724e-05,
|
|
"loss": 0.0937,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03759889677166939,
|
|
"step": 5565,
|
|
"valid_targets_mean": 3555.2,
|
|
"valid_targets_min": 2887
|
|
},
|
|
{
|
|
"epoch": 2.9177579884756417,
|
|
"grad_norm": 0.33964128112886155,
|
|
"learning_rate": 2.8973043783709628e-05,
|
|
"loss": 0.0946,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03595292568206787,
|
|
"step": 5570,
|
|
"valid_targets_mean": 3754.2,
|
|
"valid_targets_min": 2543
|
|
},
|
|
{
|
|
"epoch": 2.920377160817182,
|
|
"grad_norm": 0.48461253619580297,
|
|
"learning_rate": 2.894968954024497e-05,
|
|
"loss": 0.126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05458501726388931,
|
|
"step": 5575,
|
|
"valid_targets_mean": 1742.1,
|
|
"valid_targets_min": 446
|
|
},
|
|
{
|
|
"epoch": 2.922996333158722,
|
|
"grad_norm": 0.36508330284009327,
|
|
"learning_rate": 2.8926320027967763e-05,
|
|
"loss": 0.0872,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03947306424379349,
|
|
"step": 5580,
|
|
"valid_targets_mean": 2742.4,
|
|
"valid_targets_min": 681
|
|
},
|
|
{
|
|
"epoch": 2.9256155055002617,
|
|
"grad_norm": 0.39691509130909497,
|
|
"learning_rate": 2.8902935286748066e-05,
|
|
"loss": 0.0964,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07151953876018524,
|
|
"step": 5585,
|
|
"valid_targets_mean": 3844.6,
|
|
"valid_targets_min": 799
|
|
},
|
|
{
|
|
"epoch": 2.928234677841802,
|
|
"grad_norm": 0.37319668001393097,
|
|
"learning_rate": 2.8879535356481933e-05,
|
|
"loss": 0.0942,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04579343646764755,
|
|
"step": 5590,
|
|
"valid_targets_mean": 3546.4,
|
|
"valid_targets_min": 940
|
|
},
|
|
{
|
|
"epoch": 2.9308538501833423,
|
|
"grad_norm": 0.3072038585673081,
|
|
"learning_rate": 2.8856120277091326e-05,
|
|
"loss": 0.0829,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034740135073661804,
|
|
"step": 5595,
|
|
"valid_targets_mean": 3592.6,
|
|
"valid_targets_min": 861
|
|
},
|
|
{
|
|
"epoch": 2.933473022524882,
|
|
"grad_norm": 0.4037290376400598,
|
|
"learning_rate": 2.8832690088524045e-05,
|
|
"loss": 0.0924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06839783489704132,
|
|
"step": 5600,
|
|
"valid_targets_mean": 4078.1,
|
|
"valid_targets_min": 3024
|
|
},
|
|
{
|
|
"epoch": 2.936092194866422,
|
|
"grad_norm": 0.3119248430637808,
|
|
"learning_rate": 2.8809244830753667e-05,
|
|
"loss": 0.0974,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05406394228339195,
|
|
"step": 5605,
|
|
"valid_targets_mean": 3932.1,
|
|
"valid_targets_min": 3175
|
|
},
|
|
{
|
|
"epoch": 2.9387113672079623,
|
|
"grad_norm": 0.3498723220699873,
|
|
"learning_rate": 2.8785784543779497e-05,
|
|
"loss": 0.1182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04904625564813614,
|
|
"step": 5610,
|
|
"valid_targets_mean": 3798.4,
|
|
"valid_targets_min": 2652
|
|
},
|
|
{
|
|
"epoch": 2.9413305395495026,
|
|
"grad_norm": 0.4385505136502549,
|
|
"learning_rate": 2.8762309267626465e-05,
|
|
"loss": 0.121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11120840162038803,
|
|
"step": 5615,
|
|
"valid_targets_mean": 2981.9,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 2.9439497118910425,
|
|
"grad_norm": 0.4153393952520855,
|
|
"learning_rate": 2.8738819042345073e-05,
|
|
"loss": 0.0793,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05048034340143204,
|
|
"step": 5620,
|
|
"valid_targets_mean": 1483.8,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 2.9465688842325823,
|
|
"grad_norm": 0.37736301208032946,
|
|
"learning_rate": 2.8715313908011332e-05,
|
|
"loss": 0.1187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05640570446848869,
|
|
"step": 5625,
|
|
"valid_targets_mean": 3776.4,
|
|
"valid_targets_min": 2447
|
|
},
|
|
{
|
|
"epoch": 2.9491880565741226,
|
|
"grad_norm": 0.41033167104265644,
|
|
"learning_rate": 2.8691793904726685e-05,
|
|
"loss": 0.0941,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053387630730867386,
|
|
"step": 5630,
|
|
"valid_targets_mean": 3455.1,
|
|
"valid_targets_min": 854
|
|
},
|
|
{
|
|
"epoch": 2.9518072289156625,
|
|
"grad_norm": 0.33174583059520024,
|
|
"learning_rate": 2.8668259072617946e-05,
|
|
"loss": 0.1028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03358113765716553,
|
|
"step": 5635,
|
|
"valid_targets_mean": 3706.0,
|
|
"valid_targets_min": 743
|
|
},
|
|
{
|
|
"epoch": 2.954426401257203,
|
|
"grad_norm": 0.5185327579576747,
|
|
"learning_rate": 2.864470945183724e-05,
|
|
"loss": 0.1167,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09212501347064972,
|
|
"step": 5640,
|
|
"valid_targets_mean": 2587.1,
|
|
"valid_targets_min": 706
|
|
},
|
|
{
|
|
"epoch": 2.9570455735987426,
|
|
"grad_norm": 0.4414609757950001,
|
|
"learning_rate": 2.8621145082561886e-05,
|
|
"loss": 0.1415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09895250201225281,
|
|
"step": 5645,
|
|
"valid_targets_mean": 3028.0,
|
|
"valid_targets_min": 2457
|
|
},
|
|
{
|
|
"epoch": 2.959664745940283,
|
|
"grad_norm": 0.41969596363187245,
|
|
"learning_rate": 2.8597566004994407e-05,
|
|
"loss": 0.1081,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0406753309071064,
|
|
"step": 5650,
|
|
"valid_targets_mean": 2147.2,
|
|
"valid_targets_min": 663
|
|
},
|
|
{
|
|
"epoch": 2.962283918281823,
|
|
"grad_norm": 0.4634764618622666,
|
|
"learning_rate": 2.8573972259362405e-05,
|
|
"loss": 0.1002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049560319632291794,
|
|
"step": 5655,
|
|
"valid_targets_mean": 1507.4,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 2.964903090623363,
|
|
"grad_norm": 0.3753956288776308,
|
|
"learning_rate": 2.855036388591848e-05,
|
|
"loss": 0.0942,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057809121906757355,
|
|
"step": 5660,
|
|
"valid_targets_mean": 3088.1,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 2.967522262964903,
|
|
"grad_norm": 0.3927551212565586,
|
|
"learning_rate": 2.8526740924940234e-05,
|
|
"loss": 0.0959,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052389007061719894,
|
|
"step": 5665,
|
|
"valid_targets_mean": 3074.9,
|
|
"valid_targets_min": 920
|
|
},
|
|
{
|
|
"epoch": 2.9701414353064433,
|
|
"grad_norm": 0.3920311639286199,
|
|
"learning_rate": 2.8503103416730123e-05,
|
|
"loss": 0.0845,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032711006700992584,
|
|
"step": 5670,
|
|
"valid_targets_mean": 997.5,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 2.972760607647983,
|
|
"grad_norm": 0.34637553831989293,
|
|
"learning_rate": 2.8479451401615434e-05,
|
|
"loss": 0.1191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04361840337514877,
|
|
"step": 5675,
|
|
"valid_targets_mean": 2942.0,
|
|
"valid_targets_min": 920
|
|
},
|
|
{
|
|
"epoch": 2.9753797799895234,
|
|
"grad_norm": 0.39651341139648016,
|
|
"learning_rate": 2.8455784919948206e-05,
|
|
"loss": 0.0857,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042224809527397156,
|
|
"step": 5680,
|
|
"valid_targets_mean": 1271.1,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 2.9779989523310633,
|
|
"grad_norm": 0.38317061342586256,
|
|
"learning_rate": 2.8432104012105144e-05,
|
|
"loss": 0.0943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034398600459098816,
|
|
"step": 5685,
|
|
"valid_targets_mean": 2149.4,
|
|
"valid_targets_min": 529
|
|
},
|
|
{
|
|
"epoch": 2.9806181246726036,
|
|
"grad_norm": 0.3884152721470461,
|
|
"learning_rate": 2.8408408718487582e-05,
|
|
"loss": 0.1029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05746208131313324,
|
|
"step": 5690,
|
|
"valid_targets_mean": 3471.9,
|
|
"valid_targets_min": 2017
|
|
},
|
|
{
|
|
"epoch": 2.9832372970141434,
|
|
"grad_norm": 0.30370806170817466,
|
|
"learning_rate": 2.8384699079521398e-05,
|
|
"loss": 0.0931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0494544580578804,
|
|
"step": 5695,
|
|
"valid_targets_mean": 4240.2,
|
|
"valid_targets_min": 516
|
|
},
|
|
{
|
|
"epoch": 2.9858564693556837,
|
|
"grad_norm": 0.30344571122795916,
|
|
"learning_rate": 2.836097513565693e-05,
|
|
"loss": 0.086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04066718369722366,
|
|
"step": 5700,
|
|
"valid_targets_mean": 2583.9,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 2.9884756416972236,
|
|
"grad_norm": 0.29743190177468465,
|
|
"learning_rate": 2.833723692736893e-05,
|
|
"loss": 0.0896,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04501290246844292,
|
|
"step": 5705,
|
|
"valid_targets_mean": 3741.6,
|
|
"valid_targets_min": 605
|
|
},
|
|
{
|
|
"epoch": 2.991094814038764,
|
|
"grad_norm": 0.47781571969317105,
|
|
"learning_rate": 2.8313484495156486e-05,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16558252274990082,
|
|
"step": 5710,
|
|
"valid_targets_mean": 2971.0,
|
|
"valid_targets_min": 1907
|
|
},
|
|
{
|
|
"epoch": 2.9937139863803037,
|
|
"grad_norm": 0.3470372996331737,
|
|
"learning_rate": 2.828971787954295e-05,
|
|
"loss": 0.1148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044766783714294434,
|
|
"step": 5715,
|
|
"valid_targets_mean": 3138.5,
|
|
"valid_targets_min": 1235
|
|
},
|
|
{
|
|
"epoch": 2.996333158721844,
|
|
"grad_norm": 0.3702526905954634,
|
|
"learning_rate": 2.8265937121075873e-05,
|
|
"loss": 0.1012,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04486331343650818,
|
|
"step": 5720,
|
|
"valid_targets_mean": 2935.4,
|
|
"valid_targets_min": 759
|
|
},
|
|
{
|
|
"epoch": 2.998952331063384,
|
|
"grad_norm": 0.41766615047107897,
|
|
"learning_rate": 2.8242142260326937e-05,
|
|
"loss": 0.0931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06796953827142715,
|
|
"step": 5725,
|
|
"valid_targets_mean": 2050.0,
|
|
"valid_targets_min": 927
|
|
},
|
|
{
|
|
"epoch": 3.001571503404924,
|
|
"grad_norm": 0.7981101464361773,
|
|
"learning_rate": 2.8218333337891882e-05,
|
|
"loss": 0.1914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0790286436676979,
|
|
"step": 5730,
|
|
"valid_targets_mean": 1270.1,
|
|
"valid_targets_min": 788
|
|
},
|
|
{
|
|
"epoch": 3.004190675746464,
|
|
"grad_norm": 0.7560583152250143,
|
|
"learning_rate": 2.8194510394390443e-05,
|
|
"loss": 0.1857,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08779201656579971,
|
|
"step": 5735,
|
|
"valid_targets_mean": 1545.5,
|
|
"valid_targets_min": 957
|
|
},
|
|
{
|
|
"epoch": 3.0068098480880043,
|
|
"grad_norm": 0.7348656221096983,
|
|
"learning_rate": 2.817067347046626e-05,
|
|
"loss": 0.1755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08141259104013443,
|
|
"step": 5740,
|
|
"valid_targets_mean": 1351.6,
|
|
"valid_targets_min": 796
|
|
},
|
|
{
|
|
"epoch": 3.009429020429544,
|
|
"grad_norm": 0.7448213958029152,
|
|
"learning_rate": 2.814682260678684e-05,
|
|
"loss": 0.1734,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09365697205066681,
|
|
"step": 5745,
|
|
"valid_targets_mean": 1480.6,
|
|
"valid_targets_min": 839
|
|
},
|
|
{
|
|
"epoch": 3.0120481927710845,
|
|
"grad_norm": 0.7348628545274174,
|
|
"learning_rate": 2.8122957844043474e-05,
|
|
"loss": 0.1744,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07291596382856369,
|
|
"step": 5750,
|
|
"valid_targets_mean": 1344.9,
|
|
"valid_targets_min": 638
|
|
},
|
|
{
|
|
"epoch": 3.0146673651126243,
|
|
"grad_norm": 0.6685850834334908,
|
|
"learning_rate": 2.809907922295115e-05,
|
|
"loss": 0.1729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08419374376535416,
|
|
"step": 5755,
|
|
"valid_targets_mean": 1538.4,
|
|
"valid_targets_min": 727
|
|
},
|
|
{
|
|
"epoch": 3.0172865374541646,
|
|
"grad_norm": 0.6793100238003746,
|
|
"learning_rate": 2.8075186784248517e-05,
|
|
"loss": 0.1767,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07691366970539093,
|
|
"step": 5760,
|
|
"valid_targets_mean": 1195.4,
|
|
"valid_targets_min": 761
|
|
},
|
|
{
|
|
"epoch": 3.0199057097957045,
|
|
"grad_norm": 0.7629492371583007,
|
|
"learning_rate": 2.805128056869779e-05,
|
|
"loss": 0.1641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08975130319595337,
|
|
"step": 5765,
|
|
"valid_targets_mean": 1309.1,
|
|
"valid_targets_min": 714
|
|
},
|
|
{
|
|
"epoch": 3.022524882137245,
|
|
"grad_norm": 0.8394951597437683,
|
|
"learning_rate": 2.8027360617084684e-05,
|
|
"loss": 0.1846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08716186136007309,
|
|
"step": 5770,
|
|
"valid_targets_mean": 1417.9,
|
|
"valid_targets_min": 741
|
|
},
|
|
{
|
|
"epoch": 3.0251440544787846,
|
|
"grad_norm": 0.7196217010538367,
|
|
"learning_rate": 2.800342697021835e-05,
|
|
"loss": 0.1758,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10522545874118805,
|
|
"step": 5775,
|
|
"valid_targets_mean": 2031.6,
|
|
"valid_targets_min": 663
|
|
},
|
|
{
|
|
"epoch": 3.027763226820325,
|
|
"grad_norm": 0.6603324262189301,
|
|
"learning_rate": 2.797947966893131e-05,
|
|
"loss": 0.1673,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07629004120826721,
|
|
"step": 5780,
|
|
"valid_targets_mean": 1393.6,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 3.030382399161865,
|
|
"grad_norm": 0.6645598545760699,
|
|
"learning_rate": 2.7955518754079382e-05,
|
|
"loss": 0.1785,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09426233172416687,
|
|
"step": 5785,
|
|
"valid_targets_mean": 1895.8,
|
|
"valid_targets_min": 1009
|
|
},
|
|
{
|
|
"epoch": 3.033001571503405,
|
|
"grad_norm": 0.6434178337672452,
|
|
"learning_rate": 2.7931544266541596e-05,
|
|
"loss": 0.1687,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07660476118326187,
|
|
"step": 5790,
|
|
"valid_targets_mean": 1525.5,
|
|
"valid_targets_min": 717
|
|
},
|
|
{
|
|
"epoch": 3.035620743844945,
|
|
"grad_norm": 0.7319443766974159,
|
|
"learning_rate": 2.7907556247220166e-05,
|
|
"loss": 0.1608,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08354084193706512,
|
|
"step": 5795,
|
|
"valid_targets_mean": 1484.6,
|
|
"valid_targets_min": 659
|
|
},
|
|
{
|
|
"epoch": 3.0382399161864853,
|
|
"grad_norm": 0.6469303704614396,
|
|
"learning_rate": 2.788355473704036e-05,
|
|
"loss": 0.1608,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.064138263463974,
|
|
"step": 5800,
|
|
"valid_targets_mean": 1206.5,
|
|
"valid_targets_min": 642
|
|
},
|
|
{
|
|
"epoch": 3.040859088528025,
|
|
"grad_norm": 0.6479047117986019,
|
|
"learning_rate": 2.7859539776950478e-05,
|
|
"loss": 0.1673,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059673525393009186,
|
|
"step": 5805,
|
|
"valid_targets_mean": 1201.8,
|
|
"valid_targets_min": 765
|
|
},
|
|
{
|
|
"epoch": 3.0434782608695654,
|
|
"grad_norm": 0.7345623554093044,
|
|
"learning_rate": 2.7835511407921765e-05,
|
|
"loss": 0.1776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07721047103404999,
|
|
"step": 5810,
|
|
"valid_targets_mean": 1359.4,
|
|
"valid_targets_min": 631
|
|
},
|
|
{
|
|
"epoch": 3.0460974332111053,
|
|
"grad_norm": 0.7742188256489263,
|
|
"learning_rate": 2.7811469670948348e-05,
|
|
"loss": 0.1755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09204640984535217,
|
|
"step": 5815,
|
|
"valid_targets_mean": 1498.5,
|
|
"valid_targets_min": 623
|
|
},
|
|
{
|
|
"epoch": 3.0487166055526456,
|
|
"grad_norm": 0.7153340181131183,
|
|
"learning_rate": 2.778741460704715e-05,
|
|
"loss": 0.1711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11384181678295135,
|
|
"step": 5820,
|
|
"valid_targets_mean": 2016.6,
|
|
"valid_targets_min": 973
|
|
},
|
|
{
|
|
"epoch": 3.0513357778941854,
|
|
"grad_norm": 0.6562771495413292,
|
|
"learning_rate": 2.7763346257257843e-05,
|
|
"loss": 0.171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07798384130001068,
|
|
"step": 5825,
|
|
"valid_targets_mean": 1504.4,
|
|
"valid_targets_min": 776
|
|
},
|
|
{
|
|
"epoch": 3.0539549502357257,
|
|
"grad_norm": 0.7155882622011773,
|
|
"learning_rate": 2.7739264662642754e-05,
|
|
"loss": 0.1696,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10642405599355698,
|
|
"step": 5830,
|
|
"valid_targets_mean": 1658.8,
|
|
"valid_targets_min": 623
|
|
},
|
|
{
|
|
"epoch": 3.0565741225772656,
|
|
"grad_norm": 0.7226093969321024,
|
|
"learning_rate": 2.771516986428681e-05,
|
|
"loss": 0.173,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09716735780239105,
|
|
"step": 5835,
|
|
"valid_targets_mean": 1586.8,
|
|
"valid_targets_min": 1243
|
|
},
|
|
{
|
|
"epoch": 3.059193294918806,
|
|
"grad_norm": 0.7236637463167985,
|
|
"learning_rate": 2.7691061903297466e-05,
|
|
"loss": 0.17,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058030158281326294,
|
|
"step": 5840,
|
|
"valid_targets_mean": 1026.9,
|
|
"valid_targets_min": 611
|
|
},
|
|
{
|
|
"epoch": 3.0618124672603457,
|
|
"grad_norm": 0.7376730852977844,
|
|
"learning_rate": 2.7666940820804633e-05,
|
|
"loss": 0.1619,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0830594152212143,
|
|
"step": 5845,
|
|
"valid_targets_mean": 1498.2,
|
|
"valid_targets_min": 793
|
|
},
|
|
{
|
|
"epoch": 3.064431639601886,
|
|
"grad_norm": 0.6609142339695006,
|
|
"learning_rate": 2.7642806657960613e-05,
|
|
"loss": 0.1658,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.091240793466568,
|
|
"step": 5850,
|
|
"valid_targets_mean": 1812.6,
|
|
"valid_targets_min": 886
|
|
},
|
|
{
|
|
"epoch": 3.067050811943426,
|
|
"grad_norm": 0.7051693250614872,
|
|
"learning_rate": 2.761865945594002e-05,
|
|
"loss": 0.164,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08153796195983887,
|
|
"step": 5855,
|
|
"valid_targets_mean": 1643.8,
|
|
"valid_targets_min": 991
|
|
},
|
|
{
|
|
"epoch": 3.0696699842849657,
|
|
"grad_norm": 0.9082944258312199,
|
|
"learning_rate": 2.759449925593971e-05,
|
|
"loss": 0.1678,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10334397107362747,
|
|
"step": 5860,
|
|
"valid_targets_mean": 1784.2,
|
|
"valid_targets_min": 1107
|
|
},
|
|
{
|
|
"epoch": 3.072289156626506,
|
|
"grad_norm": 0.7095035432530095,
|
|
"learning_rate": 2.7570326099178716e-05,
|
|
"loss": 0.1761,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0884983241558075,
|
|
"step": 5865,
|
|
"valid_targets_mean": 1524.6,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 3.074908328968046,
|
|
"grad_norm": 0.6555581714051659,
|
|
"learning_rate": 2.754614002689818e-05,
|
|
"loss": 0.1599,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1063360646367073,
|
|
"step": 5870,
|
|
"valid_targets_mean": 1774.9,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 3.077527501309586,
|
|
"grad_norm": 0.735815034221504,
|
|
"learning_rate": 2.7521941080361275e-05,
|
|
"loss": 0.1689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09698130190372467,
|
|
"step": 5875,
|
|
"valid_targets_mean": 1748.0,
|
|
"valid_targets_min": 937
|
|
},
|
|
{
|
|
"epoch": 3.080146673651126,
|
|
"grad_norm": 0.7125484974425914,
|
|
"learning_rate": 2.749772930085315e-05,
|
|
"loss": 0.1692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09398886561393738,
|
|
"step": 5880,
|
|
"valid_targets_mean": 1739.2,
|
|
"valid_targets_min": 804
|
|
},
|
|
{
|
|
"epoch": 3.0827658459926663,
|
|
"grad_norm": 0.9696180961709159,
|
|
"learning_rate": 2.7473504729680823e-05,
|
|
"loss": 0.1715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06936175376176834,
|
|
"step": 5885,
|
|
"valid_targets_mean": 1306.5,
|
|
"valid_targets_min": 675
|
|
},
|
|
{
|
|
"epoch": 3.085385018334206,
|
|
"grad_norm": 0.7256259095947776,
|
|
"learning_rate": 2.744926740817316e-05,
|
|
"loss": 0.1635,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07942957431077957,
|
|
"step": 5890,
|
|
"valid_targets_mean": 1250.9,
|
|
"valid_targets_min": 882
|
|
},
|
|
{
|
|
"epoch": 3.0880041906757465,
|
|
"grad_norm": 0.6750034166270854,
|
|
"learning_rate": 2.742501737768077e-05,
|
|
"loss": 0.1632,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08731904625892639,
|
|
"step": 5895,
|
|
"valid_targets_mean": 1721.5,
|
|
"valid_targets_min": 1116
|
|
},
|
|
{
|
|
"epoch": 3.0906233630172864,
|
|
"grad_norm": 0.7339452739226756,
|
|
"learning_rate": 2.7400754679575948e-05,
|
|
"loss": 0.1528,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06412196904420853,
|
|
"step": 5900,
|
|
"valid_targets_mean": 1218.5,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 3.0932425353588267,
|
|
"grad_norm": 0.7104961468611212,
|
|
"learning_rate": 2.7376479355252588e-05,
|
|
"loss": 0.1689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06227774918079376,
|
|
"step": 5905,
|
|
"valid_targets_mean": 1085.4,
|
|
"valid_targets_min": 676
|
|
},
|
|
{
|
|
"epoch": 3.0958617077003665,
|
|
"grad_norm": 0.6733244215312041,
|
|
"learning_rate": 2.7352191446126145e-05,
|
|
"loss": 0.1653,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07951797544956207,
|
|
"step": 5910,
|
|
"valid_targets_mean": 1521.0,
|
|
"valid_targets_min": 849
|
|
},
|
|
{
|
|
"epoch": 3.098480880041907,
|
|
"grad_norm": 0.6663067364300812,
|
|
"learning_rate": 2.732789099363353e-05,
|
|
"loss": 0.1643,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08559563755989075,
|
|
"step": 5915,
|
|
"valid_targets_mean": 1485.2,
|
|
"valid_targets_min": 742
|
|
},
|
|
{
|
|
"epoch": 3.1011000523834467,
|
|
"grad_norm": 0.709938600907913,
|
|
"learning_rate": 2.7303578039233055e-05,
|
|
"loss": 0.1636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10190951079130173,
|
|
"step": 5920,
|
|
"valid_targets_mean": 1642.5,
|
|
"valid_targets_min": 1077
|
|
},
|
|
{
|
|
"epoch": 3.103719224724987,
|
|
"grad_norm": 0.7754935357002035,
|
|
"learning_rate": 2.7279252624404374e-05,
|
|
"loss": 0.1727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1033768430352211,
|
|
"step": 5925,
|
|
"valid_targets_mean": 1766.6,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 3.106338397066527,
|
|
"grad_norm": 0.7090344892805982,
|
|
"learning_rate": 2.7254914790648387e-05,
|
|
"loss": 0.1547,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08158548176288605,
|
|
"step": 5930,
|
|
"valid_targets_mean": 1349.6,
|
|
"valid_targets_min": 689
|
|
},
|
|
{
|
|
"epoch": 3.108957569408067,
|
|
"grad_norm": 0.7331205271019855,
|
|
"learning_rate": 2.723056457948718e-05,
|
|
"loss": 0.1735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08869096636772156,
|
|
"step": 5935,
|
|
"valid_targets_mean": 1466.2,
|
|
"valid_targets_min": 768
|
|
},
|
|
{
|
|
"epoch": 3.111576741749607,
|
|
"grad_norm": 0.7026004684688643,
|
|
"learning_rate": 2.7206202032463963e-05,
|
|
"loss": 0.1669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07913877815008163,
|
|
"step": 5940,
|
|
"valid_targets_mean": 1471.5,
|
|
"valid_targets_min": 1058
|
|
},
|
|
{
|
|
"epoch": 3.1141959140911473,
|
|
"grad_norm": 0.6928949635881819,
|
|
"learning_rate": 2.7181827191142988e-05,
|
|
"loss": 0.1778,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08693772554397583,
|
|
"step": 5945,
|
|
"valid_targets_mean": 1477.9,
|
|
"valid_targets_min": 618
|
|
},
|
|
{
|
|
"epoch": 3.116815086432687,
|
|
"grad_norm": 0.7072944606587698,
|
|
"learning_rate": 2.7157440097109496e-05,
|
|
"loss": 0.1586,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08366671204566956,
|
|
"step": 5950,
|
|
"valid_targets_mean": 1510.1,
|
|
"valid_targets_min": 629
|
|
},
|
|
{
|
|
"epoch": 3.1194342587742274,
|
|
"grad_norm": 0.7155777290977046,
|
|
"learning_rate": 2.7133040791969597e-05,
|
|
"loss": 0.165,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0935017466545105,
|
|
"step": 5955,
|
|
"valid_targets_mean": 1804.2,
|
|
"valid_targets_min": 1167
|
|
},
|
|
{
|
|
"epoch": 3.1220534311157673,
|
|
"grad_norm": 0.677362020785958,
|
|
"learning_rate": 2.7108629317350278e-05,
|
|
"loss": 0.1669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08500826358795166,
|
|
"step": 5960,
|
|
"valid_targets_mean": 1497.8,
|
|
"valid_targets_min": 699
|
|
},
|
|
{
|
|
"epoch": 3.1246726034573076,
|
|
"grad_norm": 0.7286002740634437,
|
|
"learning_rate": 2.7084205714899247e-05,
|
|
"loss": 0.1582,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06782686710357666,
|
|
"step": 5965,
|
|
"valid_targets_mean": 1328.5,
|
|
"valid_targets_min": 679
|
|
},
|
|
{
|
|
"epoch": 3.1272917757988474,
|
|
"grad_norm": 0.7021339243865264,
|
|
"learning_rate": 2.705977002628494e-05,
|
|
"loss": 0.1596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08048438280820847,
|
|
"step": 5970,
|
|
"valid_targets_mean": 1501.8,
|
|
"valid_targets_min": 504
|
|
},
|
|
{
|
|
"epoch": 3.1299109481403877,
|
|
"grad_norm": 0.6830143227137973,
|
|
"learning_rate": 2.703532229319638e-05,
|
|
"loss": 0.1643,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0930556207895279,
|
|
"step": 5975,
|
|
"valid_targets_mean": 1900.4,
|
|
"valid_targets_min": 1331
|
|
},
|
|
{
|
|
"epoch": 3.1325301204819276,
|
|
"grad_norm": 0.6861135630337549,
|
|
"learning_rate": 2.7010862557343168e-05,
|
|
"loss": 0.1643,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04893141984939575,
|
|
"step": 5980,
|
|
"valid_targets_mean": 1172.0,
|
|
"valid_targets_min": 661
|
|
},
|
|
{
|
|
"epoch": 3.135149292823468,
|
|
"grad_norm": 0.6776084061989935,
|
|
"learning_rate": 2.698639086045536e-05,
|
|
"loss": 0.1694,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06864239275455475,
|
|
"step": 5985,
|
|
"valid_targets_mean": 1402.1,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 3.1377684651650077,
|
|
"grad_norm": 0.6443724306278145,
|
|
"learning_rate": 2.6961907244283424e-05,
|
|
"loss": 0.1564,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09214092791080475,
|
|
"step": 5990,
|
|
"valid_targets_mean": 1866.2,
|
|
"valid_targets_min": 963
|
|
},
|
|
{
|
|
"epoch": 3.140387637506548,
|
|
"grad_norm": 0.6757253945114164,
|
|
"learning_rate": 2.6937411750598172e-05,
|
|
"loss": 0.1657,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06545227020978928,
|
|
"step": 5995,
|
|
"valid_targets_mean": 1191.1,
|
|
"valid_targets_min": 793
|
|
},
|
|
{
|
|
"epoch": 3.143006809848088,
|
|
"grad_norm": 0.7973697455117805,
|
|
"learning_rate": 2.6912904421190672e-05,
|
|
"loss": 0.1669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11898652464151382,
|
|
"step": 6000,
|
|
"valid_targets_mean": 1563.2,
|
|
"valid_targets_min": 917
|
|
},
|
|
{
|
|
"epoch": 3.145625982189628,
|
|
"grad_norm": 0.722464004593059,
|
|
"learning_rate": 2.6888385297872194e-05,
|
|
"loss": 0.1629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06986261904239655,
|
|
"step": 6005,
|
|
"valid_targets_mean": 1274.4,
|
|
"valid_targets_min": 752
|
|
},
|
|
{
|
|
"epoch": 3.148245154531168,
|
|
"grad_norm": 0.6946891567070904,
|
|
"learning_rate": 2.6863854422474108e-05,
|
|
"loss": 0.1647,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10306160897016525,
|
|
"step": 6010,
|
|
"valid_targets_mean": 1790.4,
|
|
"valid_targets_min": 1220
|
|
},
|
|
{
|
|
"epoch": 3.1508643268727083,
|
|
"grad_norm": 0.6969095258678409,
|
|
"learning_rate": 2.6839311836847856e-05,
|
|
"loss": 0.1516,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09284568578004837,
|
|
"step": 6015,
|
|
"valid_targets_mean": 1727.0,
|
|
"valid_targets_min": 953
|
|
},
|
|
{
|
|
"epoch": 3.153483499214248,
|
|
"grad_norm": 0.7707142300486267,
|
|
"learning_rate": 2.6814757582864847e-05,
|
|
"loss": 0.1582,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08388621360063553,
|
|
"step": 6020,
|
|
"valid_targets_mean": 1370.0,
|
|
"valid_targets_min": 639
|
|
},
|
|
{
|
|
"epoch": 3.1561026715557885,
|
|
"grad_norm": 0.666834244973884,
|
|
"learning_rate": 2.6790191702416405e-05,
|
|
"loss": 0.1704,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08771908283233643,
|
|
"step": 6025,
|
|
"valid_targets_mean": 1918.8,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 3.1587218438973284,
|
|
"grad_norm": 0.7102588320192474,
|
|
"learning_rate": 2.676561423741368e-05,
|
|
"loss": 0.1574,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07976895570755005,
|
|
"step": 6030,
|
|
"valid_targets_mean": 1233.9,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 3.1613410162388687,
|
|
"grad_norm": 0.6906103632403421,
|
|
"learning_rate": 2.6741025229787596e-05,
|
|
"loss": 0.1664,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08830182254314423,
|
|
"step": 6035,
|
|
"valid_targets_mean": 1608.4,
|
|
"valid_targets_min": 936
|
|
},
|
|
{
|
|
"epoch": 3.1639601885804085,
|
|
"grad_norm": 0.6967305793043933,
|
|
"learning_rate": 2.6716424721488752e-05,
|
|
"loss": 0.1628,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06965943425893784,
|
|
"step": 6040,
|
|
"valid_targets_mean": 1455.1,
|
|
"valid_targets_min": 902
|
|
},
|
|
{
|
|
"epoch": 3.166579360921949,
|
|
"grad_norm": 0.7506317658861008,
|
|
"learning_rate": 2.6691812754487395e-05,
|
|
"loss": 0.1675,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07598876953125,
|
|
"step": 6045,
|
|
"valid_targets_mean": 1266.8,
|
|
"valid_targets_min": 986
|
|
},
|
|
{
|
|
"epoch": 3.1691985332634887,
|
|
"grad_norm": 0.7362679513515235,
|
|
"learning_rate": 2.6667189370773302e-05,
|
|
"loss": 0.1637,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08343760669231415,
|
|
"step": 6050,
|
|
"valid_targets_mean": 1488.4,
|
|
"valid_targets_min": 1079
|
|
},
|
|
{
|
|
"epoch": 3.171817705605029,
|
|
"grad_norm": 0.6744924339491063,
|
|
"learning_rate": 2.664255461235573e-05,
|
|
"loss": 0.1608,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07914218306541443,
|
|
"step": 6055,
|
|
"valid_targets_mean": 1512.0,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 3.174436877946569,
|
|
"grad_norm": 0.7327687496774006,
|
|
"learning_rate": 2.6617908521263344e-05,
|
|
"loss": 0.1675,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08062729239463806,
|
|
"step": 6060,
|
|
"valid_targets_mean": 1439.5,
|
|
"valid_targets_min": 655
|
|
},
|
|
{
|
|
"epoch": 3.177056050288109,
|
|
"grad_norm": 0.7418671205008001,
|
|
"learning_rate": 2.659325113954415e-05,
|
|
"loss": 0.163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06643374264240265,
|
|
"step": 6065,
|
|
"valid_targets_mean": 1297.6,
|
|
"valid_targets_min": 667
|
|
},
|
|
{
|
|
"epoch": 3.179675222629649,
|
|
"grad_norm": 0.7196932247817096,
|
|
"learning_rate": 2.6568582509265403e-05,
|
|
"loss": 0.1712,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0984589010477066,
|
|
"step": 6070,
|
|
"valid_targets_mean": 1918.1,
|
|
"valid_targets_min": 1041
|
|
},
|
|
{
|
|
"epoch": 3.1822943949711893,
|
|
"grad_norm": 0.7471858549275774,
|
|
"learning_rate": 2.6543902672513562e-05,
|
|
"loss": 0.1682,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.062356751412153244,
|
|
"step": 6075,
|
|
"valid_targets_mean": 1216.2,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 3.184913567312729,
|
|
"grad_norm": 0.6739018916560132,
|
|
"learning_rate": 2.6519211671394196e-05,
|
|
"loss": 0.1662,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08063490688800812,
|
|
"step": 6080,
|
|
"valid_targets_mean": 1757.2,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 3.1875327396542694,
|
|
"grad_norm": 0.7460699909826294,
|
|
"learning_rate": 2.6494509548031926e-05,
|
|
"loss": 0.1579,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0652986541390419,
|
|
"step": 6085,
|
|
"valid_targets_mean": 1021.8,
|
|
"valid_targets_min": 620
|
|
},
|
|
{
|
|
"epoch": 3.1901519119958093,
|
|
"grad_norm": 0.691870261331958,
|
|
"learning_rate": 2.6469796344570348e-05,
|
|
"loss": 0.1616,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09676294028759003,
|
|
"step": 6090,
|
|
"valid_targets_mean": 1853.4,
|
|
"valid_targets_min": 656
|
|
},
|
|
{
|
|
"epoch": 3.1927710843373496,
|
|
"grad_norm": 0.7070404191392444,
|
|
"learning_rate": 2.6445072103171952e-05,
|
|
"loss": 0.1657,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06507851183414459,
|
|
"step": 6095,
|
|
"valid_targets_mean": 1437.9,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 3.1953902566788894,
|
|
"grad_norm": 0.7294314865976155,
|
|
"learning_rate": 2.6420336866018087e-05,
|
|
"loss": 0.1599,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08893925696611404,
|
|
"step": 6100,
|
|
"valid_targets_mean": 1716.4,
|
|
"valid_targets_min": 816
|
|
},
|
|
{
|
|
"epoch": 3.1980094290204297,
|
|
"grad_norm": 0.6705758424773739,
|
|
"learning_rate": 2.6395590675308826e-05,
|
|
"loss": 0.1553,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07986105233430862,
|
|
"step": 6105,
|
|
"valid_targets_mean": 1542.0,
|
|
"valid_targets_min": 959
|
|
},
|
|
{
|
|
"epoch": 3.2006286013619696,
|
|
"grad_norm": 0.7027299991646958,
|
|
"learning_rate": 2.6370833573262946e-05,
|
|
"loss": 0.1536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06690247356891632,
|
|
"step": 6110,
|
|
"valid_targets_mean": 1193.2,
|
|
"valid_targets_min": 747
|
|
},
|
|
{
|
|
"epoch": 3.20324777370351,
|
|
"grad_norm": 0.7354964208411295,
|
|
"learning_rate": 2.634606560211785e-05,
|
|
"loss": 0.172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07954400777816772,
|
|
"step": 6115,
|
|
"valid_targets_mean": 1542.4,
|
|
"valid_targets_min": 933
|
|
},
|
|
{
|
|
"epoch": 3.2058669460450497,
|
|
"grad_norm": 0.7110894716658694,
|
|
"learning_rate": 2.6321286804129465e-05,
|
|
"loss": 0.161,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08626635372638702,
|
|
"step": 6120,
|
|
"valid_targets_mean": 1575.0,
|
|
"valid_targets_min": 827
|
|
},
|
|
{
|
|
"epoch": 3.20848611838659,
|
|
"grad_norm": 0.6549484866958899,
|
|
"learning_rate": 2.62964972215722e-05,
|
|
"loss": 0.155,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06384006142616272,
|
|
"step": 6125,
|
|
"valid_targets_mean": 1332.8,
|
|
"valid_targets_min": 741
|
|
},
|
|
{
|
|
"epoch": 3.21110529072813,
|
|
"grad_norm": 0.6375353110419558,
|
|
"learning_rate": 2.6271696896738874e-05,
|
|
"loss": 0.1563,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06473011523485184,
|
|
"step": 6130,
|
|
"valid_targets_mean": 1391.2,
|
|
"valid_targets_min": 639
|
|
},
|
|
{
|
|
"epoch": 3.21372446306967,
|
|
"grad_norm": 0.7257128878637935,
|
|
"learning_rate": 2.62468858719406e-05,
|
|
"loss": 0.1481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06425201892852783,
|
|
"step": 6135,
|
|
"valid_targets_mean": 1068.9,
|
|
"valid_targets_min": 743
|
|
},
|
|
{
|
|
"epoch": 3.21634363541121,
|
|
"grad_norm": 0.6736038952728773,
|
|
"learning_rate": 2.6222064189506782e-05,
|
|
"loss": 0.163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06724032014608383,
|
|
"step": 6140,
|
|
"valid_targets_mean": 1378.8,
|
|
"valid_targets_min": 610
|
|
},
|
|
{
|
|
"epoch": 3.2189628077527503,
|
|
"grad_norm": 0.6862167359194624,
|
|
"learning_rate": 2.619723189178498e-05,
|
|
"loss": 0.1582,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07766151428222656,
|
|
"step": 6145,
|
|
"valid_targets_mean": 1417.6,
|
|
"valid_targets_min": 695
|
|
},
|
|
{
|
|
"epoch": 3.22158198009429,
|
|
"grad_norm": 0.6667822521852074,
|
|
"learning_rate": 2.6172389021140886e-05,
|
|
"loss": 0.1645,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10055532306432724,
|
|
"step": 6150,
|
|
"valid_targets_mean": 1966.2,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 3.22420115243583,
|
|
"grad_norm": 0.6843734767884763,
|
|
"learning_rate": 2.6147535619958216e-05,
|
|
"loss": 0.1602,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07212713360786438,
|
|
"step": 6155,
|
|
"valid_targets_mean": 1374.0,
|
|
"valid_targets_min": 765
|
|
},
|
|
{
|
|
"epoch": 3.2268203247773704,
|
|
"grad_norm": 0.7478392776507355,
|
|
"learning_rate": 2.6122671730638657e-05,
|
|
"loss": 0.1606,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08449780195951462,
|
|
"step": 6160,
|
|
"valid_targets_mean": 1407.0,
|
|
"valid_targets_min": 816
|
|
},
|
|
{
|
|
"epoch": 3.2294394971189107,
|
|
"grad_norm": 0.6873028289484555,
|
|
"learning_rate": 2.6097797395601782e-05,
|
|
"loss": 0.1631,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07395478338003159,
|
|
"step": 6165,
|
|
"valid_targets_mean": 1474.1,
|
|
"valid_targets_min": 1054
|
|
},
|
|
{
|
|
"epoch": 3.2320586694604505,
|
|
"grad_norm": 0.4776467763544407,
|
|
"learning_rate": 2.6072912657285002e-05,
|
|
"loss": 0.1818,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0907076895236969,
|
|
"step": 6170,
|
|
"valid_targets_mean": 5532.4,
|
|
"valid_targets_min": 4247
|
|
},
|
|
{
|
|
"epoch": 3.2346778418019904,
|
|
"grad_norm": 0.3732618307241714,
|
|
"learning_rate": 2.6048017558143462e-05,
|
|
"loss": 0.1792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09417170286178589,
|
|
"step": 6175,
|
|
"valid_targets_mean": 6763.9,
|
|
"valid_targets_min": 4329
|
|
},
|
|
{
|
|
"epoch": 3.2372970141435307,
|
|
"grad_norm": 0.38289420746438846,
|
|
"learning_rate": 2.602311214064998e-05,
|
|
"loss": 0.1681,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09114161133766174,
|
|
"step": 6180,
|
|
"valid_targets_mean": 5724.6,
|
|
"valid_targets_min": 4480
|
|
},
|
|
{
|
|
"epoch": 3.2399161864850705,
|
|
"grad_norm": 0.3836635906229439,
|
|
"learning_rate": 2.5998196447294996e-05,
|
|
"loss": 0.1735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0969650000333786,
|
|
"step": 6185,
|
|
"valid_targets_mean": 5883.4,
|
|
"valid_targets_min": 4826
|
|
},
|
|
{
|
|
"epoch": 3.242535358826611,
|
|
"grad_norm": 0.3975060540969869,
|
|
"learning_rate": 2.5973270520586465e-05,
|
|
"loss": 0.1781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09373161196708679,
|
|
"step": 6190,
|
|
"valid_targets_mean": 5766.0,
|
|
"valid_targets_min": 3983
|
|
},
|
|
{
|
|
"epoch": 3.2451545311681507,
|
|
"grad_norm": 0.37385141092856894,
|
|
"learning_rate": 2.5948334403049806e-05,
|
|
"loss": 0.1756,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08556351065635681,
|
|
"step": 6195,
|
|
"valid_targets_mean": 5980.8,
|
|
"valid_targets_min": 4998
|
|
},
|
|
{
|
|
"epoch": 3.247773703509691,
|
|
"grad_norm": 0.37349177636254594,
|
|
"learning_rate": 2.5923388137227825e-05,
|
|
"loss": 0.1826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08874762803316116,
|
|
"step": 6200,
|
|
"valid_targets_mean": 5429.8,
|
|
"valid_targets_min": 4907
|
|
},
|
|
{
|
|
"epoch": 3.250392875851231,
|
|
"grad_norm": 0.4114694801622512,
|
|
"learning_rate": 2.5898431765680644e-05,
|
|
"loss": 0.1681,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11268821358680725,
|
|
"step": 6205,
|
|
"valid_targets_mean": 6900.6,
|
|
"valid_targets_min": 5066
|
|
},
|
|
{
|
|
"epoch": 3.253012048192771,
|
|
"grad_norm": 0.3587170860139122,
|
|
"learning_rate": 2.5873465330985625e-05,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08456286787986755,
|
|
"step": 6210,
|
|
"valid_targets_mean": 7580.8,
|
|
"valid_targets_min": 4500
|
|
},
|
|
{
|
|
"epoch": 3.255631220534311,
|
|
"grad_norm": 0.35279536697085795,
|
|
"learning_rate": 2.5848488875737293e-05,
|
|
"loss": 0.1711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08426110446453094,
|
|
"step": 6215,
|
|
"valid_targets_mean": 7422.6,
|
|
"valid_targets_min": 5455
|
|
},
|
|
{
|
|
"epoch": 3.2582503928758513,
|
|
"grad_norm": 0.35830411133344736,
|
|
"learning_rate": 2.582350244254728e-05,
|
|
"loss": 0.1814,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08689013123512268,
|
|
"step": 6220,
|
|
"valid_targets_mean": 7564.4,
|
|
"valid_targets_min": 4236
|
|
},
|
|
{
|
|
"epoch": 3.260869565217391,
|
|
"grad_norm": 0.34967856809357584,
|
|
"learning_rate": 2.5798506074044225e-05,
|
|
"loss": 0.1622,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08358441293239594,
|
|
"step": 6225,
|
|
"valid_targets_mean": 6671.9,
|
|
"valid_targets_min": 5315
|
|
},
|
|
{
|
|
"epoch": 3.2634887375589314,
|
|
"grad_norm": 0.35882018746504213,
|
|
"learning_rate": 2.577349981287373e-05,
|
|
"loss": 0.177,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08513113856315613,
|
|
"step": 6230,
|
|
"valid_targets_mean": 6717.1,
|
|
"valid_targets_min": 4353
|
|
},
|
|
{
|
|
"epoch": 3.2661079099004713,
|
|
"grad_norm": 0.35269541833504336,
|
|
"learning_rate": 2.5748483701698268e-05,
|
|
"loss": 0.1558,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0673292949795723,
|
|
"step": 6235,
|
|
"valid_targets_mean": 5959.0,
|
|
"valid_targets_min": 4890
|
|
},
|
|
{
|
|
"epoch": 3.2687270822420116,
|
|
"grad_norm": 0.3991951759169992,
|
|
"learning_rate": 2.5723457783197123e-05,
|
|
"loss": 0.1583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06948716938495636,
|
|
"step": 6240,
|
|
"valid_targets_mean": 5842.4,
|
|
"valid_targets_min": 4526
|
|
},
|
|
{
|
|
"epoch": 3.2713462545835514,
|
|
"grad_norm": 0.3992626001867334,
|
|
"learning_rate": 2.569842210006631e-05,
|
|
"loss": 0.1743,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09178012609481812,
|
|
"step": 6245,
|
|
"valid_targets_mean": 5798.9,
|
|
"valid_targets_min": 4793
|
|
},
|
|
{
|
|
"epoch": 3.2739654269250917,
|
|
"grad_norm": 0.4381857179992536,
|
|
"learning_rate": 2.56733766950185e-05,
|
|
"loss": 0.1749,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08497138321399689,
|
|
"step": 6250,
|
|
"valid_targets_mean": 6176.4,
|
|
"valid_targets_min": 3967
|
|
},
|
|
{
|
|
"epoch": 3.2765845992666316,
|
|
"grad_norm": 0.38177694056915856,
|
|
"learning_rate": 2.5648321610782943e-05,
|
|
"loss": 0.1518,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08642652630805969,
|
|
"step": 6255,
|
|
"valid_targets_mean": 6440.4,
|
|
"valid_targets_min": 5009
|
|
},
|
|
{
|
|
"epoch": 3.279203771608172,
|
|
"grad_norm": 0.385214253629684,
|
|
"learning_rate": 2.5623256890105422e-05,
|
|
"loss": 0.1568,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07483935356140137,
|
|
"step": 6260,
|
|
"valid_targets_mean": 5954.2,
|
|
"valid_targets_min": 4703
|
|
},
|
|
{
|
|
"epoch": 3.2818229439497117,
|
|
"grad_norm": 0.41522487533898733,
|
|
"learning_rate": 2.559818257574814e-05,
|
|
"loss": 0.1597,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07940084487199783,
|
|
"step": 6265,
|
|
"valid_targets_mean": 5829.8,
|
|
"valid_targets_min": 4385
|
|
},
|
|
{
|
|
"epoch": 3.284442116291252,
|
|
"grad_norm": 0.3264798440861103,
|
|
"learning_rate": 2.5573098710489683e-05,
|
|
"loss": 0.1849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08041777461767197,
|
|
"step": 6270,
|
|
"valid_targets_mean": 7655.0,
|
|
"valid_targets_min": 5133
|
|
},
|
|
{
|
|
"epoch": 3.287061288632792,
|
|
"grad_norm": 0.35577133900676294,
|
|
"learning_rate": 2.5548005337124924e-05,
|
|
"loss": 0.1673,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08044205605983734,
|
|
"step": 6275,
|
|
"valid_targets_mean": 6274.6,
|
|
"valid_targets_min": 4792
|
|
},
|
|
{
|
|
"epoch": 3.289680460974332,
|
|
"grad_norm": 0.3676349884831182,
|
|
"learning_rate": 2.552290249846496e-05,
|
|
"loss": 0.1662,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07401085644960403,
|
|
"step": 6280,
|
|
"valid_targets_mean": 6152.8,
|
|
"valid_targets_min": 3742
|
|
},
|
|
{
|
|
"epoch": 3.292299633315872,
|
|
"grad_norm": 0.367132421344028,
|
|
"learning_rate": 2.5497790237337024e-05,
|
|
"loss": 0.1776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09174760431051254,
|
|
"step": 6285,
|
|
"valid_targets_mean": 7111.8,
|
|
"valid_targets_min": 4062
|
|
},
|
|
{
|
|
"epoch": 3.2949188056574124,
|
|
"grad_norm": 0.3723449977990775,
|
|
"learning_rate": 2.5472668596584448e-05,
|
|
"loss": 0.1928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10043409466743469,
|
|
"step": 6290,
|
|
"valid_targets_mean": 6711.2,
|
|
"valid_targets_min": 4743
|
|
},
|
|
{
|
|
"epoch": 3.297537977998952,
|
|
"grad_norm": 0.37632408759519975,
|
|
"learning_rate": 2.5447537619066543e-05,
|
|
"loss": 0.2007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10895048826932907,
|
|
"step": 6295,
|
|
"valid_targets_mean": 7476.6,
|
|
"valid_targets_min": 5764
|
|
},
|
|
{
|
|
"epoch": 3.3001571503404925,
|
|
"grad_norm": 0.3641846475597183,
|
|
"learning_rate": 2.5422397347658575e-05,
|
|
"loss": 0.1763,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08448459208011627,
|
|
"step": 6300,
|
|
"valid_targets_mean": 6018.8,
|
|
"valid_targets_min": 4789
|
|
},
|
|
{
|
|
"epoch": 3.3027763226820324,
|
|
"grad_norm": 0.3750578971526196,
|
|
"learning_rate": 2.539724782525164e-05,
|
|
"loss": 0.1751,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08280646055936813,
|
|
"step": 6305,
|
|
"valid_targets_mean": 6307.5,
|
|
"valid_targets_min": 4828
|
|
},
|
|
{
|
|
"epoch": 3.3053954950235727,
|
|
"grad_norm": 0.3903161585255933,
|
|
"learning_rate": 2.537208909475263e-05,
|
|
"loss": 0.1715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0846589058637619,
|
|
"step": 6310,
|
|
"valid_targets_mean": 6349.4,
|
|
"valid_targets_min": 4671
|
|
},
|
|
{
|
|
"epoch": 3.3080146673651125,
|
|
"grad_norm": 0.363785247862675,
|
|
"learning_rate": 2.534692119908414e-05,
|
|
"loss": 0.1805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10132154822349548,
|
|
"step": 6315,
|
|
"valid_targets_mean": 7190.0,
|
|
"valid_targets_min": 3850
|
|
},
|
|
{
|
|
"epoch": 3.310633839706653,
|
|
"grad_norm": 0.3708807504191291,
|
|
"learning_rate": 2.532174418118442e-05,
|
|
"loss": 0.1691,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09385242313146591,
|
|
"step": 6320,
|
|
"valid_targets_mean": 6765.5,
|
|
"valid_targets_min": 4951
|
|
},
|
|
{
|
|
"epoch": 3.3132530120481927,
|
|
"grad_norm": 0.3752398723516199,
|
|
"learning_rate": 2.5296558084007253e-05,
|
|
"loss": 0.1778,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09168010950088501,
|
|
"step": 6325,
|
|
"valid_targets_mean": 8050.8,
|
|
"valid_targets_min": 4794
|
|
},
|
|
{
|
|
"epoch": 3.315872184389733,
|
|
"grad_norm": 0.42307307827222046,
|
|
"learning_rate": 2.5271362950521935e-05,
|
|
"loss": 0.178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09137772023677826,
|
|
"step": 6330,
|
|
"valid_targets_mean": 5186.9,
|
|
"valid_targets_min": 593
|
|
},
|
|
{
|
|
"epoch": 3.318491356731273,
|
|
"grad_norm": 0.3322795500595901,
|
|
"learning_rate": 2.5246158823713176e-05,
|
|
"loss": 0.1707,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07264706492424011,
|
|
"step": 6335,
|
|
"valid_targets_mean": 7296.5,
|
|
"valid_targets_min": 4633
|
|
},
|
|
{
|
|
"epoch": 3.321110529072813,
|
|
"grad_norm": 0.38299965795851554,
|
|
"learning_rate": 2.5220945746581014e-05,
|
|
"loss": 0.1528,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07251882553100586,
|
|
"step": 6340,
|
|
"valid_targets_mean": 6049.0,
|
|
"valid_targets_min": 5031
|
|
},
|
|
{
|
|
"epoch": 3.323729701414353,
|
|
"grad_norm": 0.41932734607392075,
|
|
"learning_rate": 2.5195723762140777e-05,
|
|
"loss": 0.1661,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08227471262216568,
|
|
"step": 6345,
|
|
"valid_targets_mean": 6290.0,
|
|
"valid_targets_min": 3987
|
|
},
|
|
{
|
|
"epoch": 3.3263488737558933,
|
|
"grad_norm": 0.404682250819354,
|
|
"learning_rate": 2.5170492913422973e-05,
|
|
"loss": 0.1794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08566948771476746,
|
|
"step": 6350,
|
|
"valid_targets_mean": 6865.1,
|
|
"valid_targets_min": 5568
|
|
},
|
|
{
|
|
"epoch": 3.328968046097433,
|
|
"grad_norm": 0.4011956473179966,
|
|
"learning_rate": 2.514525324347325e-05,
|
|
"loss": 0.1775,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09244292974472046,
|
|
"step": 6355,
|
|
"valid_targets_mean": 6699.5,
|
|
"valid_targets_min": 4839
|
|
},
|
|
{
|
|
"epoch": 3.3315872184389734,
|
|
"grad_norm": 0.4092538401932468,
|
|
"learning_rate": 2.5120004795352287e-05,
|
|
"loss": 0.163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07024848461151123,
|
|
"step": 6360,
|
|
"valid_targets_mean": 4793.9,
|
|
"valid_targets_min": 3517
|
|
},
|
|
{
|
|
"epoch": 3.3342063907805133,
|
|
"grad_norm": 0.3919155985895916,
|
|
"learning_rate": 2.509474761213576e-05,
|
|
"loss": 0.1621,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08399725705385208,
|
|
"step": 6365,
|
|
"valid_targets_mean": 6496.4,
|
|
"valid_targets_min": 4150
|
|
},
|
|
{
|
|
"epoch": 3.3368255631220536,
|
|
"grad_norm": 0.3956119853167681,
|
|
"learning_rate": 2.5069481736914218e-05,
|
|
"loss": 0.1712,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07363933324813843,
|
|
"step": 6370,
|
|
"valid_targets_mean": 5308.6,
|
|
"valid_targets_min": 4497
|
|
},
|
|
{
|
|
"epoch": 3.3394447354635934,
|
|
"grad_norm": 0.3820452027487794,
|
|
"learning_rate": 2.5044207212793073e-05,
|
|
"loss": 0.1754,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09037287533283234,
|
|
"step": 6375,
|
|
"valid_targets_mean": 7091.4,
|
|
"valid_targets_min": 6151
|
|
},
|
|
{
|
|
"epoch": 3.3420639078051337,
|
|
"grad_norm": 0.3997129999370939,
|
|
"learning_rate": 2.5018924082892465e-05,
|
|
"loss": 0.1641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08182957768440247,
|
|
"step": 6380,
|
|
"valid_targets_mean": 5894.9,
|
|
"valid_targets_min": 4661
|
|
},
|
|
{
|
|
"epoch": 3.3446830801466736,
|
|
"grad_norm": 0.4120300525624395,
|
|
"learning_rate": 2.4993632390347234e-05,
|
|
"loss": 0.1639,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07934527844190598,
|
|
"step": 6385,
|
|
"valid_targets_mean": 5986.8,
|
|
"valid_targets_min": 4634
|
|
},
|
|
{
|
|
"epoch": 3.347302252488214,
|
|
"grad_norm": 0.3614429555300503,
|
|
"learning_rate": 2.4968332178306826e-05,
|
|
"loss": 0.181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06840213388204575,
|
|
"step": 6390,
|
|
"valid_targets_mean": 5988.5,
|
|
"valid_targets_min": 4711
|
|
},
|
|
{
|
|
"epoch": 3.3499214248297537,
|
|
"grad_norm": 0.3601445269199851,
|
|
"learning_rate": 2.4943023489935215e-05,
|
|
"loss": 0.1499,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0641680657863617,
|
|
"step": 6395,
|
|
"valid_targets_mean": 5294.5,
|
|
"valid_targets_min": 4430
|
|
},
|
|
{
|
|
"epoch": 3.352540597171294,
|
|
"grad_norm": 0.3443049477813852,
|
|
"learning_rate": 2.4917706368410838e-05,
|
|
"loss": 0.1473,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07176784425973892,
|
|
"step": 6400,
|
|
"valid_targets_mean": 5903.9,
|
|
"valid_targets_min": 4701
|
|
},
|
|
{
|
|
"epoch": 3.355159769512834,
|
|
"grad_norm": 0.3638713448262528,
|
|
"learning_rate": 2.4892380856926517e-05,
|
|
"loss": 0.1559,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.084088534116745,
|
|
"step": 6405,
|
|
"valid_targets_mean": 6680.9,
|
|
"valid_targets_min": 4501
|
|
},
|
|
{
|
|
"epoch": 3.357778941854374,
|
|
"grad_norm": 0.3849187164924035,
|
|
"learning_rate": 2.48670469986894e-05,
|
|
"loss": 0.1594,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07608664035797119,
|
|
"step": 6410,
|
|
"valid_targets_mean": 5788.4,
|
|
"valid_targets_min": 3287
|
|
},
|
|
{
|
|
"epoch": 3.360398114195914,
|
|
"grad_norm": 0.3772639442062634,
|
|
"learning_rate": 2.4841704836920862e-05,
|
|
"loss": 0.1769,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08706492185592651,
|
|
"step": 6415,
|
|
"valid_targets_mean": 6215.6,
|
|
"valid_targets_min": 4881
|
|
},
|
|
{
|
|
"epoch": 3.3630172865374544,
|
|
"grad_norm": 0.41983338681278204,
|
|
"learning_rate": 2.481635441485645e-05,
|
|
"loss": 0.2065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10334602743387222,
|
|
"step": 6420,
|
|
"valid_targets_mean": 7103.5,
|
|
"valid_targets_min": 4715
|
|
},
|
|
{
|
|
"epoch": 3.365636458878994,
|
|
"grad_norm": 0.3910223540701964,
|
|
"learning_rate": 2.479099577574581e-05,
|
|
"loss": 0.1692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07273611426353455,
|
|
"step": 6425,
|
|
"valid_targets_mean": 5771.0,
|
|
"valid_targets_min": 4403
|
|
},
|
|
{
|
|
"epoch": 3.368255631220534,
|
|
"grad_norm": 0.41699640553441875,
|
|
"learning_rate": 2.4765628962852586e-05,
|
|
"loss": 0.1684,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08446206897497177,
|
|
"step": 6430,
|
|
"valid_targets_mean": 5611.8,
|
|
"valid_targets_min": 4657
|
|
},
|
|
{
|
|
"epoch": 3.3708748035620744,
|
|
"grad_norm": 0.36893431404927785,
|
|
"learning_rate": 2.474025401945439e-05,
|
|
"loss": 0.1673,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07603064179420471,
|
|
"step": 6435,
|
|
"valid_targets_mean": 6171.8,
|
|
"valid_targets_min": 4293
|
|
},
|
|
{
|
|
"epoch": 3.3734939759036147,
|
|
"grad_norm": 0.35851489921835866,
|
|
"learning_rate": 2.4714870988842696e-05,
|
|
"loss": 0.1805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1075579822063446,
|
|
"step": 6440,
|
|
"valid_targets_mean": 6910.0,
|
|
"valid_targets_min": 6043
|
|
},
|
|
{
|
|
"epoch": 3.3761131482451545,
|
|
"grad_norm": 0.3877951110234821,
|
|
"learning_rate": 2.4689479914322777e-05,
|
|
"loss": 0.1807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09149368107318878,
|
|
"step": 6445,
|
|
"valid_targets_mean": 7532.6,
|
|
"valid_targets_min": 4876
|
|
},
|
|
{
|
|
"epoch": 3.3787323205866944,
|
|
"grad_norm": 0.3565687477905426,
|
|
"learning_rate": 2.466408083921363e-05,
|
|
"loss": 0.1714,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07170824706554413,
|
|
"step": 6450,
|
|
"valid_targets_mean": 5248.0,
|
|
"valid_targets_min": 2545
|
|
},
|
|
{
|
|
"epoch": 3.3813514929282347,
|
|
"grad_norm": 0.38380382136855723,
|
|
"learning_rate": 2.4638673806847883e-05,
|
|
"loss": 0.1782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08993254601955414,
|
|
"step": 6455,
|
|
"valid_targets_mean": 6254.9,
|
|
"valid_targets_min": 4800
|
|
},
|
|
{
|
|
"epoch": 3.383970665269775,
|
|
"grad_norm": 0.34414237325334873,
|
|
"learning_rate": 2.4613258860571782e-05,
|
|
"loss": 0.1533,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07681073248386383,
|
|
"step": 6460,
|
|
"valid_targets_mean": 6607.8,
|
|
"valid_targets_min": 4228
|
|
},
|
|
{
|
|
"epoch": 3.386589837611315,
|
|
"grad_norm": 0.3908809599404283,
|
|
"learning_rate": 2.4587836043745027e-05,
|
|
"loss": 0.1657,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07173924148082733,
|
|
"step": 6465,
|
|
"valid_targets_mean": 5782.5,
|
|
"valid_targets_min": 5099
|
|
},
|
|
{
|
|
"epoch": 3.3892090099528547,
|
|
"grad_norm": 0.37696518161027326,
|
|
"learning_rate": 2.4562405399740778e-05,
|
|
"loss": 0.2121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06864018738269806,
|
|
"step": 6470,
|
|
"valid_targets_mean": 5704.1,
|
|
"valid_targets_min": 4157
|
|
},
|
|
{
|
|
"epoch": 3.391828182294395,
|
|
"grad_norm": 0.3668486070721226,
|
|
"learning_rate": 2.4536966971945536e-05,
|
|
"loss": 0.162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07351751625537872,
|
|
"step": 6475,
|
|
"valid_targets_mean": 5938.6,
|
|
"valid_targets_min": 3930
|
|
},
|
|
{
|
|
"epoch": 3.3944473546359353,
|
|
"grad_norm": 0.44902743122082356,
|
|
"learning_rate": 2.4511520803759083e-05,
|
|
"loss": 0.1692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08883160352706909,
|
|
"step": 6480,
|
|
"valid_targets_mean": 5714.6,
|
|
"valid_targets_min": 4876
|
|
},
|
|
{
|
|
"epoch": 3.397066526977475,
|
|
"grad_norm": 0.35888992094307914,
|
|
"learning_rate": 2.448606693859442e-05,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08407334983348846,
|
|
"step": 6485,
|
|
"valid_targets_mean": 6711.4,
|
|
"valid_targets_min": 3054
|
|
},
|
|
{
|
|
"epoch": 3.399685699319015,
|
|
"grad_norm": 0.3671497425717947,
|
|
"learning_rate": 2.4460605419877652e-05,
|
|
"loss": 0.1675,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07031820714473724,
|
|
"step": 6490,
|
|
"valid_targets_mean": 6126.9,
|
|
"valid_targets_min": 4670
|
|
},
|
|
{
|
|
"epoch": 3.4023048716605553,
|
|
"grad_norm": 0.37887286091801736,
|
|
"learning_rate": 2.443513629104796e-05,
|
|
"loss": 0.1521,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08363084495067596,
|
|
"step": 6495,
|
|
"valid_targets_mean": 7143.2,
|
|
"valid_targets_min": 4806
|
|
},
|
|
{
|
|
"epoch": 3.4049240440020956,
|
|
"grad_norm": 0.3911340464093035,
|
|
"learning_rate": 2.4409659595557507e-05,
|
|
"loss": 0.1633,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07207661122083664,
|
|
"step": 6500,
|
|
"valid_targets_mean": 5455.6,
|
|
"valid_targets_min": 4453
|
|
},
|
|
{
|
|
"epoch": 3.4075432163436354,
|
|
"grad_norm": 0.3721682476509893,
|
|
"learning_rate": 2.4384175376871366e-05,
|
|
"loss": 0.1597,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08851202577352524,
|
|
"step": 6505,
|
|
"valid_targets_mean": 7124.2,
|
|
"valid_targets_min": 4935
|
|
},
|
|
{
|
|
"epoch": 3.4101623886851753,
|
|
"grad_norm": 0.39766714127910935,
|
|
"learning_rate": 2.4358683678467442e-05,
|
|
"loss": 0.1544,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07665763795375824,
|
|
"step": 6510,
|
|
"valid_targets_mean": 7344.5,
|
|
"valid_targets_min": 4964
|
|
},
|
|
{
|
|
"epoch": 3.4127815610267156,
|
|
"grad_norm": 0.3385291733527547,
|
|
"learning_rate": 2.43331845438364e-05,
|
|
"loss": 0.1649,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08365072309970856,
|
|
"step": 6515,
|
|
"valid_targets_mean": 7247.2,
|
|
"valid_targets_min": 5625
|
|
},
|
|
{
|
|
"epoch": 3.4154007333682554,
|
|
"grad_norm": 0.35181977929673425,
|
|
"learning_rate": 2.4307678016481595e-05,
|
|
"loss": 0.1668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.087459996342659,
|
|
"step": 6520,
|
|
"valid_targets_mean": 7612.6,
|
|
"valid_targets_min": 5389
|
|
},
|
|
{
|
|
"epoch": 3.4180199057097957,
|
|
"grad_norm": 0.3599454329463783,
|
|
"learning_rate": 2.428216413991899e-05,
|
|
"loss": 0.1545,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07323252409696579,
|
|
"step": 6525,
|
|
"valid_targets_mean": 6254.5,
|
|
"valid_targets_min": 4220
|
|
},
|
|
{
|
|
"epoch": 3.4206390780513356,
|
|
"grad_norm": 0.3638509969197595,
|
|
"learning_rate": 2.4256642957677096e-05,
|
|
"loss": 0.1704,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09068848192691803,
|
|
"step": 6530,
|
|
"valid_targets_mean": 7109.0,
|
|
"valid_targets_min": 5071
|
|
},
|
|
{
|
|
"epoch": 3.423258250392876,
|
|
"grad_norm": 0.39786519172694557,
|
|
"learning_rate": 2.4231114513296868e-05,
|
|
"loss": 0.1699,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0803876668214798,
|
|
"step": 6535,
|
|
"valid_targets_mean": 5811.4,
|
|
"valid_targets_min": 2560
|
|
},
|
|
{
|
|
"epoch": 3.4258774227344158,
|
|
"grad_norm": 0.5287233305230999,
|
|
"learning_rate": 2.420557885033168e-05,
|
|
"loss": 0.1821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.106256864964962,
|
|
"step": 6540,
|
|
"valid_targets_mean": 4137.9,
|
|
"valid_targets_min": 1320
|
|
},
|
|
{
|
|
"epoch": 3.428496595075956,
|
|
"grad_norm": 0.4864357882653739,
|
|
"learning_rate": 2.4180036012347197e-05,
|
|
"loss": 0.2588,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08376993238925934,
|
|
"step": 6545,
|
|
"valid_targets_mean": 3768.8,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 3.431115767417496,
|
|
"grad_norm": 0.4495192508104731,
|
|
"learning_rate": 2.415448604292133e-05,
|
|
"loss": 0.2463,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11839775741100311,
|
|
"step": 6550,
|
|
"valid_targets_mean": 5086.6,
|
|
"valid_targets_min": 1846
|
|
},
|
|
{
|
|
"epoch": 3.433734939759036,
|
|
"grad_norm": 0.45948985894295946,
|
|
"learning_rate": 2.4128928985644165e-05,
|
|
"loss": 0.2382,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0898493304848671,
|
|
"step": 6555,
|
|
"valid_targets_mean": 3826.8,
|
|
"valid_targets_min": 1321
|
|
},
|
|
{
|
|
"epoch": 3.436354112100576,
|
|
"grad_norm": 0.47832825522476635,
|
|
"learning_rate": 2.4103364884117878e-05,
|
|
"loss": 0.2433,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13066086173057556,
|
|
"step": 6560,
|
|
"valid_targets_mean": 6475.0,
|
|
"valid_targets_min": 3049
|
|
},
|
|
{
|
|
"epoch": 3.4389732844421164,
|
|
"grad_norm": 0.5336403968678465,
|
|
"learning_rate": 2.4077793781956652e-05,
|
|
"loss": 0.2531,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12808701395988464,
|
|
"step": 6565,
|
|
"valid_targets_mean": 4328.6,
|
|
"valid_targets_min": 2075
|
|
},
|
|
{
|
|
"epoch": 3.441592456783656,
|
|
"grad_norm": 0.4749600720867551,
|
|
"learning_rate": 2.4052215722786635e-05,
|
|
"loss": 0.2466,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12571166455745697,
|
|
"step": 6570,
|
|
"valid_targets_mean": 5135.4,
|
|
"valid_targets_min": 1991
|
|
},
|
|
{
|
|
"epoch": 3.4442116291251965,
|
|
"grad_norm": 0.5402532183284965,
|
|
"learning_rate": 2.4026630750245823e-05,
|
|
"loss": 0.239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11091949045658112,
|
|
"step": 6575,
|
|
"valid_targets_mean": 3699.0,
|
|
"valid_targets_min": 1665
|
|
},
|
|
{
|
|
"epoch": 3.4468308014667364,
|
|
"grad_norm": 0.49991999722513,
|
|
"learning_rate": 2.4001038907984015e-05,
|
|
"loss": 0.2452,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11975577473640442,
|
|
"step": 6580,
|
|
"valid_targets_mean": 4666.8,
|
|
"valid_targets_min": 578
|
|
},
|
|
{
|
|
"epoch": 3.4494499738082767,
|
|
"grad_norm": 0.49919660827363055,
|
|
"learning_rate": 2.3975440239662724e-05,
|
|
"loss": 0.2529,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11430370062589645,
|
|
"step": 6585,
|
|
"valid_targets_mean": 4916.1,
|
|
"valid_targets_min": 1241
|
|
},
|
|
{
|
|
"epoch": 3.4520691461498165,
|
|
"grad_norm": 0.5215501546589721,
|
|
"learning_rate": 2.3949834788955136e-05,
|
|
"loss": 0.2468,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12626492977142334,
|
|
"step": 6590,
|
|
"valid_targets_mean": 4525.4,
|
|
"valid_targets_min": 2726
|
|
},
|
|
{
|
|
"epoch": 3.454688318491357,
|
|
"grad_norm": 0.4371477610579171,
|
|
"learning_rate": 2.3924222599545967e-05,
|
|
"loss": 0.2403,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11062407493591309,
|
|
"step": 6595,
|
|
"valid_targets_mean": 6055.0,
|
|
"valid_targets_min": 2639
|
|
},
|
|
{
|
|
"epoch": 3.4573074908328967,
|
|
"grad_norm": 0.513368675015223,
|
|
"learning_rate": 2.3898603715131455e-05,
|
|
"loss": 0.2417,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1264042854309082,
|
|
"step": 6600,
|
|
"valid_targets_mean": 5447.0,
|
|
"valid_targets_min": 2244
|
|
},
|
|
{
|
|
"epoch": 3.459926663174437,
|
|
"grad_norm": 0.49105031757605666,
|
|
"learning_rate": 2.387297817941926e-05,
|
|
"loss": 0.2469,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1250518560409546,
|
|
"step": 6605,
|
|
"valid_targets_mean": 5634.1,
|
|
"valid_targets_min": 3076
|
|
},
|
|
{
|
|
"epoch": 3.462545835515977,
|
|
"grad_norm": 0.5547062342359849,
|
|
"learning_rate": 2.384734603612838e-05,
|
|
"loss": 0.2418,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13765278458595276,
|
|
"step": 6610,
|
|
"valid_targets_mean": 4215.6,
|
|
"valid_targets_min": 1721
|
|
},
|
|
{
|
|
"epoch": 3.465165007857517,
|
|
"grad_norm": 0.5817376589491438,
|
|
"learning_rate": 2.382170732898909e-05,
|
|
"loss": 0.2483,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11693128943443298,
|
|
"step": 6615,
|
|
"valid_targets_mean": 3667.0,
|
|
"valid_targets_min": 1705
|
|
},
|
|
{
|
|
"epoch": 3.467784180199057,
|
|
"grad_norm": 0.658845436501259,
|
|
"learning_rate": 2.3796062101742864e-05,
|
|
"loss": 0.2313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10137315094470978,
|
|
"step": 6620,
|
|
"valid_targets_mean": 3136.5,
|
|
"valid_targets_min": 1538
|
|
},
|
|
{
|
|
"epoch": 3.4704033525405973,
|
|
"grad_norm": 0.5716954504474003,
|
|
"learning_rate": 2.3770410398142307e-05,
|
|
"loss": 0.2416,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09106269478797913,
|
|
"step": 6625,
|
|
"valid_targets_mean": 3045.5,
|
|
"valid_targets_min": 1451
|
|
},
|
|
{
|
|
"epoch": 3.473022524882137,
|
|
"grad_norm": 0.560592583138159,
|
|
"learning_rate": 2.374475226195106e-05,
|
|
"loss": 0.2523,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12927128374576569,
|
|
"step": 6630,
|
|
"valid_targets_mean": 4635.6,
|
|
"valid_targets_min": 2560
|
|
},
|
|
{
|
|
"epoch": 3.4756416972236774,
|
|
"grad_norm": 0.579320179595869,
|
|
"learning_rate": 2.371908773694375e-05,
|
|
"loss": 0.2516,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11526928842067719,
|
|
"step": 6635,
|
|
"valid_targets_mean": 3925.4,
|
|
"valid_targets_min": 1612
|
|
},
|
|
{
|
|
"epoch": 3.4782608695652173,
|
|
"grad_norm": 0.5839141166219242,
|
|
"learning_rate": 2.3693416866905885e-05,
|
|
"loss": 0.2478,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14135050773620605,
|
|
"step": 6640,
|
|
"valid_targets_mean": 4462.9,
|
|
"valid_targets_min": 1921
|
|
},
|
|
{
|
|
"epoch": 3.4808800419067576,
|
|
"grad_norm": 0.5435071521477286,
|
|
"learning_rate": 2.3667739695633825e-05,
|
|
"loss": 0.2324,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11711271107196808,
|
|
"step": 6645,
|
|
"valid_targets_mean": 4697.6,
|
|
"valid_targets_min": 2004
|
|
},
|
|
{
|
|
"epoch": 3.4834992142482974,
|
|
"grad_norm": 0.5813285853264689,
|
|
"learning_rate": 2.364205626693466e-05,
|
|
"loss": 0.2436,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09046830236911774,
|
|
"step": 6650,
|
|
"valid_targets_mean": 3302.0,
|
|
"valid_targets_min": 1896
|
|
},
|
|
{
|
|
"epoch": 3.4861183865898377,
|
|
"grad_norm": 0.6399815223156761,
|
|
"learning_rate": 2.3616366624626165e-05,
|
|
"loss": 0.2234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1311061531305313,
|
|
"step": 6655,
|
|
"valid_targets_mean": 3412.4,
|
|
"valid_targets_min": 1846
|
|
},
|
|
{
|
|
"epoch": 3.4887375589313776,
|
|
"grad_norm": 0.6133645420794462,
|
|
"learning_rate": 2.3590670812536713e-05,
|
|
"loss": 0.2467,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15300464630126953,
|
|
"step": 6660,
|
|
"valid_targets_mean": 4289.0,
|
|
"valid_targets_min": 3360
|
|
},
|
|
{
|
|
"epoch": 3.491356731272918,
|
|
"grad_norm": 0.551325447894351,
|
|
"learning_rate": 2.3564968874505202e-05,
|
|
"loss": 0.2386,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11628319323062897,
|
|
"step": 6665,
|
|
"valid_targets_mean": 3717.1,
|
|
"valid_targets_min": 1475
|
|
},
|
|
{
|
|
"epoch": 3.4939759036144578,
|
|
"grad_norm": 0.5698229078957439,
|
|
"learning_rate": 2.3539260854380982e-05,
|
|
"loss": 0.2414,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1304197609424591,
|
|
"step": 6670,
|
|
"valid_targets_mean": 4400.9,
|
|
"valid_targets_min": 2229
|
|
},
|
|
{
|
|
"epoch": 3.496595075955998,
|
|
"grad_norm": 0.5828878520133568,
|
|
"learning_rate": 2.3513546796023783e-05,
|
|
"loss": 0.2437,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10436005890369415,
|
|
"step": 6675,
|
|
"valid_targets_mean": 3238.2,
|
|
"valid_targets_min": 1233
|
|
},
|
|
{
|
|
"epoch": 3.499214248297538,
|
|
"grad_norm": 0.5919028438370402,
|
|
"learning_rate": 2.348782674330363e-05,
|
|
"loss": 0.2283,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10672833770513535,
|
|
"step": 6680,
|
|
"valid_targets_mean": 3109.8,
|
|
"valid_targets_min": 869
|
|
},
|
|
{
|
|
"epoch": 3.501833420639078,
|
|
"grad_norm": 0.5296272359340188,
|
|
"learning_rate": 2.346210074010078e-05,
|
|
"loss": 0.2397,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1168830394744873,
|
|
"step": 6685,
|
|
"valid_targets_mean": 4352.0,
|
|
"valid_targets_min": 1781
|
|
},
|
|
{
|
|
"epoch": 3.504452592980618,
|
|
"grad_norm": 0.6155343206368286,
|
|
"learning_rate": 2.343636883030564e-05,
|
|
"loss": 0.2366,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1330322027206421,
|
|
"step": 6690,
|
|
"valid_targets_mean": 3687.0,
|
|
"valid_targets_min": 1108
|
|
},
|
|
{
|
|
"epoch": 3.5070717653221584,
|
|
"grad_norm": 0.636116669556504,
|
|
"learning_rate": 2.3410631057818703e-05,
|
|
"loss": 0.2323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11047638952732086,
|
|
"step": 6695,
|
|
"valid_targets_mean": 2814.9,
|
|
"valid_targets_min": 1913
|
|
},
|
|
{
|
|
"epoch": 3.509690937663698,
|
|
"grad_norm": 0.5987881246802838,
|
|
"learning_rate": 2.338488746655044e-05,
|
|
"loss": 0.2361,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13345681130886078,
|
|
"step": 6700,
|
|
"valid_targets_mean": 4176.2,
|
|
"valid_targets_min": 1828
|
|
},
|
|
{
|
|
"epoch": 3.512310110005238,
|
|
"grad_norm": 0.5727757188547329,
|
|
"learning_rate": 2.3359138100421274e-05,
|
|
"loss": 0.2405,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10748310387134552,
|
|
"step": 6705,
|
|
"valid_targets_mean": 3962.1,
|
|
"valid_targets_min": 1589
|
|
},
|
|
{
|
|
"epoch": 3.5149292823467784,
|
|
"grad_norm": 0.6323122660725055,
|
|
"learning_rate": 2.333338300336147e-05,
|
|
"loss": 0.2282,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11577625572681427,
|
|
"step": 6710,
|
|
"valid_targets_mean": 3307.2,
|
|
"valid_targets_min": 2289
|
|
},
|
|
{
|
|
"epoch": 3.5175484546883187,
|
|
"grad_norm": 0.605942931678382,
|
|
"learning_rate": 2.3307622219311078e-05,
|
|
"loss": 0.2331,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08664129674434662,
|
|
"step": 6715,
|
|
"valid_targets_mean": 2650.5,
|
|
"valid_targets_min": 1248
|
|
},
|
|
{
|
|
"epoch": 3.5201676270298585,
|
|
"grad_norm": 0.5885722207923769,
|
|
"learning_rate": 2.3281855792219834e-05,
|
|
"loss": 0.2449,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10523630678653717,
|
|
"step": 6720,
|
|
"valid_targets_mean": 3181.0,
|
|
"valid_targets_min": 1727
|
|
},
|
|
{
|
|
"epoch": 3.5227867993713984,
|
|
"grad_norm": 0.569475051288075,
|
|
"learning_rate": 2.3256083766047113e-05,
|
|
"loss": 0.2351,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13042029738426208,
|
|
"step": 6725,
|
|
"valid_targets_mean": 4625.5,
|
|
"valid_targets_min": 1965
|
|
},
|
|
{
|
|
"epoch": 3.5254059717129387,
|
|
"grad_norm": 0.583999831780277,
|
|
"learning_rate": 2.3230306184761853e-05,
|
|
"loss": 0.2294,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13146890699863434,
|
|
"step": 6730,
|
|
"valid_targets_mean": 3945.2,
|
|
"valid_targets_min": 2018
|
|
},
|
|
{
|
|
"epoch": 3.528025144054479,
|
|
"grad_norm": 0.565115221902501,
|
|
"learning_rate": 2.3204523092342442e-05,
|
|
"loss": 0.2331,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11057507246732712,
|
|
"step": 6735,
|
|
"valid_targets_mean": 4180.6,
|
|
"valid_targets_min": 1730
|
|
},
|
|
{
|
|
"epoch": 3.530644316396019,
|
|
"grad_norm": 0.5907782420435588,
|
|
"learning_rate": 2.3178734532776685e-05,
|
|
"loss": 0.2432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12093518674373627,
|
|
"step": 6740,
|
|
"valid_targets_mean": 3331.2,
|
|
"valid_targets_min": 2097
|
|
},
|
|
{
|
|
"epoch": 3.5332634887375587,
|
|
"grad_norm": 0.6428180267374662,
|
|
"learning_rate": 2.3152940550061728e-05,
|
|
"loss": 0.2298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13893792033195496,
|
|
"step": 6745,
|
|
"valid_targets_mean": 3802.9,
|
|
"valid_targets_min": 1648
|
|
},
|
|
{
|
|
"epoch": 3.535882661079099,
|
|
"grad_norm": 0.5816173407043932,
|
|
"learning_rate": 2.3127141188203948e-05,
|
|
"loss": 0.2391,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10701972246170044,
|
|
"step": 6750,
|
|
"valid_targets_mean": 3595.1,
|
|
"valid_targets_min": 1421
|
|
},
|
|
{
|
|
"epoch": 3.5385018334206393,
|
|
"grad_norm": 0.6058804924806588,
|
|
"learning_rate": 2.3101336491218916e-05,
|
|
"loss": 0.2344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12693704664707184,
|
|
"step": 6755,
|
|
"valid_targets_mean": 3734.4,
|
|
"valid_targets_min": 2284
|
|
},
|
|
{
|
|
"epoch": 3.541121005762179,
|
|
"grad_norm": 0.5241777612491162,
|
|
"learning_rate": 2.3075526503131283e-05,
|
|
"loss": 0.2231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10609862208366394,
|
|
"step": 6760,
|
|
"valid_targets_mean": 4147.4,
|
|
"valid_targets_min": 2095
|
|
},
|
|
{
|
|
"epoch": 3.543740178103719,
|
|
"grad_norm": 0.5964242963664286,
|
|
"learning_rate": 2.304971126797475e-05,
|
|
"loss": 0.2321,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09846395254135132,
|
|
"step": 6765,
|
|
"valid_targets_mean": 3100.2,
|
|
"valid_targets_min": 1673
|
|
},
|
|
{
|
|
"epoch": 3.5463593504452593,
|
|
"grad_norm": 0.5303796472746766,
|
|
"learning_rate": 2.302389082979196e-05,
|
|
"loss": 0.24,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13052743673324585,
|
|
"step": 6770,
|
|
"valid_targets_mean": 4951.2,
|
|
"valid_targets_min": 2067
|
|
},
|
|
{
|
|
"epoch": 3.5489785227867996,
|
|
"grad_norm": 0.5751260888215624,
|
|
"learning_rate": 2.2998065232634428e-05,
|
|
"loss": 0.2326,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.087217777967453,
|
|
"step": 6775,
|
|
"valid_targets_mean": 2375.2,
|
|
"valid_targets_min": 742
|
|
},
|
|
{
|
|
"epoch": 3.5515976951283394,
|
|
"grad_norm": 0.5669176050593572,
|
|
"learning_rate": 2.2972234520562485e-05,
|
|
"loss": 0.2438,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1378200203180313,
|
|
"step": 6780,
|
|
"valid_targets_mean": 4786.8,
|
|
"valid_targets_min": 1766
|
|
},
|
|
{
|
|
"epoch": 3.5542168674698793,
|
|
"grad_norm": 0.6453716565668436,
|
|
"learning_rate": 2.2946398737645173e-05,
|
|
"loss": 0.2309,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11073620617389679,
|
|
"step": 6785,
|
|
"valid_targets_mean": 3331.2,
|
|
"valid_targets_min": 1504
|
|
},
|
|
{
|
|
"epoch": 3.5568360398114196,
|
|
"grad_norm": 0.5577886544254315,
|
|
"learning_rate": 2.292055792796019e-05,
|
|
"loss": 0.2386,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10144495964050293,
|
|
"step": 6790,
|
|
"valid_targets_mean": 3742.8,
|
|
"valid_targets_min": 1170
|
|
},
|
|
{
|
|
"epoch": 3.55945521215296,
|
|
"grad_norm": 0.6007736438160111,
|
|
"learning_rate": 2.2894712135593816e-05,
|
|
"loss": 0.2297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11968302726745605,
|
|
"step": 6795,
|
|
"valid_targets_mean": 3477.5,
|
|
"valid_targets_min": 2669
|
|
},
|
|
{
|
|
"epoch": 3.5620743844944998,
|
|
"grad_norm": 0.5852908871471947,
|
|
"learning_rate": 2.2868861404640826e-05,
|
|
"loss": 0.2415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11945214867591858,
|
|
"step": 6800,
|
|
"valid_targets_mean": 3669.4,
|
|
"valid_targets_min": 2923
|
|
},
|
|
{
|
|
"epoch": 3.5646935568360396,
|
|
"grad_norm": 0.527808683195503,
|
|
"learning_rate": 2.2843005779204424e-05,
|
|
"loss": 0.225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09395307302474976,
|
|
"step": 6805,
|
|
"valid_targets_mean": 3643.6,
|
|
"valid_targets_min": 2756
|
|
},
|
|
{
|
|
"epoch": 3.56731272917758,
|
|
"grad_norm": 0.5576057861287207,
|
|
"learning_rate": 2.2817145303396155e-05,
|
|
"loss": 0.2229,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11294896900653839,
|
|
"step": 6810,
|
|
"valid_targets_mean": 4241.8,
|
|
"valid_targets_min": 2526
|
|
},
|
|
{
|
|
"epoch": 3.56993190151912,
|
|
"grad_norm": 0.5501081807405083,
|
|
"learning_rate": 2.279128002133586e-05,
|
|
"loss": 0.2163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12715858221054077,
|
|
"step": 6815,
|
|
"valid_targets_mean": 5317.1,
|
|
"valid_targets_min": 464
|
|
},
|
|
{
|
|
"epoch": 3.57255107386066,
|
|
"grad_norm": 0.5597244011271181,
|
|
"learning_rate": 2.276540997715155e-05,
|
|
"loss": 0.2481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1311587691307068,
|
|
"step": 6820,
|
|
"valid_targets_mean": 4387.4,
|
|
"valid_targets_min": 1003
|
|
},
|
|
{
|
|
"epoch": 3.5751702462022,
|
|
"grad_norm": 0.5704061903557548,
|
|
"learning_rate": 2.273953521497938e-05,
|
|
"loss": 0.231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12439266592264175,
|
|
"step": 6825,
|
|
"valid_targets_mean": 3934.6,
|
|
"valid_targets_min": 1632
|
|
},
|
|
{
|
|
"epoch": 3.57778941854374,
|
|
"grad_norm": 0.5752906314810322,
|
|
"learning_rate": 2.2713655778963563e-05,
|
|
"loss": 0.2361,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1213194727897644,
|
|
"step": 6830,
|
|
"valid_targets_mean": 3808.1,
|
|
"valid_targets_min": 1969
|
|
},
|
|
{
|
|
"epoch": 3.5804085908852805,
|
|
"grad_norm": 0.559386966041707,
|
|
"learning_rate": 2.268777171325626e-05,
|
|
"loss": 0.2282,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09403820335865021,
|
|
"step": 6835,
|
|
"valid_targets_mean": 3318.8,
|
|
"valid_targets_min": 1491
|
|
},
|
|
{
|
|
"epoch": 3.5830277632268204,
|
|
"grad_norm": 0.6160088400557124,
|
|
"learning_rate": 2.2661883062017558e-05,
|
|
"loss": 0.247,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13603980839252472,
|
|
"step": 6840,
|
|
"valid_targets_mean": 4169.9,
|
|
"valid_targets_min": 1637
|
|
},
|
|
{
|
|
"epoch": 3.5856469355683602,
|
|
"grad_norm": 0.522822036080833,
|
|
"learning_rate": 2.263598986941534e-05,
|
|
"loss": 0.2344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10942547023296356,
|
|
"step": 6845,
|
|
"valid_targets_mean": 3640.1,
|
|
"valid_targets_min": 1818
|
|
},
|
|
{
|
|
"epoch": 3.5882661079099005,
|
|
"grad_norm": 0.5866234190352576,
|
|
"learning_rate": 2.261009217962526e-05,
|
|
"loss": 0.2264,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12125527113676071,
|
|
"step": 6850,
|
|
"valid_targets_mean": 3803.0,
|
|
"valid_targets_min": 1523
|
|
},
|
|
{
|
|
"epoch": 3.5908852802514404,
|
|
"grad_norm": 0.5786354258156285,
|
|
"learning_rate": 2.2584190036830635e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1317993700504303,
|
|
"step": 6855,
|
|
"valid_targets_mean": 4696.1,
|
|
"valid_targets_min": 1814
|
|
},
|
|
{
|
|
"epoch": 3.5935044525929807,
|
|
"grad_norm": 0.6453915038756044,
|
|
"learning_rate": 2.2558283485222367e-05,
|
|
"loss": 0.2222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12326325476169586,
|
|
"step": 6860,
|
|
"valid_targets_mean": 3022.9,
|
|
"valid_targets_min": 1073
|
|
},
|
|
{
|
|
"epoch": 3.5961236249345205,
|
|
"grad_norm": 0.5588512120546468,
|
|
"learning_rate": 2.2532372568998903e-05,
|
|
"loss": 0.2207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10281728953123093,
|
|
"step": 6865,
|
|
"valid_targets_mean": 3355.5,
|
|
"valid_targets_min": 1654
|
|
},
|
|
{
|
|
"epoch": 3.598742797276061,
|
|
"grad_norm": 0.6307951883780548,
|
|
"learning_rate": 2.2506457332366124e-05,
|
|
"loss": 0.2149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1167975515127182,
|
|
"step": 6870,
|
|
"valid_targets_mean": 3669.1,
|
|
"valid_targets_min": 1079
|
|
},
|
|
{
|
|
"epoch": 3.6013619696176007,
|
|
"grad_norm": 0.6086990493021137,
|
|
"learning_rate": 2.2480537819537293e-05,
|
|
"loss": 0.2127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08841513097286224,
|
|
"step": 6875,
|
|
"valid_targets_mean": 3073.5,
|
|
"valid_targets_min": 1143
|
|
},
|
|
{
|
|
"epoch": 3.603981141959141,
|
|
"grad_norm": 0.581842629892961,
|
|
"learning_rate": 2.2454614074732936e-05,
|
|
"loss": 0.2208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0971125066280365,
|
|
"step": 6880,
|
|
"valid_targets_mean": 3300.2,
|
|
"valid_targets_min": 1853
|
|
},
|
|
{
|
|
"epoch": 3.606600314300681,
|
|
"grad_norm": 0.6784781899860668,
|
|
"learning_rate": 2.242868614218084e-05,
|
|
"loss": 0.2494,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1171766147017479,
|
|
"step": 6885,
|
|
"valid_targets_mean": 3122.8,
|
|
"valid_targets_min": 1410
|
|
},
|
|
{
|
|
"epoch": 3.609219486642221,
|
|
"grad_norm": 0.6675489122087376,
|
|
"learning_rate": 2.2402754066115905e-05,
|
|
"loss": 0.2223,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09704920649528503,
|
|
"step": 6890,
|
|
"valid_targets_mean": 2564.8,
|
|
"valid_targets_min": 1566
|
|
},
|
|
{
|
|
"epoch": 3.611838658983761,
|
|
"grad_norm": 0.603871032055132,
|
|
"learning_rate": 2.2376817890780128e-05,
|
|
"loss": 0.2345,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11461576819419861,
|
|
"step": 6895,
|
|
"valid_targets_mean": 3865.4,
|
|
"valid_targets_min": 2335
|
|
},
|
|
{
|
|
"epoch": 3.6144578313253013,
|
|
"grad_norm": 0.5921554960682128,
|
|
"learning_rate": 2.2350877660422475e-05,
|
|
"loss": 0.2244,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11522124707698822,
|
|
"step": 6900,
|
|
"valid_targets_mean": 3269.5,
|
|
"valid_targets_min": 1255
|
|
},
|
|
{
|
|
"epoch": 3.617077003666841,
|
|
"grad_norm": 0.6134455521032701,
|
|
"learning_rate": 2.232493341929885e-05,
|
|
"loss": 0.2233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12292369455099106,
|
|
"step": 6905,
|
|
"valid_targets_mean": 3915.5,
|
|
"valid_targets_min": 2355
|
|
},
|
|
{
|
|
"epoch": 3.6196961760083814,
|
|
"grad_norm": 0.5872835124432654,
|
|
"learning_rate": 2.2298985211671975e-05,
|
|
"loss": 0.222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09926629066467285,
|
|
"step": 6910,
|
|
"valid_targets_mean": 3704.9,
|
|
"valid_targets_min": 1281
|
|
},
|
|
{
|
|
"epoch": 3.6223153483499213,
|
|
"grad_norm": 0.8559885474385842,
|
|
"learning_rate": 2.2273033081811367e-05,
|
|
"loss": 0.2233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12465713173151016,
|
|
"step": 6915,
|
|
"valid_targets_mean": 3788.6,
|
|
"valid_targets_min": 1836
|
|
},
|
|
{
|
|
"epoch": 3.6249345206914616,
|
|
"grad_norm": 0.5607531279508331,
|
|
"learning_rate": 2.2247077073993217e-05,
|
|
"loss": 0.2284,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15465277433395386,
|
|
"step": 6920,
|
|
"valid_targets_mean": 5778.1,
|
|
"valid_targets_min": 4218
|
|
},
|
|
{
|
|
"epoch": 3.6275536930330015,
|
|
"grad_norm": 0.5709923678031399,
|
|
"learning_rate": 2.222111723250034e-05,
|
|
"loss": 0.2229,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11252649128437042,
|
|
"step": 6925,
|
|
"valid_targets_mean": 3982.6,
|
|
"valid_targets_min": 2717
|
|
},
|
|
{
|
|
"epoch": 3.6301728653745418,
|
|
"grad_norm": 0.6328013307602074,
|
|
"learning_rate": 2.219515360162208e-05,
|
|
"loss": 0.2268,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11584963649511337,
|
|
"step": 6930,
|
|
"valid_targets_mean": 2804.9,
|
|
"valid_targets_min": 1171
|
|
},
|
|
{
|
|
"epoch": 3.6327920377160816,
|
|
"grad_norm": 0.5784454533017686,
|
|
"learning_rate": 2.2169186225654266e-05,
|
|
"loss": 0.2368,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12169316411018372,
|
|
"step": 6935,
|
|
"valid_targets_mean": 4265.1,
|
|
"valid_targets_min": 2177
|
|
},
|
|
{
|
|
"epoch": 3.635411210057622,
|
|
"grad_norm": 0.5440171631544455,
|
|
"learning_rate": 2.2143215148899094e-05,
|
|
"loss": 0.2311,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11516569554805756,
|
|
"step": 6940,
|
|
"valid_targets_mean": 4090.0,
|
|
"valid_targets_min": 2368
|
|
},
|
|
{
|
|
"epoch": 3.6380303823991618,
|
|
"grad_norm": 0.5768498905763396,
|
|
"learning_rate": 2.2117240415665093e-05,
|
|
"loss": 0.2383,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10153605788946152,
|
|
"step": 6945,
|
|
"valid_targets_mean": 4149.6,
|
|
"valid_targets_min": 2294
|
|
},
|
|
{
|
|
"epoch": 3.640649554740702,
|
|
"grad_norm": 0.6149359693531732,
|
|
"learning_rate": 2.2091262070267017e-05,
|
|
"loss": 0.2373,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13069310784339905,
|
|
"step": 6950,
|
|
"valid_targets_mean": 3977.1,
|
|
"valid_targets_min": 1135
|
|
},
|
|
{
|
|
"epoch": 3.643268727082242,
|
|
"grad_norm": 0.5984225983630644,
|
|
"learning_rate": 2.2065280157025786e-05,
|
|
"loss": 0.233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12294389307498932,
|
|
"step": 6955,
|
|
"valid_targets_mean": 4319.0,
|
|
"valid_targets_min": 1449
|
|
},
|
|
{
|
|
"epoch": 3.645887899423782,
|
|
"grad_norm": 0.6470752148570302,
|
|
"learning_rate": 2.2039294720268413e-05,
|
|
"loss": 0.2417,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13006079196929932,
|
|
"step": 6960,
|
|
"valid_targets_mean": 3227.5,
|
|
"valid_targets_min": 1400
|
|
},
|
|
{
|
|
"epoch": 3.648507071765322,
|
|
"grad_norm": 0.5732468220240098,
|
|
"learning_rate": 2.2013305804327923e-05,
|
|
"loss": 0.2405,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0939599871635437,
|
|
"step": 6965,
|
|
"valid_targets_mean": 3268.4,
|
|
"valid_targets_min": 1433
|
|
},
|
|
{
|
|
"epoch": 3.6511262441068624,
|
|
"grad_norm": 0.6066086118447095,
|
|
"learning_rate": 2.1987313453543254e-05,
|
|
"loss": 0.2222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08269931375980377,
|
|
"step": 6970,
|
|
"valid_targets_mean": 2946.5,
|
|
"valid_targets_min": 614
|
|
},
|
|
{
|
|
"epoch": 3.6537454164484022,
|
|
"grad_norm": 0.6271052804557808,
|
|
"learning_rate": 2.1961317712259237e-05,
|
|
"loss": 0.2292,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09952224791049957,
|
|
"step": 6975,
|
|
"valid_targets_mean": 3125.0,
|
|
"valid_targets_min": 2041
|
|
},
|
|
{
|
|
"epoch": 3.6563645887899425,
|
|
"grad_norm": 0.6305679688576481,
|
|
"learning_rate": 2.193531862482646e-05,
|
|
"loss": 0.2385,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12361295521259308,
|
|
"step": 6980,
|
|
"valid_targets_mean": 3852.9,
|
|
"valid_targets_min": 1333
|
|
},
|
|
{
|
|
"epoch": 3.6589837611314824,
|
|
"grad_norm": 0.5632433130439336,
|
|
"learning_rate": 2.1909316235601246e-05,
|
|
"loss": 0.2278,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09937632083892822,
|
|
"step": 6985,
|
|
"valid_targets_mean": 3973.2,
|
|
"valid_targets_min": 1325
|
|
},
|
|
{
|
|
"epoch": 3.6616029334730227,
|
|
"grad_norm": 0.5264059565376541,
|
|
"learning_rate": 2.1883310588945523e-05,
|
|
"loss": 0.2266,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08674407750368118,
|
|
"step": 6990,
|
|
"valid_targets_mean": 3510.1,
|
|
"valid_targets_min": 2439
|
|
},
|
|
{
|
|
"epoch": 3.6642221058145625,
|
|
"grad_norm": 0.5763460513999769,
|
|
"learning_rate": 2.18573017292268e-05,
|
|
"loss": 0.2196,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11072199046611786,
|
|
"step": 6995,
|
|
"valid_targets_mean": 3699.4,
|
|
"valid_targets_min": 2651
|
|
},
|
|
{
|
|
"epoch": 3.666841278156103,
|
|
"grad_norm": 0.6203362612539552,
|
|
"learning_rate": 2.1831289700818058e-05,
|
|
"loss": 0.2313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1367991864681244,
|
|
"step": 7000,
|
|
"valid_targets_mean": 4253.4,
|
|
"valid_targets_min": 1076
|
|
},
|
|
{
|
|
"epoch": 3.6694604504976427,
|
|
"grad_norm": 0.5640191516108097,
|
|
"learning_rate": 2.1805274548097675e-05,
|
|
"loss": 0.2136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12644751369953156,
|
|
"step": 7005,
|
|
"valid_targets_mean": 5076.9,
|
|
"valid_targets_min": 2127
|
|
},
|
|
{
|
|
"epoch": 3.672079622839183,
|
|
"grad_norm": 0.5677460744297587,
|
|
"learning_rate": 2.177925631544937e-05,
|
|
"loss": 0.2324,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11261679232120514,
|
|
"step": 7010,
|
|
"valid_targets_mean": 4415.2,
|
|
"valid_targets_min": 3904
|
|
},
|
|
{
|
|
"epoch": 3.674698795180723,
|
|
"grad_norm": 0.6044780997676237,
|
|
"learning_rate": 2.1753235047262124e-05,
|
|
"loss": 0.2288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08222821354866028,
|
|
"step": 7015,
|
|
"valid_targets_mean": 2814.0,
|
|
"valid_targets_min": 793
|
|
},
|
|
{
|
|
"epoch": 3.6773179675222627,
|
|
"grad_norm": 0.5698177819651197,
|
|
"learning_rate": 2.172721078793008e-05,
|
|
"loss": 0.2298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11177408695220947,
|
|
"step": 7020,
|
|
"valid_targets_mean": 4488.0,
|
|
"valid_targets_min": 1777
|
|
},
|
|
{
|
|
"epoch": 3.679937139863803,
|
|
"grad_norm": 0.5685644784817925,
|
|
"learning_rate": 2.1701183581852497e-05,
|
|
"loss": 0.2233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10050319135189056,
|
|
"step": 7025,
|
|
"valid_targets_mean": 3447.1,
|
|
"valid_targets_min": 1965
|
|
},
|
|
{
|
|
"epoch": 3.6825563122053433,
|
|
"grad_norm": 0.5914313903231709,
|
|
"learning_rate": 2.1675153473433653e-05,
|
|
"loss": 0.2323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1289793699979782,
|
|
"step": 7030,
|
|
"valid_targets_mean": 4516.6,
|
|
"valid_targets_min": 1206
|
|
},
|
|
{
|
|
"epoch": 3.685175484546883,
|
|
"grad_norm": 0.5416343694362858,
|
|
"learning_rate": 2.1649120507082776e-05,
|
|
"loss": 0.2272,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1228301003575325,
|
|
"step": 7035,
|
|
"valid_targets_mean": 5388.2,
|
|
"valid_targets_min": 3599
|
|
},
|
|
{
|
|
"epoch": 3.687794656888423,
|
|
"grad_norm": 0.5824407282519591,
|
|
"learning_rate": 2.162308472721398e-05,
|
|
"loss": 0.2155,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10571618378162384,
|
|
"step": 7040,
|
|
"valid_targets_mean": 3637.1,
|
|
"valid_targets_min": 1250
|
|
},
|
|
{
|
|
"epoch": 3.6904138292299633,
|
|
"grad_norm": 0.5941806822380946,
|
|
"learning_rate": 2.1597046178246174e-05,
|
|
"loss": 0.2169,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11149287968873978,
|
|
"step": 7045,
|
|
"valid_targets_mean": 4201.2,
|
|
"valid_targets_min": 1876
|
|
},
|
|
{
|
|
"epoch": 3.6930330015715036,
|
|
"grad_norm": 0.5393176593744349,
|
|
"learning_rate": 2.1571004904602993e-05,
|
|
"loss": 0.2183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11081297695636749,
|
|
"step": 7050,
|
|
"valid_targets_mean": 4462.0,
|
|
"valid_targets_min": 2487
|
|
},
|
|
{
|
|
"epoch": 3.6956521739130435,
|
|
"grad_norm": 0.5873772296357762,
|
|
"learning_rate": 2.154496095071272e-05,
|
|
"loss": 0.2192,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09898903965950012,
|
|
"step": 7055,
|
|
"valid_targets_mean": 3380.9,
|
|
"valid_targets_min": 1360
|
|
},
|
|
{
|
|
"epoch": 3.6982713462545833,
|
|
"grad_norm": 0.589105360240938,
|
|
"learning_rate": 2.1518914361008197e-05,
|
|
"loss": 0.2138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11599930375814438,
|
|
"step": 7060,
|
|
"valid_targets_mean": 4094.6,
|
|
"valid_targets_min": 1496
|
|
},
|
|
{
|
|
"epoch": 3.7008905185961236,
|
|
"grad_norm": 0.6242763783101974,
|
|
"learning_rate": 2.1492865179926786e-05,
|
|
"loss": 0.2138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12099158763885498,
|
|
"step": 7065,
|
|
"valid_targets_mean": 3533.5,
|
|
"valid_targets_min": 2059
|
|
},
|
|
{
|
|
"epoch": 3.703509690937664,
|
|
"grad_norm": 0.579993260527459,
|
|
"learning_rate": 2.1466813451910253e-05,
|
|
"loss": 0.217,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11872360855340958,
|
|
"step": 7070,
|
|
"valid_targets_mean": 4585.9,
|
|
"valid_targets_min": 2890
|
|
},
|
|
{
|
|
"epoch": 3.7061288632792038,
|
|
"grad_norm": 0.5772907462842918,
|
|
"learning_rate": 2.144075922140472e-05,
|
|
"loss": 0.224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08896462619304657,
|
|
"step": 7075,
|
|
"valid_targets_mean": 3349.1,
|
|
"valid_targets_min": 1598
|
|
},
|
|
{
|
|
"epoch": 3.7087480356207436,
|
|
"grad_norm": 0.5543791887751796,
|
|
"learning_rate": 2.1414702532860576e-05,
|
|
"loss": 0.2366,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09576257318258286,
|
|
"step": 7080,
|
|
"valid_targets_mean": 3761.4,
|
|
"valid_targets_min": 2267
|
|
},
|
|
{
|
|
"epoch": 3.711367207962284,
|
|
"grad_norm": 0.6546215629725258,
|
|
"learning_rate": 2.1388643430732396e-05,
|
|
"loss": 0.2152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1018991619348526,
|
|
"step": 7085,
|
|
"valid_targets_mean": 2590.0,
|
|
"valid_targets_min": 1279
|
|
},
|
|
{
|
|
"epoch": 3.713986380303824,
|
|
"grad_norm": 0.6337142610145704,
|
|
"learning_rate": 2.1362581959478877e-05,
|
|
"loss": 0.2172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11606380343437195,
|
|
"step": 7090,
|
|
"valid_targets_mean": 3247.0,
|
|
"valid_targets_min": 1730
|
|
},
|
|
{
|
|
"epoch": 3.716605552645364,
|
|
"grad_norm": 0.6550756025812633,
|
|
"learning_rate": 2.133651816356276e-05,
|
|
"loss": 0.2171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11128291487693787,
|
|
"step": 7095,
|
|
"valid_targets_mean": 3144.8,
|
|
"valid_targets_min": 1664
|
|
},
|
|
{
|
|
"epoch": 3.719224724986904,
|
|
"grad_norm": 0.6749025831860757,
|
|
"learning_rate": 2.1310452087450755e-05,
|
|
"loss": 0.2175,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08359639346599579,
|
|
"step": 7100,
|
|
"valid_targets_mean": 3659.5,
|
|
"valid_targets_min": 1945
|
|
},
|
|
{
|
|
"epoch": 3.7218438973284442,
|
|
"grad_norm": 0.5907450396294541,
|
|
"learning_rate": 2.1284383775613456e-05,
|
|
"loss": 0.2306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12242050468921661,
|
|
"step": 7105,
|
|
"valid_targets_mean": 4016.2,
|
|
"valid_targets_min": 2418
|
|
},
|
|
{
|
|
"epoch": 3.7244630696699845,
|
|
"grad_norm": 0.5766563890693965,
|
|
"learning_rate": 2.1258313272525277e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11721499264240265,
|
|
"step": 7110,
|
|
"valid_targets_mean": 4044.8,
|
|
"valid_targets_min": 2100
|
|
},
|
|
{
|
|
"epoch": 3.7270822420115244,
|
|
"grad_norm": 0.5765636261546505,
|
|
"learning_rate": 2.123224062266436e-05,
|
|
"loss": 0.2209,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1124374270439148,
|
|
"step": 7115,
|
|
"valid_targets_mean": 3743.8,
|
|
"valid_targets_min": 1365
|
|
},
|
|
{
|
|
"epoch": 3.7297014143530642,
|
|
"grad_norm": 0.554981450703291,
|
|
"learning_rate": 2.1206165870512522e-05,
|
|
"loss": 0.2239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10184729099273682,
|
|
"step": 7120,
|
|
"valid_targets_mean": 4017.1,
|
|
"valid_targets_min": 1572
|
|
},
|
|
{
|
|
"epoch": 3.7323205866946045,
|
|
"grad_norm": 0.6135910413795018,
|
|
"learning_rate": 2.1180089060555162e-05,
|
|
"loss": 0.2292,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10245160758495331,
|
|
"step": 7125,
|
|
"valid_targets_mean": 3171.4,
|
|
"valid_targets_min": 876
|
|
},
|
|
{
|
|
"epoch": 3.734939759036145,
|
|
"grad_norm": 0.4202717422025161,
|
|
"learning_rate": 2.115401023728118e-05,
|
|
"loss": 0.1219,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049300335347652435,
|
|
"step": 7130,
|
|
"valid_targets_mean": 2722.6,
|
|
"valid_targets_min": 458
|
|
},
|
|
{
|
|
"epoch": 3.7375589313776847,
|
|
"grad_norm": 0.42149866016405785,
|
|
"learning_rate": 2.112792944518293e-05,
|
|
"loss": 0.0921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0460713729262352,
|
|
"step": 7135,
|
|
"valid_targets_mean": 3113.1,
|
|
"valid_targets_min": 2461
|
|
},
|
|
{
|
|
"epoch": 3.7401781037192245,
|
|
"grad_norm": 0.375211518708931,
|
|
"learning_rate": 2.1101846728756107e-05,
|
|
"loss": 0.0844,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04961033910512924,
|
|
"step": 7140,
|
|
"valid_targets_mean": 3648.1,
|
|
"valid_targets_min": 2068
|
|
},
|
|
{
|
|
"epoch": 3.742797276060765,
|
|
"grad_norm": 0.43357057509450503,
|
|
"learning_rate": 2.1075762132499702e-05,
|
|
"loss": 0.1057,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049306970089673996,
|
|
"step": 7145,
|
|
"valid_targets_mean": 3423.6,
|
|
"valid_targets_min": 2425
|
|
},
|
|
{
|
|
"epoch": 3.745416448402305,
|
|
"grad_norm": 0.46401565751483315,
|
|
"learning_rate": 2.104967570091591e-05,
|
|
"loss": 0.085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0431523360311985,
|
|
"step": 7150,
|
|
"valid_targets_mean": 2724.1,
|
|
"valid_targets_min": 1198
|
|
},
|
|
{
|
|
"epoch": 3.748035620743845,
|
|
"grad_norm": 0.7317628730767393,
|
|
"learning_rate": 2.1023587478510045e-05,
|
|
"loss": 0.1008,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07258038967847824,
|
|
"step": 7155,
|
|
"valid_targets_mean": 1232.5,
|
|
"valid_targets_min": 807
|
|
},
|
|
{
|
|
"epoch": 3.750654793085385,
|
|
"grad_norm": 0.4452913466256158,
|
|
"learning_rate": 2.0997497509790498e-05,
|
|
"loss": 0.1387,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05453608185052872,
|
|
"step": 7160,
|
|
"valid_targets_mean": 3487.4,
|
|
"valid_targets_min": 2351
|
|
},
|
|
{
|
|
"epoch": 3.753273965426925,
|
|
"grad_norm": 0.47561872194048394,
|
|
"learning_rate": 2.097140583926862e-05,
|
|
"loss": 0.0938,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07181800901889801,
|
|
"step": 7165,
|
|
"valid_targets_mean": 3126.4,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 3.755893137768465,
|
|
"grad_norm": 0.34047866543595,
|
|
"learning_rate": 2.094531251145868e-05,
|
|
"loss": 0.1044,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046262361109256744,
|
|
"step": 7170,
|
|
"valid_targets_mean": 3474.8,
|
|
"valid_targets_min": 799
|
|
},
|
|
{
|
|
"epoch": 3.7585123101100053,
|
|
"grad_norm": 0.3093023961260679,
|
|
"learning_rate": 2.0919217570877758e-05,
|
|
"loss": 0.1091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04107516258955002,
|
|
"step": 7175,
|
|
"valid_targets_mean": 5026.1,
|
|
"valid_targets_min": 2802
|
|
},
|
|
{
|
|
"epoch": 3.761131482451545,
|
|
"grad_norm": 0.33879537081383754,
|
|
"learning_rate": 2.0893121062045708e-05,
|
|
"loss": 0.0827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032565630972385406,
|
|
"step": 7180,
|
|
"valid_targets_mean": 2611.0,
|
|
"valid_targets_min": 574
|
|
},
|
|
{
|
|
"epoch": 3.7637506547930855,
|
|
"grad_norm": 0.34388373179864956,
|
|
"learning_rate": 2.0867023029485032e-05,
|
|
"loss": 0.0901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042953215539455414,
|
|
"step": 7185,
|
|
"valid_targets_mean": 4036.5,
|
|
"valid_targets_min": 2642
|
|
},
|
|
{
|
|
"epoch": 3.7663698271346253,
|
|
"grad_norm": 0.5253315597352873,
|
|
"learning_rate": 2.0840923517720854e-05,
|
|
"loss": 0.0936,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06158360466361046,
|
|
"step": 7190,
|
|
"valid_targets_mean": 1960.5,
|
|
"valid_targets_min": 593
|
|
},
|
|
{
|
|
"epoch": 3.7689889994761656,
|
|
"grad_norm": 0.3903639599020791,
|
|
"learning_rate": 2.081482257128082e-05,
|
|
"loss": 0.1187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03658132255077362,
|
|
"step": 7195,
|
|
"valid_targets_mean": 3354.2,
|
|
"valid_targets_min": 1353
|
|
},
|
|
{
|
|
"epoch": 3.7716081718177055,
|
|
"grad_norm": 0.4866331067485609,
|
|
"learning_rate": 2.078872023469501e-05,
|
|
"loss": 0.0865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05128805711865425,
|
|
"step": 7200,
|
|
"valid_targets_mean": 2300.9,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 3.7742273441592458,
|
|
"grad_norm": 0.3786582391494195,
|
|
"learning_rate": 2.076261655249588e-05,
|
|
"loss": 0.086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030418336391448975,
|
|
"step": 7205,
|
|
"valid_targets_mean": 2295.8,
|
|
"valid_targets_min": 934
|
|
},
|
|
{
|
|
"epoch": 3.7768465165007856,
|
|
"grad_norm": 0.34771139894436154,
|
|
"learning_rate": 2.0736511569218186e-05,
|
|
"loss": 0.0787,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04290395230054855,
|
|
"step": 7210,
|
|
"valid_targets_mean": 3524.4,
|
|
"valid_targets_min": 2798
|
|
},
|
|
{
|
|
"epoch": 3.779465688842326,
|
|
"grad_norm": 0.4755408422218981,
|
|
"learning_rate": 2.071040532939891e-05,
|
|
"loss": 0.099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03995596989989281,
|
|
"step": 7215,
|
|
"valid_targets_mean": 2810.1,
|
|
"valid_targets_min": 471
|
|
},
|
|
{
|
|
"epoch": 3.7820848611838658,
|
|
"grad_norm": 0.4321925357201202,
|
|
"learning_rate": 2.0684297877577165e-05,
|
|
"loss": 0.1004,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06376747786998749,
|
|
"step": 7220,
|
|
"valid_targets_mean": 3042.8,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 3.784704033525406,
|
|
"grad_norm": 0.4072510935049014,
|
|
"learning_rate": 2.0658189258294144e-05,
|
|
"loss": 0.0996,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04013584181666374,
|
|
"step": 7225,
|
|
"valid_targets_mean": 3081.6,
|
|
"valid_targets_min": 474
|
|
},
|
|
{
|
|
"epoch": 3.787323205866946,
|
|
"grad_norm": 0.3969728795752201,
|
|
"learning_rate": 2.0632079516093015e-05,
|
|
"loss": 0.0999,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04627121612429619,
|
|
"step": 7230,
|
|
"valid_targets_mean": 3468.0,
|
|
"valid_targets_min": 640
|
|
},
|
|
{
|
|
"epoch": 3.7899423782084862,
|
|
"grad_norm": 0.3400736743267032,
|
|
"learning_rate": 2.0605968695518873e-05,
|
|
"loss": 0.1339,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03258595988154411,
|
|
"step": 7235,
|
|
"valid_targets_mean": 3438.4,
|
|
"valid_targets_min": 902
|
|
},
|
|
{
|
|
"epoch": 3.792561550550026,
|
|
"grad_norm": 0.4024576266908952,
|
|
"learning_rate": 2.057985684111866e-05,
|
|
"loss": 0.0842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042630430310964584,
|
|
"step": 7240,
|
|
"valid_targets_mean": 3212.8,
|
|
"valid_targets_min": 1837
|
|
},
|
|
{
|
|
"epoch": 3.7951807228915664,
|
|
"grad_norm": 0.6884395421314213,
|
|
"learning_rate": 2.0553743997441066e-05,
|
|
"loss": 0.1157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08674518764019012,
|
|
"step": 7245,
|
|
"valid_targets_mean": 1358.8,
|
|
"valid_targets_min": 756
|
|
},
|
|
{
|
|
"epoch": 3.7977998952331062,
|
|
"grad_norm": 0.5349913749235634,
|
|
"learning_rate": 2.0527630209036474e-05,
|
|
"loss": 0.1116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08277156949043274,
|
|
"step": 7250,
|
|
"valid_targets_mean": 2896.2,
|
|
"valid_targets_min": 1072
|
|
},
|
|
{
|
|
"epoch": 3.8004190675746465,
|
|
"grad_norm": 0.38558069149318935,
|
|
"learning_rate": 2.0501515520456895e-05,
|
|
"loss": 0.0913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02911374717950821,
|
|
"step": 7255,
|
|
"valid_targets_mean": 2908.1,
|
|
"valid_targets_min": 895
|
|
},
|
|
{
|
|
"epoch": 3.8030382399161864,
|
|
"grad_norm": 0.5790567115299913,
|
|
"learning_rate": 2.047539997625584e-05,
|
|
"loss": 0.2065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1842120736837387,
|
|
"step": 7260,
|
|
"valid_targets_mean": 2562.2,
|
|
"valid_targets_min": 878
|
|
},
|
|
{
|
|
"epoch": 3.8056574122577267,
|
|
"grad_norm": 0.4662391343390448,
|
|
"learning_rate": 2.0449283620988322e-05,
|
|
"loss": 0.1919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04233997315168381,
|
|
"step": 7265,
|
|
"valid_targets_mean": 2087.6,
|
|
"valid_targets_min": 756
|
|
},
|
|
{
|
|
"epoch": 3.8082765845992665,
|
|
"grad_norm": 0.34993598136122933,
|
|
"learning_rate": 2.04231664992107e-05,
|
|
"loss": 0.1022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.036809686571359634,
|
|
"step": 7270,
|
|
"valid_targets_mean": 3421.8,
|
|
"valid_targets_min": 2963
|
|
},
|
|
{
|
|
"epoch": 3.810895756940807,
|
|
"grad_norm": 0.7997591185983286,
|
|
"learning_rate": 2.039704865548067e-05,
|
|
"loss": 0.1721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22461523115634918,
|
|
"step": 7275,
|
|
"valid_targets_mean": 2375.2,
|
|
"valid_targets_min": 1153
|
|
},
|
|
{
|
|
"epoch": 3.8135149292823467,
|
|
"grad_norm": 0.3025477416479924,
|
|
"learning_rate": 2.0370930134357142e-05,
|
|
"loss": 0.1111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03618280589580536,
|
|
"step": 7280,
|
|
"valid_targets_mean": 4886.2,
|
|
"valid_targets_min": 1700
|
|
},
|
|
{
|
|
"epoch": 3.816134101623887,
|
|
"grad_norm": 0.3336682191169776,
|
|
"learning_rate": 2.0344810980400185e-05,
|
|
"loss": 0.0902,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02956889010965824,
|
|
"step": 7285,
|
|
"valid_targets_mean": 3965.1,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 3.818753273965427,
|
|
"grad_norm": 0.38169150485491554,
|
|
"learning_rate": 2.031869123817096e-05,
|
|
"loss": 0.0977,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050395525991916656,
|
|
"step": 7290,
|
|
"valid_targets_mean": 3771.5,
|
|
"valid_targets_min": 995
|
|
},
|
|
{
|
|
"epoch": 3.821372446306967,
|
|
"grad_norm": 0.3567381546882948,
|
|
"learning_rate": 2.029257095223161e-05,
|
|
"loss": 0.087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030895696952939034,
|
|
"step": 7295,
|
|
"valid_targets_mean": 2220.8,
|
|
"valid_targets_min": 488
|
|
},
|
|
{
|
|
"epoch": 3.823991618648507,
|
|
"grad_norm": 0.45080057733047246,
|
|
"learning_rate": 2.026645016714522e-05,
|
|
"loss": 0.0887,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05217716470360756,
|
|
"step": 7300,
|
|
"valid_targets_mean": 3635.4,
|
|
"valid_targets_min": 2437
|
|
},
|
|
{
|
|
"epoch": 3.8266107909900473,
|
|
"grad_norm": 0.48102035649069846,
|
|
"learning_rate": 2.024032892747573e-05,
|
|
"loss": 0.0846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03712937980890274,
|
|
"step": 7305,
|
|
"valid_targets_mean": 1752.1,
|
|
"valid_targets_min": 712
|
|
},
|
|
{
|
|
"epoch": 3.829229963331587,
|
|
"grad_norm": 0.39783111724258796,
|
|
"learning_rate": 2.021420727778784e-05,
|
|
"loss": 0.0868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04353148490190506,
|
|
"step": 7310,
|
|
"valid_targets_mean": 3713.4,
|
|
"valid_targets_min": 2422
|
|
},
|
|
{
|
|
"epoch": 3.831849135673127,
|
|
"grad_norm": 0.3553079012266278,
|
|
"learning_rate": 2.018808526264697e-05,
|
|
"loss": 0.0761,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04106612503528595,
|
|
"step": 7315,
|
|
"valid_targets_mean": 3998.5,
|
|
"valid_targets_min": 3643
|
|
},
|
|
{
|
|
"epoch": 3.8344683080146673,
|
|
"grad_norm": 0.3244256257152452,
|
|
"learning_rate": 2.0161962926619144e-05,
|
|
"loss": 0.1311,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030028818175196648,
|
|
"step": 7320,
|
|
"valid_targets_mean": 3319.0,
|
|
"valid_targets_min": 772
|
|
},
|
|
{
|
|
"epoch": 3.8370874803562076,
|
|
"grad_norm": 0.38063110072772477,
|
|
"learning_rate": 2.013584031427094e-05,
|
|
"loss": 0.0705,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029718592762947083,
|
|
"step": 7325,
|
|
"valid_targets_mean": 2934.8,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 3.8397066526977475,
|
|
"grad_norm": 0.31637284793444953,
|
|
"learning_rate": 2.010971747016941e-05,
|
|
"loss": 0.079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031116031110286713,
|
|
"step": 7330,
|
|
"valid_targets_mean": 3656.4,
|
|
"valid_targets_min": 2788
|
|
},
|
|
{
|
|
"epoch": 3.8423258250392873,
|
|
"grad_norm": 0.37898954142065566,
|
|
"learning_rate": 2.0083594438882006e-05,
|
|
"loss": 0.0805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04169466719031334,
|
|
"step": 7335,
|
|
"valid_targets_mean": 3202.5,
|
|
"valid_targets_min": 948
|
|
},
|
|
{
|
|
"epoch": 3.8449449973808276,
|
|
"grad_norm": 0.5617243089869247,
|
|
"learning_rate": 2.0057471264976493e-05,
|
|
"loss": 0.0983,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04837562143802643,
|
|
"step": 7340,
|
|
"valid_targets_mean": 1002.0,
|
|
"valid_targets_min": 490
|
|
},
|
|
{
|
|
"epoch": 3.847564169722368,
|
|
"grad_norm": 0.34839600190461567,
|
|
"learning_rate": 2.0031347993020876e-05,
|
|
"loss": 0.1101,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04791203886270523,
|
|
"step": 7345,
|
|
"valid_targets_mean": 3751.9,
|
|
"valid_targets_min": 2800
|
|
},
|
|
{
|
|
"epoch": 3.8501833420639078,
|
|
"grad_norm": 0.4885743778972602,
|
|
"learning_rate": 2.0005224667583337e-05,
|
|
"loss": 0.1192,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07227519154548645,
|
|
"step": 7350,
|
|
"valid_targets_mean": 2641.8,
|
|
"valid_targets_min": 896
|
|
},
|
|
{
|
|
"epoch": 3.8528025144054476,
|
|
"grad_norm": 0.4448499530668694,
|
|
"learning_rate": 1.9979101333232136e-05,
|
|
"loss": 0.0949,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052534934133291245,
|
|
"step": 7355,
|
|
"valid_targets_mean": 3199.0,
|
|
"valid_targets_min": 1333
|
|
},
|
|
{
|
|
"epoch": 3.855421686746988,
|
|
"grad_norm": 0.4615518720521214,
|
|
"learning_rate": 1.9952978034535556e-05,
|
|
"loss": 0.1093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04913472756743431,
|
|
"step": 7360,
|
|
"valid_targets_mean": 2548.1,
|
|
"valid_targets_min": 693
|
|
},
|
|
{
|
|
"epoch": 3.8580408590885282,
|
|
"grad_norm": 0.4005004069162051,
|
|
"learning_rate": 1.9926854816061833e-05,
|
|
"loss": 0.1243,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03859833627939224,
|
|
"step": 7365,
|
|
"valid_targets_mean": 2482.9,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 3.860660031430068,
|
|
"grad_norm": 0.4047788117435986,
|
|
"learning_rate": 1.990073172237904e-05,
|
|
"loss": 0.0931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04876485466957092,
|
|
"step": 7370,
|
|
"valid_targets_mean": 3239.2,
|
|
"valid_targets_min": 2525
|
|
},
|
|
{
|
|
"epoch": 3.863279203771608,
|
|
"grad_norm": 0.5068732170981666,
|
|
"learning_rate": 1.9874608798055045e-05,
|
|
"loss": 0.1149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04208771511912346,
|
|
"step": 7375,
|
|
"valid_targets_mean": 3806.1,
|
|
"valid_targets_min": 2393
|
|
},
|
|
{
|
|
"epoch": 3.8658983761131482,
|
|
"grad_norm": 0.3844447129019727,
|
|
"learning_rate": 1.9848486087657447e-05,
|
|
"loss": 0.0969,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042661719024181366,
|
|
"step": 7380,
|
|
"valid_targets_mean": 2990.9,
|
|
"valid_targets_min": 804
|
|
},
|
|
{
|
|
"epoch": 3.8685175484546885,
|
|
"grad_norm": 0.3774126041995829,
|
|
"learning_rate": 1.982236363575345e-05,
|
|
"loss": 0.0899,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04523811489343643,
|
|
"step": 7385,
|
|
"valid_targets_mean": 4885.8,
|
|
"valid_targets_min": 3695
|
|
},
|
|
{
|
|
"epoch": 3.8711367207962284,
|
|
"grad_norm": 0.43172860492708126,
|
|
"learning_rate": 1.9796241486909836e-05,
|
|
"loss": 0.0805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04847349226474762,
|
|
"step": 7390,
|
|
"valid_targets_mean": 3128.8,
|
|
"valid_targets_min": 596
|
|
},
|
|
{
|
|
"epoch": 3.8737558931377682,
|
|
"grad_norm": 0.2801146081181737,
|
|
"learning_rate": 1.9770119685692874e-05,
|
|
"loss": 0.0786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03003329038619995,
|
|
"step": 7395,
|
|
"valid_targets_mean": 4567.8,
|
|
"valid_targets_min": 3804
|
|
},
|
|
{
|
|
"epoch": 3.8763750654793085,
|
|
"grad_norm": 0.35307733798335517,
|
|
"learning_rate": 1.974399827666822e-05,
|
|
"loss": 0.0861,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03416905924677849,
|
|
"step": 7400,
|
|
"valid_targets_mean": 3853.1,
|
|
"valid_targets_min": 2862
|
|
},
|
|
{
|
|
"epoch": 3.878994237820849,
|
|
"grad_norm": 0.5552958434920389,
|
|
"learning_rate": 1.971787730440087e-05,
|
|
"loss": 0.1055,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06091686338186264,
|
|
"step": 7405,
|
|
"valid_targets_mean": 1072.6,
|
|
"valid_targets_min": 497
|
|
},
|
|
{
|
|
"epoch": 3.8816134101623887,
|
|
"grad_norm": 0.3890583830957186,
|
|
"learning_rate": 1.9691756813455084e-05,
|
|
"loss": 0.0897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05139869451522827,
|
|
"step": 7410,
|
|
"valid_targets_mean": 4191.0,
|
|
"valid_targets_min": 3918
|
|
},
|
|
{
|
|
"epoch": 3.8842325825039286,
|
|
"grad_norm": 0.5350479091654851,
|
|
"learning_rate": 1.9665636848394297e-05,
|
|
"loss": 0.101,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045283541083335876,
|
|
"step": 7415,
|
|
"valid_targets_mean": 2121.9,
|
|
"valid_targets_min": 785
|
|
},
|
|
{
|
|
"epoch": 3.886851754845469,
|
|
"grad_norm": 0.33836036040841005,
|
|
"learning_rate": 1.9639517453781032e-05,
|
|
"loss": 0.0734,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.036470964550971985,
|
|
"step": 7420,
|
|
"valid_targets_mean": 3451.6,
|
|
"valid_targets_min": 905
|
|
},
|
|
{
|
|
"epoch": 3.889470927187009,
|
|
"grad_norm": 0.4235972082823505,
|
|
"learning_rate": 1.961339867417687e-05,
|
|
"loss": 0.0953,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06840990483760834,
|
|
"step": 7425,
|
|
"valid_targets_mean": 2929.2,
|
|
"valid_targets_min": 657
|
|
},
|
|
{
|
|
"epoch": 3.892090099528549,
|
|
"grad_norm": 0.4199428163888667,
|
|
"learning_rate": 1.9587280554142307e-05,
|
|
"loss": 0.0887,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04536977410316467,
|
|
"step": 7430,
|
|
"valid_targets_mean": 3798.9,
|
|
"valid_targets_min": 3184
|
|
},
|
|
{
|
|
"epoch": 3.894709271870089,
|
|
"grad_norm": 0.49563760382145616,
|
|
"learning_rate": 1.9561163138236723e-05,
|
|
"loss": 0.079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04851203039288521,
|
|
"step": 7435,
|
|
"valid_targets_mean": 1463.9,
|
|
"valid_targets_min": 627
|
|
},
|
|
{
|
|
"epoch": 3.897328444211629,
|
|
"grad_norm": 0.46140585434038806,
|
|
"learning_rate": 1.9535046471018326e-05,
|
|
"loss": 0.0888,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04280813783407211,
|
|
"step": 7440,
|
|
"valid_targets_mean": 1701.5,
|
|
"valid_targets_min": 827
|
|
},
|
|
{
|
|
"epoch": 3.8999476165531695,
|
|
"grad_norm": 0.4696922128308981,
|
|
"learning_rate": 1.9508930597044e-05,
|
|
"loss": 0.0917,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04109931364655495,
|
|
"step": 7445,
|
|
"valid_targets_mean": 1789.9,
|
|
"valid_targets_min": 940
|
|
},
|
|
{
|
|
"epoch": 3.9025667888947093,
|
|
"grad_norm": 0.4046158647857266,
|
|
"learning_rate": 1.9482815560869327e-05,
|
|
"loss": 0.1068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039725400507450104,
|
|
"step": 7450,
|
|
"valid_targets_mean": 2943.8,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 3.905185961236249,
|
|
"grad_norm": 0.44635335729054926,
|
|
"learning_rate": 1.9456701407048414e-05,
|
|
"loss": 0.0879,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03700082004070282,
|
|
"step": 7455,
|
|
"valid_targets_mean": 1859.2,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 3.9078051335777895,
|
|
"grad_norm": 0.48368109190536873,
|
|
"learning_rate": 1.943058818013388e-05,
|
|
"loss": 0.1076,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034503936767578125,
|
|
"step": 7460,
|
|
"valid_targets_mean": 1703.9,
|
|
"valid_targets_min": 759
|
|
},
|
|
{
|
|
"epoch": 3.9104243059193298,
|
|
"grad_norm": 0.749090481722063,
|
|
"learning_rate": 1.9404475924676777e-05,
|
|
"loss": 0.1094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11215709894895554,
|
|
"step": 7465,
|
|
"valid_targets_mean": 1556.6,
|
|
"valid_targets_min": 521
|
|
},
|
|
{
|
|
"epoch": 3.9130434782608696,
|
|
"grad_norm": 0.43379214835409946,
|
|
"learning_rate": 1.9378364685226468e-05,
|
|
"loss": 0.0894,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05452614277601242,
|
|
"step": 7470,
|
|
"valid_targets_mean": 3267.6,
|
|
"valid_targets_min": 1032
|
|
},
|
|
{
|
|
"epoch": 3.9156626506024095,
|
|
"grad_norm": 0.30781486338337427,
|
|
"learning_rate": 1.9352254506330626e-05,
|
|
"loss": 0.0755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03051396831870079,
|
|
"step": 7475,
|
|
"valid_targets_mean": 3609.0,
|
|
"valid_targets_min": 2741
|
|
},
|
|
{
|
|
"epoch": 3.9182818229439498,
|
|
"grad_norm": 0.409820699570783,
|
|
"learning_rate": 1.932614543253506e-05,
|
|
"loss": 0.0868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.036954060196876526,
|
|
"step": 7480,
|
|
"valid_targets_mean": 2735.8,
|
|
"valid_targets_min": 1692
|
|
},
|
|
{
|
|
"epoch": 3.9209009952854896,
|
|
"grad_norm": 0.3983827381797914,
|
|
"learning_rate": 1.9300037508383753e-05,
|
|
"loss": 0.1165,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04921487346291542,
|
|
"step": 7485,
|
|
"valid_targets_mean": 3693.4,
|
|
"valid_targets_min": 2215
|
|
},
|
|
{
|
|
"epoch": 3.92352016762703,
|
|
"grad_norm": 0.34887223837531534,
|
|
"learning_rate": 1.9273930778418685e-05,
|
|
"loss": 0.0726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03230620175600052,
|
|
"step": 7490,
|
|
"valid_targets_mean": 3109.9,
|
|
"valid_targets_min": 723
|
|
},
|
|
{
|
|
"epoch": 3.92613933996857,
|
|
"grad_norm": 0.3142371236980945,
|
|
"learning_rate": 1.9247825287179796e-05,
|
|
"loss": 0.0883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03691818192601204,
|
|
"step": 7495,
|
|
"valid_targets_mean": 4984.2,
|
|
"valid_targets_min": 3282
|
|
},
|
|
{
|
|
"epoch": 3.92875851231011,
|
|
"grad_norm": 0.29035038233260285,
|
|
"learning_rate": 1.922172107920495e-05,
|
|
"loss": 0.0836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035114917904138565,
|
|
"step": 7500,
|
|
"valid_targets_mean": 3863.4,
|
|
"valid_targets_min": 995
|
|
},
|
|
{
|
|
"epoch": 3.93137768465165,
|
|
"grad_norm": 0.2625531451388003,
|
|
"learning_rate": 1.9195618199029783e-05,
|
|
"loss": 0.0714,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033638183027505875,
|
|
"step": 7505,
|
|
"valid_targets_mean": 4201.0,
|
|
"valid_targets_min": 949
|
|
},
|
|
{
|
|
"epoch": 3.9339968569931902,
|
|
"grad_norm": 0.4064411831354708,
|
|
"learning_rate": 1.916951669118769e-05,
|
|
"loss": 0.0934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058708321303129196,
|
|
"step": 7510,
|
|
"valid_targets_mean": 4321.6,
|
|
"valid_targets_min": 2724
|
|
},
|
|
{
|
|
"epoch": 3.93661602933473,
|
|
"grad_norm": 0.305752485102919,
|
|
"learning_rate": 1.9143416600209716e-05,
|
|
"loss": 0.0777,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02644570730626583,
|
|
"step": 7515,
|
|
"valid_targets_mean": 3384.5,
|
|
"valid_targets_min": 424
|
|
},
|
|
{
|
|
"epoch": 3.9392352016762704,
|
|
"grad_norm": 0.38059536630926644,
|
|
"learning_rate": 1.9117317970624484e-05,
|
|
"loss": 0.1106,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0355902835726738,
|
|
"step": 7520,
|
|
"valid_targets_mean": 2987.2,
|
|
"valid_targets_min": 531
|
|
},
|
|
{
|
|
"epoch": 3.9418543740178102,
|
|
"grad_norm": 0.3175284385746139,
|
|
"learning_rate": 1.9091220846958133e-05,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03525662422180176,
|
|
"step": 7525,
|
|
"valid_targets_mean": 3469.5,
|
|
"valid_targets_min": 1195
|
|
},
|
|
{
|
|
"epoch": 3.9444735463593505,
|
|
"grad_norm": 0.36278027512058986,
|
|
"learning_rate": 1.9065125273734228e-05,
|
|
"loss": 0.0737,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06611423939466476,
|
|
"step": 7530,
|
|
"valid_targets_mean": 2682.8,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 3.9470927187008904,
|
|
"grad_norm": 0.36773275928544025,
|
|
"learning_rate": 1.903903129547369e-05,
|
|
"loss": 0.1036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037536874413490295,
|
|
"step": 7535,
|
|
"valid_targets_mean": 3496.9,
|
|
"valid_targets_min": 1251
|
|
},
|
|
{
|
|
"epoch": 3.9497118910424307,
|
|
"grad_norm": 0.4047648503696616,
|
|
"learning_rate": 1.9012938956694714e-05,
|
|
"loss": 0.0834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04590514674782753,
|
|
"step": 7540,
|
|
"valid_targets_mean": 2912.5,
|
|
"valid_targets_min": 970
|
|
},
|
|
{
|
|
"epoch": 3.9523310633839706,
|
|
"grad_norm": 0.5394327284039387,
|
|
"learning_rate": 1.8986848301912707e-05,
|
|
"loss": 0.0954,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04667893052101135,
|
|
"step": 7545,
|
|
"valid_targets_mean": 1423.6,
|
|
"valid_targets_min": 634
|
|
},
|
|
{
|
|
"epoch": 3.954950235725511,
|
|
"grad_norm": 0.5774321003985943,
|
|
"learning_rate": 1.8960759375640204e-05,
|
|
"loss": 0.109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07255034148693085,
|
|
"step": 7550,
|
|
"valid_targets_mean": 1423.9,
|
|
"valid_targets_min": 685
|
|
},
|
|
{
|
|
"epoch": 3.9575694080670507,
|
|
"grad_norm": 0.3963183109928674,
|
|
"learning_rate": 1.8934672222386762e-05,
|
|
"loss": 0.1191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03373725712299347,
|
|
"step": 7555,
|
|
"valid_targets_mean": 2932.5,
|
|
"valid_targets_min": 593
|
|
},
|
|
{
|
|
"epoch": 3.960188580408591,
|
|
"grad_norm": 0.3216521588594565,
|
|
"learning_rate": 1.8908586886658967e-05,
|
|
"loss": 0.0908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033638566732406616,
|
|
"step": 7560,
|
|
"valid_targets_mean": 3945.8,
|
|
"valid_targets_min": 3129
|
|
},
|
|
{
|
|
"epoch": 3.962807752750131,
|
|
"grad_norm": 0.4168944532648461,
|
|
"learning_rate": 1.8882503412960247e-05,
|
|
"loss": 0.0939,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05136759579181671,
|
|
"step": 7565,
|
|
"valid_targets_mean": 3621.1,
|
|
"valid_targets_min": 859
|
|
},
|
|
{
|
|
"epoch": 3.965426925091671,
|
|
"grad_norm": 0.4678632826293506,
|
|
"learning_rate": 1.88564218457909e-05,
|
|
"loss": 0.0846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04501301422715187,
|
|
"step": 7570,
|
|
"valid_targets_mean": 2553.0,
|
|
"valid_targets_min": 616
|
|
},
|
|
{
|
|
"epoch": 3.968046097433211,
|
|
"grad_norm": 0.3954471615047967,
|
|
"learning_rate": 1.8830342229647945e-05,
|
|
"loss": 0.0826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03754996508359909,
|
|
"step": 7575,
|
|
"valid_targets_mean": 2690.5,
|
|
"valid_targets_min": 541
|
|
},
|
|
{
|
|
"epoch": 3.9706652697747513,
|
|
"grad_norm": 0.7106174824045168,
|
|
"learning_rate": 1.880426460902507e-05,
|
|
"loss": 0.0806,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05315547436475754,
|
|
"step": 7580,
|
|
"valid_targets_mean": 906.4,
|
|
"valid_targets_min": 598
|
|
},
|
|
{
|
|
"epoch": 3.973284442116291,
|
|
"grad_norm": 0.4165032943697711,
|
|
"learning_rate": 1.8778189028412574e-05,
|
|
"loss": 0.0972,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04184139892458916,
|
|
"step": 7585,
|
|
"valid_targets_mean": 3147.2,
|
|
"valid_targets_min": 1101
|
|
},
|
|
{
|
|
"epoch": 3.9759036144578315,
|
|
"grad_norm": 0.4802834714852666,
|
|
"learning_rate": 1.8752115532297263e-05,
|
|
"loss": 0.0789,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03538201004266739,
|
|
"step": 7590,
|
|
"valid_targets_mean": 2331.8,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 3.9785227867993713,
|
|
"grad_norm": 0.393894429443511,
|
|
"learning_rate": 1.8726044165162406e-05,
|
|
"loss": 0.0813,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04353872686624527,
|
|
"step": 7595,
|
|
"valid_targets_mean": 2577.0,
|
|
"valid_targets_min": 774
|
|
},
|
|
{
|
|
"epoch": 3.9811419591409116,
|
|
"grad_norm": 0.38984281411916594,
|
|
"learning_rate": 1.869997497148761e-05,
|
|
"loss": 0.0952,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050356339663267136,
|
|
"step": 7600,
|
|
"valid_targets_mean": 4140.1,
|
|
"valid_targets_min": 1474
|
|
},
|
|
{
|
|
"epoch": 3.9837611314824515,
|
|
"grad_norm": 0.2384552263579987,
|
|
"learning_rate": 1.8673907995748802e-05,
|
|
"loss": 0.0743,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02689472585916519,
|
|
"step": 7605,
|
|
"valid_targets_mean": 5498.9,
|
|
"valid_targets_min": 3677
|
|
},
|
|
{
|
|
"epoch": 3.9863803038239918,
|
|
"grad_norm": 0.38140640854743973,
|
|
"learning_rate": 1.8647843282418114e-05,
|
|
"loss": 0.082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04897107183933258,
|
|
"step": 7610,
|
|
"valid_targets_mean": 3352.5,
|
|
"valid_targets_min": 675
|
|
},
|
|
{
|
|
"epoch": 3.9889994761655316,
|
|
"grad_norm": 0.354715669859789,
|
|
"learning_rate": 1.86217808759638e-05,
|
|
"loss": 0.0788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03356082737445831,
|
|
"step": 7615,
|
|
"valid_targets_mean": 3902.5,
|
|
"valid_targets_min": 795
|
|
},
|
|
{
|
|
"epoch": 3.991618648507072,
|
|
"grad_norm": 0.5360196420481077,
|
|
"learning_rate": 1.8595720820850214e-05,
|
|
"loss": 0.1827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08176171779632568,
|
|
"step": 7620,
|
|
"valid_targets_mean": 2495.0,
|
|
"valid_targets_min": 605
|
|
},
|
|
{
|
|
"epoch": 3.994237820848612,
|
|
"grad_norm": 0.3272683951673967,
|
|
"learning_rate": 1.8569663161537665e-05,
|
|
"loss": 0.0778,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039716288447380066,
|
|
"step": 7625,
|
|
"valid_targets_mean": 3298.6,
|
|
"valid_targets_min": 985
|
|
},
|
|
{
|
|
"epoch": 3.9968569931901516,
|
|
"grad_norm": 0.4054114117257748,
|
|
"learning_rate": 1.854360794248241e-05,
|
|
"loss": 0.0915,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04544153809547424,
|
|
"step": 7630,
|
|
"valid_targets_mean": 3505.2,
|
|
"valid_targets_min": 878
|
|
},
|
|
{
|
|
"epoch": 3.999476165531692,
|
|
"grad_norm": 0.574865559161945,
|
|
"learning_rate": 1.8517555208136508e-05,
|
|
"loss": 0.1113,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12018486857414246,
|
|
"step": 7635,
|
|
"valid_targets_mean": 2289.5,
|
|
"valid_targets_min": 1214
|
|
},
|
|
{
|
|
"epoch": 4.002095337873232,
|
|
"grad_norm": 0.8147505408907869,
|
|
"learning_rate": 1.8491505002947784e-05,
|
|
"loss": 0.1637,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07107695192098618,
|
|
"step": 7640,
|
|
"valid_targets_mean": 1391.0,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 4.0047145102147725,
|
|
"grad_norm": 0.7834448703599444,
|
|
"learning_rate": 1.846545737135978e-05,
|
|
"loss": 0.1604,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0666496679186821,
|
|
"step": 7645,
|
|
"valid_targets_mean": 1389.0,
|
|
"valid_targets_min": 725
|
|
},
|
|
{
|
|
"epoch": 4.007333682556312,
|
|
"grad_norm": 0.7176476370684532,
|
|
"learning_rate": 1.8439412357811607e-05,
|
|
"loss": 0.1542,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07671309262514114,
|
|
"step": 7650,
|
|
"valid_targets_mean": 1432.1,
|
|
"valid_targets_min": 933
|
|
},
|
|
{
|
|
"epoch": 4.009952854897852,
|
|
"grad_norm": 0.7138731947033008,
|
|
"learning_rate": 1.8413370006737934e-05,
|
|
"loss": 0.1506,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07407839596271515,
|
|
"step": 7655,
|
|
"valid_targets_mean": 1396.4,
|
|
"valid_targets_min": 615
|
|
},
|
|
{
|
|
"epoch": 4.0125720272393925,
|
|
"grad_norm": 0.7590080786379892,
|
|
"learning_rate": 1.838733036256888e-05,
|
|
"loss": 0.1538,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07466895878314972,
|
|
"step": 7660,
|
|
"valid_targets_mean": 1552.1,
|
|
"valid_targets_min": 715
|
|
},
|
|
{
|
|
"epoch": 4.015191199580933,
|
|
"grad_norm": 0.832969840506494,
|
|
"learning_rate": 1.836129346972995e-05,
|
|
"loss": 0.1577,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11136734485626221,
|
|
"step": 7665,
|
|
"valid_targets_mean": 1899.1,
|
|
"valid_targets_min": 1090
|
|
},
|
|
{
|
|
"epoch": 4.017810371922472,
|
|
"grad_norm": 0.7439198656657051,
|
|
"learning_rate": 1.8335259372641945e-05,
|
|
"loss": 0.146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06305424869060516,
|
|
"step": 7670,
|
|
"valid_targets_mean": 1217.5,
|
|
"valid_targets_min": 761
|
|
},
|
|
{
|
|
"epoch": 4.0204295442640126,
|
|
"grad_norm": 0.6686768105488541,
|
|
"learning_rate": 1.8309228115720907e-05,
|
|
"loss": 0.1609,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06820094585418701,
|
|
"step": 7675,
|
|
"valid_targets_mean": 1600.6,
|
|
"valid_targets_min": 1182
|
|
},
|
|
{
|
|
"epoch": 4.023048716605553,
|
|
"grad_norm": 0.7986341260316357,
|
|
"learning_rate": 1.8283199743378028e-05,
|
|
"loss": 0.1496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07157409191131592,
|
|
"step": 7680,
|
|
"valid_targets_mean": 1476.8,
|
|
"valid_targets_min": 733
|
|
},
|
|
{
|
|
"epoch": 4.025667888947093,
|
|
"grad_norm": 0.7368932947097813,
|
|
"learning_rate": 1.8257174300019565e-05,
|
|
"loss": 0.156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06354066729545593,
|
|
"step": 7685,
|
|
"valid_targets_mean": 1622.1,
|
|
"valid_targets_min": 964
|
|
},
|
|
{
|
|
"epoch": 4.028287061288633,
|
|
"grad_norm": 0.7356153462819714,
|
|
"learning_rate": 1.8231151830046818e-05,
|
|
"loss": 0.1427,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08492037653923035,
|
|
"step": 7690,
|
|
"valid_targets_mean": 1614.6,
|
|
"valid_targets_min": 1046
|
|
},
|
|
{
|
|
"epoch": 4.030906233630173,
|
|
"grad_norm": 0.7331956418545665,
|
|
"learning_rate": 1.8205132377855974e-05,
|
|
"loss": 0.1558,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08211348950862885,
|
|
"step": 7695,
|
|
"valid_targets_mean": 1828.8,
|
|
"valid_targets_min": 1434
|
|
},
|
|
{
|
|
"epoch": 4.033525405971713,
|
|
"grad_norm": 0.742440518015762,
|
|
"learning_rate": 1.8179115987838078e-05,
|
|
"loss": 0.143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06464903056621552,
|
|
"step": 7700,
|
|
"valid_targets_mean": 1226.8,
|
|
"valid_targets_min": 570
|
|
},
|
|
{
|
|
"epoch": 4.036144578313253,
|
|
"grad_norm": 0.8092304133955073,
|
|
"learning_rate": 1.8153102704378975e-05,
|
|
"loss": 0.1421,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05798610299825668,
|
|
"step": 7705,
|
|
"valid_targets_mean": 1114.1,
|
|
"valid_targets_min": 548
|
|
},
|
|
{
|
|
"epoch": 4.038763750654793,
|
|
"grad_norm": 0.7007036076832147,
|
|
"learning_rate": 1.8127092571859173e-05,
|
|
"loss": 0.136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05507735162973404,
|
|
"step": 7710,
|
|
"valid_targets_mean": 1238.2,
|
|
"valid_targets_min": 682
|
|
},
|
|
{
|
|
"epoch": 4.041382922996333,
|
|
"grad_norm": 0.7570528581637733,
|
|
"learning_rate": 1.8101085634653846e-05,
|
|
"loss": 0.148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05682568997144699,
|
|
"step": 7715,
|
|
"valid_targets_mean": 1509.5,
|
|
"valid_targets_min": 968
|
|
},
|
|
{
|
|
"epoch": 4.0440020953378735,
|
|
"grad_norm": 0.7655927174644584,
|
|
"learning_rate": 1.8075081937132676e-05,
|
|
"loss": 0.1525,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07743413746356964,
|
|
"step": 7720,
|
|
"valid_targets_mean": 1428.2,
|
|
"valid_targets_min": 597
|
|
},
|
|
{
|
|
"epoch": 4.046621267679413,
|
|
"grad_norm": 0.7328719613273594,
|
|
"learning_rate": 1.8049081523659855e-05,
|
|
"loss": 0.1548,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07145942747592926,
|
|
"step": 7725,
|
|
"valid_targets_mean": 1484.6,
|
|
"valid_targets_min": 772
|
|
},
|
|
{
|
|
"epoch": 4.049240440020953,
|
|
"grad_norm": 0.722312852380765,
|
|
"learning_rate": 1.8023084438593945e-05,
|
|
"loss": 0.1473,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05349155515432358,
|
|
"step": 7730,
|
|
"valid_targets_mean": 1381.1,
|
|
"valid_targets_min": 978
|
|
},
|
|
{
|
|
"epoch": 4.0518596123624935,
|
|
"grad_norm": 0.6831881102657719,
|
|
"learning_rate": 1.7997090726287828e-05,
|
|
"loss": 0.1433,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05978317931294441,
|
|
"step": 7735,
|
|
"valid_targets_mean": 1295.9,
|
|
"valid_targets_min": 785
|
|
},
|
|
{
|
|
"epoch": 4.054478784704034,
|
|
"grad_norm": 0.7914198620249139,
|
|
"learning_rate": 1.797110043108866e-05,
|
|
"loss": 0.1497,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05559869110584259,
|
|
"step": 7740,
|
|
"valid_targets_mean": 1203.9,
|
|
"valid_targets_min": 617
|
|
},
|
|
{
|
|
"epoch": 4.057097957045573,
|
|
"grad_norm": 0.8442139413690831,
|
|
"learning_rate": 1.794511359733774e-05,
|
|
"loss": 0.1522,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06827697902917862,
|
|
"step": 7745,
|
|
"valid_targets_mean": 1205.5,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 4.0597171293871135,
|
|
"grad_norm": 0.7380528812620671,
|
|
"learning_rate": 1.791913026937048e-05,
|
|
"loss": 0.1439,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07087908685207367,
|
|
"step": 7750,
|
|
"valid_targets_mean": 1532.2,
|
|
"valid_targets_min": 1190
|
|
},
|
|
{
|
|
"epoch": 4.062336301728654,
|
|
"grad_norm": 0.7754559109397527,
|
|
"learning_rate": 1.789315049151629e-05,
|
|
"loss": 0.141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07318627834320068,
|
|
"step": 7755,
|
|
"valid_targets_mean": 1501.0,
|
|
"valid_targets_min": 774
|
|
},
|
|
{
|
|
"epoch": 4.064955474070194,
|
|
"grad_norm": 0.6970757617812192,
|
|
"learning_rate": 1.7867174308098537e-05,
|
|
"loss": 0.1424,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08933094888925552,
|
|
"step": 7760,
|
|
"valid_targets_mean": 2143.0,
|
|
"valid_targets_min": 1057
|
|
},
|
|
{
|
|
"epoch": 4.0675746464117335,
|
|
"grad_norm": 0.8299370161831886,
|
|
"learning_rate": 1.7841201763434465e-05,
|
|
"loss": 0.143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06878514587879181,
|
|
"step": 7765,
|
|
"valid_targets_mean": 1453.0,
|
|
"valid_targets_min": 1004
|
|
},
|
|
{
|
|
"epoch": 4.070193818753274,
|
|
"grad_norm": 0.7525883738507897,
|
|
"learning_rate": 1.781523290183508e-05,
|
|
"loss": 0.1465,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07721415162086487,
|
|
"step": 7770,
|
|
"valid_targets_mean": 1736.9,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 4.072812991094814,
|
|
"grad_norm": 0.6769824810891483,
|
|
"learning_rate": 1.7789267767605145e-05,
|
|
"loss": 0.1526,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06848867982625961,
|
|
"step": 7775,
|
|
"valid_targets_mean": 1477.9,
|
|
"valid_targets_min": 830
|
|
},
|
|
{
|
|
"epoch": 4.075432163436354,
|
|
"grad_norm": 0.7061942178506194,
|
|
"learning_rate": 1.7763306405043027e-05,
|
|
"loss": 0.1415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.062424588948488235,
|
|
"step": 7780,
|
|
"valid_targets_mean": 1463.9,
|
|
"valid_targets_min": 646
|
|
},
|
|
{
|
|
"epoch": 4.078051335777894,
|
|
"grad_norm": 0.7614928786964195,
|
|
"learning_rate": 1.773734885844068e-05,
|
|
"loss": 0.1496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07690142095088959,
|
|
"step": 7785,
|
|
"valid_targets_mean": 1431.2,
|
|
"valid_targets_min": 591
|
|
},
|
|
{
|
|
"epoch": 4.080670508119434,
|
|
"grad_norm": 0.7726308989574523,
|
|
"learning_rate": 1.7711395172083546e-05,
|
|
"loss": 0.1514,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09195318818092346,
|
|
"step": 7790,
|
|
"valid_targets_mean": 1579.6,
|
|
"valid_targets_min": 796
|
|
},
|
|
{
|
|
"epoch": 4.083289680460974,
|
|
"grad_norm": 0.7709271587907439,
|
|
"learning_rate": 1.768544539025046e-05,
|
|
"loss": 0.1432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07206501066684723,
|
|
"step": 7795,
|
|
"valid_targets_mean": 1642.8,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 4.085908852802515,
|
|
"grad_norm": 0.7563531244723153,
|
|
"learning_rate": 1.7659499557213624e-05,
|
|
"loss": 0.1408,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07513036578893661,
|
|
"step": 7800,
|
|
"valid_targets_mean": 1417.9,
|
|
"valid_targets_min": 665
|
|
},
|
|
{
|
|
"epoch": 4.088528025144054,
|
|
"grad_norm": 0.7261910183527875,
|
|
"learning_rate": 1.763355771723849e-05,
|
|
"loss": 0.14,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06454315036535263,
|
|
"step": 7805,
|
|
"valid_targets_mean": 1389.1,
|
|
"valid_targets_min": 652
|
|
},
|
|
{
|
|
"epoch": 4.091147197485594,
|
|
"grad_norm": 0.8458040723209423,
|
|
"learning_rate": 1.76076199145837e-05,
|
|
"loss": 0.141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058387018740177155,
|
|
"step": 7810,
|
|
"valid_targets_mean": 1289.8,
|
|
"valid_targets_min": 662
|
|
},
|
|
{
|
|
"epoch": 4.093766369827135,
|
|
"grad_norm": 0.7930638210354205,
|
|
"learning_rate": 1.7581686193500997e-05,
|
|
"loss": 0.1435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06643882393836975,
|
|
"step": 7815,
|
|
"valid_targets_mean": 1264.9,
|
|
"valid_targets_min": 770
|
|
},
|
|
{
|
|
"epoch": 4.096385542168675,
|
|
"grad_norm": 0.8133522359187357,
|
|
"learning_rate": 1.7555756598235178e-05,
|
|
"loss": 0.1398,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08015811443328857,
|
|
"step": 7820,
|
|
"valid_targets_mean": 1399.9,
|
|
"valid_targets_min": 966
|
|
},
|
|
{
|
|
"epoch": 4.099004714510214,
|
|
"grad_norm": 0.7293726199024233,
|
|
"learning_rate": 1.7529831173023994e-05,
|
|
"loss": 0.1427,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06379858404397964,
|
|
"step": 7825,
|
|
"valid_targets_mean": 1443.0,
|
|
"valid_targets_min": 773
|
|
},
|
|
{
|
|
"epoch": 4.101623886851755,
|
|
"grad_norm": 0.7568027073881571,
|
|
"learning_rate": 1.7503909962098075e-05,
|
|
"loss": 0.1413,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06921360641717911,
|
|
"step": 7830,
|
|
"valid_targets_mean": 1566.1,
|
|
"valid_targets_min": 940
|
|
},
|
|
{
|
|
"epoch": 4.104243059193295,
|
|
"grad_norm": 0.805512965299132,
|
|
"learning_rate": 1.7477993009680874e-05,
|
|
"loss": 0.1459,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05086955055594444,
|
|
"step": 7835,
|
|
"valid_targets_mean": 1089.4,
|
|
"valid_targets_min": 801
|
|
},
|
|
{
|
|
"epoch": 4.106862231534835,
|
|
"grad_norm": 0.7092495943845473,
|
|
"learning_rate": 1.745208035998857e-05,
|
|
"loss": 0.1381,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08520159125328064,
|
|
"step": 7840,
|
|
"valid_targets_mean": 1911.8,
|
|
"valid_targets_min": 634
|
|
},
|
|
{
|
|
"epoch": 4.109481403876375,
|
|
"grad_norm": 0.7503348918284387,
|
|
"learning_rate": 1.7426172057229987e-05,
|
|
"loss": 0.1511,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09396681934595108,
|
|
"step": 7845,
|
|
"valid_targets_mean": 1958.2,
|
|
"valid_targets_min": 919
|
|
},
|
|
{
|
|
"epoch": 4.112100576217915,
|
|
"grad_norm": 0.705556169148339,
|
|
"learning_rate": 1.7400268145606576e-05,
|
|
"loss": 0.1431,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08061669766902924,
|
|
"step": 7850,
|
|
"valid_targets_mean": 1991.5,
|
|
"valid_targets_min": 1102
|
|
},
|
|
{
|
|
"epoch": 4.114719748559455,
|
|
"grad_norm": 0.7455599835269692,
|
|
"learning_rate": 1.7374368669312245e-05,
|
|
"loss": 0.1481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05526871606707573,
|
|
"step": 7855,
|
|
"valid_targets_mean": 1266.5,
|
|
"valid_targets_min": 841
|
|
},
|
|
{
|
|
"epoch": 4.117338920900996,
|
|
"grad_norm": 0.7808324338281639,
|
|
"learning_rate": 1.7348473672533377e-05,
|
|
"loss": 0.1382,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05895335227251053,
|
|
"step": 7860,
|
|
"valid_targets_mean": 1280.8,
|
|
"valid_targets_min": 956
|
|
},
|
|
{
|
|
"epoch": 4.119958093242535,
|
|
"grad_norm": 0.8609518643491395,
|
|
"learning_rate": 1.732258319944868e-05,
|
|
"loss": 0.1439,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06586974114179611,
|
|
"step": 7865,
|
|
"valid_targets_mean": 1224.2,
|
|
"valid_targets_min": 857
|
|
},
|
|
{
|
|
"epoch": 4.122577265584075,
|
|
"grad_norm": 0.7522490704276529,
|
|
"learning_rate": 1.729669729422917e-05,
|
|
"loss": 0.1456,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07322436571121216,
|
|
"step": 7870,
|
|
"valid_targets_mean": 1312.2,
|
|
"valid_targets_min": 808
|
|
},
|
|
{
|
|
"epoch": 4.125196437925616,
|
|
"grad_norm": 0.7868336422863321,
|
|
"learning_rate": 1.727081600103806e-05,
|
|
"loss": 0.1342,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06527800858020782,
|
|
"step": 7875,
|
|
"valid_targets_mean": 1253.1,
|
|
"valid_targets_min": 817
|
|
},
|
|
{
|
|
"epoch": 4.127815610267156,
|
|
"grad_norm": 0.7608696113993708,
|
|
"learning_rate": 1.7244939364030676e-05,
|
|
"loss": 0.1365,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057997383177280426,
|
|
"step": 7880,
|
|
"valid_targets_mean": 1261.5,
|
|
"valid_targets_min": 727
|
|
},
|
|
{
|
|
"epoch": 4.130434782608695,
|
|
"grad_norm": 0.746396394408531,
|
|
"learning_rate": 1.7219067427354436e-05,
|
|
"loss": 0.1426,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0750526711344719,
|
|
"step": 7885,
|
|
"valid_targets_mean": 1557.1,
|
|
"valid_targets_min": 907
|
|
},
|
|
{
|
|
"epoch": 4.133053954950236,
|
|
"grad_norm": 0.7734343498112959,
|
|
"learning_rate": 1.7193200235148706e-05,
|
|
"loss": 0.1436,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0678153932094574,
|
|
"step": 7890,
|
|
"valid_targets_mean": 1369.2,
|
|
"valid_targets_min": 669
|
|
},
|
|
{
|
|
"epoch": 4.135673127291776,
|
|
"grad_norm": 0.688620854608366,
|
|
"learning_rate": 1.716733783154479e-05,
|
|
"loss": 0.1434,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07749447226524353,
|
|
"step": 7895,
|
|
"valid_targets_mean": 1794.9,
|
|
"valid_targets_min": 750
|
|
},
|
|
{
|
|
"epoch": 4.138292299633316,
|
|
"grad_norm": 0.7462520237842121,
|
|
"learning_rate": 1.7141480260665794e-05,
|
|
"loss": 0.1382,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06762084364891052,
|
|
"step": 7900,
|
|
"valid_targets_mean": 1542.2,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 4.140911471974856,
|
|
"grad_norm": 0.7705392056294689,
|
|
"learning_rate": 1.711562756662658e-05,
|
|
"loss": 0.1449,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08405694365501404,
|
|
"step": 7905,
|
|
"valid_targets_mean": 1618.6,
|
|
"valid_targets_min": 694
|
|
},
|
|
{
|
|
"epoch": 4.143530644316396,
|
|
"grad_norm": 0.7401935877883663,
|
|
"learning_rate": 1.7089779793533716e-05,
|
|
"loss": 0.1409,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08615291118621826,
|
|
"step": 7910,
|
|
"valid_targets_mean": 1725.2,
|
|
"valid_targets_min": 806
|
|
},
|
|
{
|
|
"epoch": 4.146149816657936,
|
|
"grad_norm": 0.8661918661157276,
|
|
"learning_rate": 1.706393698548534e-05,
|
|
"loss": 0.1437,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06610341370105743,
|
|
"step": 7915,
|
|
"valid_targets_mean": 1304.5,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 4.1487689889994765,
|
|
"grad_norm": 0.724989948253216,
|
|
"learning_rate": 1.703809918657116e-05,
|
|
"loss": 0.1404,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04395994916558266,
|
|
"step": 7920,
|
|
"valid_targets_mean": 1106.8,
|
|
"valid_targets_min": 698
|
|
},
|
|
{
|
|
"epoch": 4.151388161341016,
|
|
"grad_norm": 0.7846331292623914,
|
|
"learning_rate": 1.7012266440872294e-05,
|
|
"loss": 0.1322,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0646669939160347,
|
|
"step": 7925,
|
|
"valid_targets_mean": 1276.2,
|
|
"valid_targets_min": 750
|
|
},
|
|
{
|
|
"epoch": 4.154007333682556,
|
|
"grad_norm": 0.822275111676299,
|
|
"learning_rate": 1.6986438792461272e-05,
|
|
"loss": 0.143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09048368781805038,
|
|
"step": 7930,
|
|
"valid_targets_mean": 1869.1,
|
|
"valid_targets_min": 1003
|
|
},
|
|
{
|
|
"epoch": 4.156626506024097,
|
|
"grad_norm": 0.7688287932415243,
|
|
"learning_rate": 1.696061628540191e-05,
|
|
"loss": 0.1454,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0764274001121521,
|
|
"step": 7935,
|
|
"valid_targets_mean": 1735.8,
|
|
"valid_targets_min": 907
|
|
},
|
|
{
|
|
"epoch": 4.159245678365637,
|
|
"grad_norm": 0.7603343008442301,
|
|
"learning_rate": 1.6934798963749256e-05,
|
|
"loss": 0.1375,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0774686336517334,
|
|
"step": 7940,
|
|
"valid_targets_mean": 1505.4,
|
|
"valid_targets_min": 1160
|
|
},
|
|
{
|
|
"epoch": 4.161864850707176,
|
|
"grad_norm": 0.7324817066259853,
|
|
"learning_rate": 1.6908986871549525e-05,
|
|
"loss": 0.1445,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07530603557825089,
|
|
"step": 7945,
|
|
"valid_targets_mean": 1562.1,
|
|
"valid_targets_min": 1119
|
|
},
|
|
{
|
|
"epoch": 4.164484023048717,
|
|
"grad_norm": 0.7355857744107196,
|
|
"learning_rate": 1.6883180052839975e-05,
|
|
"loss": 0.1443,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08236520737409592,
|
|
"step": 7950,
|
|
"valid_targets_mean": 1615.5,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 4.167103195390257,
|
|
"grad_norm": 0.7494770260259638,
|
|
"learning_rate": 1.6857378551648918e-05,
|
|
"loss": 0.141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055171504616737366,
|
|
"step": 7955,
|
|
"valid_targets_mean": 1330.6,
|
|
"valid_targets_min": 859
|
|
},
|
|
{
|
|
"epoch": 4.169722367731797,
|
|
"grad_norm": 0.8679246068657159,
|
|
"learning_rate": 1.6831582411995556e-05,
|
|
"loss": 0.1459,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06985874474048615,
|
|
"step": 7960,
|
|
"valid_targets_mean": 1289.4,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 4.172341540073337,
|
|
"grad_norm": 0.8015833114321246,
|
|
"learning_rate": 1.680579167788994e-05,
|
|
"loss": 0.1371,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08213045448064804,
|
|
"step": 7965,
|
|
"valid_targets_mean": 1576.4,
|
|
"valid_targets_min": 673
|
|
},
|
|
{
|
|
"epoch": 4.174960712414877,
|
|
"grad_norm": 0.8100547386034976,
|
|
"learning_rate": 1.6780006393332934e-05,
|
|
"loss": 0.1468,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08223119378089905,
|
|
"step": 7970,
|
|
"valid_targets_mean": 1317.2,
|
|
"valid_targets_min": 851
|
|
},
|
|
{
|
|
"epoch": 4.177579884756417,
|
|
"grad_norm": 0.8021779391612973,
|
|
"learning_rate": 1.675422660231607e-05,
|
|
"loss": 0.1365,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0749107152223587,
|
|
"step": 7975,
|
|
"valid_targets_mean": 1581.0,
|
|
"valid_targets_min": 870
|
|
},
|
|
{
|
|
"epoch": 4.1801990570979575,
|
|
"grad_norm": 0.7954715334575904,
|
|
"learning_rate": 1.672845234882153e-05,
|
|
"loss": 0.1545,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10032214224338531,
|
|
"step": 7980,
|
|
"valid_targets_mean": 1962.4,
|
|
"valid_targets_min": 1156
|
|
},
|
|
{
|
|
"epoch": 4.182818229439497,
|
|
"grad_norm": 0.7187552218621887,
|
|
"learning_rate": 1.6702683676822028e-05,
|
|
"loss": 0.1404,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054006561636924744,
|
|
"step": 7985,
|
|
"valid_targets_mean": 1244.0,
|
|
"valid_targets_min": 657
|
|
},
|
|
{
|
|
"epoch": 4.185437401781037,
|
|
"grad_norm": 0.7287158136368013,
|
|
"learning_rate": 1.6676920630280785e-05,
|
|
"loss": 0.144,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07185535877943039,
|
|
"step": 7990,
|
|
"valid_targets_mean": 1536.6,
|
|
"valid_targets_min": 672
|
|
},
|
|
{
|
|
"epoch": 4.1880565741225775,
|
|
"grad_norm": 0.7539873725776786,
|
|
"learning_rate": 1.66511632531514e-05,
|
|
"loss": 0.1401,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08302894979715347,
|
|
"step": 7995,
|
|
"valid_targets_mean": 1686.9,
|
|
"valid_targets_min": 1019
|
|
},
|
|
{
|
|
"epoch": 4.190675746464118,
|
|
"grad_norm": 0.8157626078261773,
|
|
"learning_rate": 1.6625411589377795e-05,
|
|
"loss": 0.1364,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06138870120048523,
|
|
"step": 8000,
|
|
"valid_targets_mean": 1172.5,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 4.193294918805657,
|
|
"grad_norm": 0.8008925323806418,
|
|
"learning_rate": 1.659966568289417e-05,
|
|
"loss": 0.1448,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05657664313912392,
|
|
"step": 8005,
|
|
"valid_targets_mean": 1255.4,
|
|
"valid_targets_min": 634
|
|
},
|
|
{
|
|
"epoch": 4.1959140911471975,
|
|
"grad_norm": 0.7236013504321113,
|
|
"learning_rate": 1.6573925577624885e-05,
|
|
"loss": 0.1345,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07120637595653534,
|
|
"step": 8010,
|
|
"valid_targets_mean": 1793.1,
|
|
"valid_targets_min": 905
|
|
},
|
|
{
|
|
"epoch": 4.198533263488738,
|
|
"grad_norm": 0.7366915314972589,
|
|
"learning_rate": 1.6548191317484405e-05,
|
|
"loss": 0.1337,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08546614646911621,
|
|
"step": 8015,
|
|
"valid_targets_mean": 1636.1,
|
|
"valid_targets_min": 596
|
|
},
|
|
{
|
|
"epoch": 4.201152435830277,
|
|
"grad_norm": 0.7310244741400327,
|
|
"learning_rate": 1.6522462946377233e-05,
|
|
"loss": 0.1335,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06330102682113647,
|
|
"step": 8020,
|
|
"valid_targets_mean": 1320.8,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 4.2037716081718175,
|
|
"grad_norm": 0.7555194632894897,
|
|
"learning_rate": 1.6496740508197797e-05,
|
|
"loss": 0.1477,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06768176704645157,
|
|
"step": 8025,
|
|
"valid_targets_mean": 1494.8,
|
|
"valid_targets_min": 820
|
|
},
|
|
{
|
|
"epoch": 4.206390780513358,
|
|
"grad_norm": 0.7265269483373465,
|
|
"learning_rate": 1.647102404683044e-05,
|
|
"loss": 0.1384,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07998554408550262,
|
|
"step": 8030,
|
|
"valid_targets_mean": 2009.6,
|
|
"valid_targets_min": 1256
|
|
},
|
|
{
|
|
"epoch": 4.209009952854898,
|
|
"grad_norm": 0.7598587392455949,
|
|
"learning_rate": 1.6445313606149278e-05,
|
|
"loss": 0.1365,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07028426229953766,
|
|
"step": 8035,
|
|
"valid_targets_mean": 1477.8,
|
|
"valid_targets_min": 1122
|
|
},
|
|
{
|
|
"epoch": 4.2116291251964375,
|
|
"grad_norm": 0.7664829569108801,
|
|
"learning_rate": 1.6419609230018183e-05,
|
|
"loss": 0.1348,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06817641854286194,
|
|
"step": 8040,
|
|
"valid_targets_mean": 1564.5,
|
|
"valid_targets_min": 850
|
|
},
|
|
{
|
|
"epoch": 4.214248297537978,
|
|
"grad_norm": 0.7234625660676817,
|
|
"learning_rate": 1.639391096229065e-05,
|
|
"loss": 0.1262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07557559013366699,
|
|
"step": 8045,
|
|
"valid_targets_mean": 1746.5,
|
|
"valid_targets_min": 1389
|
|
},
|
|
{
|
|
"epoch": 4.216867469879518,
|
|
"grad_norm": 0.7745049628839951,
|
|
"learning_rate": 1.6368218846809777e-05,
|
|
"loss": 0.1412,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0712113231420517,
|
|
"step": 8050,
|
|
"valid_targets_mean": 1474.5,
|
|
"valid_targets_min": 1196
|
|
},
|
|
{
|
|
"epoch": 4.219486642221058,
|
|
"grad_norm": 0.7587287795680999,
|
|
"learning_rate": 1.6342532927408153e-05,
|
|
"loss": 0.1386,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07678717374801636,
|
|
"step": 8055,
|
|
"valid_targets_mean": 1668.9,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 4.222105814562598,
|
|
"grad_norm": 0.8432068094692399,
|
|
"learning_rate": 1.63168532479078e-05,
|
|
"loss": 0.1399,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07342837005853653,
|
|
"step": 8060,
|
|
"valid_targets_mean": 1506.5,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 4.224724986904138,
|
|
"grad_norm": 0.6838792512397553,
|
|
"learning_rate": 1.6291179852120095e-05,
|
|
"loss": 0.1407,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07926402986049652,
|
|
"step": 8065,
|
|
"valid_targets_mean": 1890.0,
|
|
"valid_targets_min": 1399
|
|
},
|
|
{
|
|
"epoch": 4.227344159245678,
|
|
"grad_norm": 0.7674603399756434,
|
|
"learning_rate": 1.626551278384569e-05,
|
|
"loss": 0.1387,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07749255001544952,
|
|
"step": 8070,
|
|
"valid_targets_mean": 1633.8,
|
|
"valid_targets_min": 594
|
|
},
|
|
{
|
|
"epoch": 4.229963331587219,
|
|
"grad_norm": 0.7177674834291355,
|
|
"learning_rate": 1.623985208687445e-05,
|
|
"loss": 0.141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0683123916387558,
|
|
"step": 8075,
|
|
"valid_targets_mean": 1847.6,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 4.232582503928758,
|
|
"grad_norm": 0.6146871340342175,
|
|
"learning_rate": 1.6214197804985364e-05,
|
|
"loss": 0.1738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08079074323177338,
|
|
"step": 8080,
|
|
"valid_targets_mean": 6074.6,
|
|
"valid_targets_min": 4960
|
|
},
|
|
{
|
|
"epoch": 4.235201676270298,
|
|
"grad_norm": 0.40635000420186906,
|
|
"learning_rate": 1.6188549981946475e-05,
|
|
"loss": 0.1641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07531790435314178,
|
|
"step": 8085,
|
|
"valid_targets_mean": 7117.6,
|
|
"valid_targets_min": 4541
|
|
},
|
|
{
|
|
"epoch": 4.237820848611839,
|
|
"grad_norm": 0.41096452601330413,
|
|
"learning_rate": 1.616290866151481e-05,
|
|
"loss": 0.1583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08659469336271286,
|
|
"step": 8090,
|
|
"valid_targets_mean": 6001.4,
|
|
"valid_targets_min": 4245
|
|
},
|
|
{
|
|
"epoch": 4.240440020953379,
|
|
"grad_norm": 0.35428357067602523,
|
|
"learning_rate": 1.6137273887436302e-05,
|
|
"loss": 0.1585,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0781104564666748,
|
|
"step": 8095,
|
|
"valid_targets_mean": 7032.8,
|
|
"valid_targets_min": 4108
|
|
},
|
|
{
|
|
"epoch": 4.243059193294918,
|
|
"grad_norm": 0.37198233829614286,
|
|
"learning_rate": 1.611164570344572e-05,
|
|
"loss": 0.1656,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09088402986526489,
|
|
"step": 8100,
|
|
"valid_targets_mean": 7311.5,
|
|
"valid_targets_min": 4689
|
|
},
|
|
{
|
|
"epoch": 4.245678365636459,
|
|
"grad_norm": 0.40630240395512573,
|
|
"learning_rate": 1.6086024153266574e-05,
|
|
"loss": 0.1666,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09036040306091309,
|
|
"step": 8105,
|
|
"valid_targets_mean": 6030.6,
|
|
"valid_targets_min": 4398
|
|
},
|
|
{
|
|
"epoch": 4.248297537977999,
|
|
"grad_norm": 0.3658256328381038,
|
|
"learning_rate": 1.6060409280611083e-05,
|
|
"loss": 0.1634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07792043685913086,
|
|
"step": 8110,
|
|
"valid_targets_mean": 6799.2,
|
|
"valid_targets_min": 4826
|
|
},
|
|
{
|
|
"epoch": 4.250916710319539,
|
|
"grad_norm": 0.3865735745584883,
|
|
"learning_rate": 1.6034801129180047e-05,
|
|
"loss": 0.1613,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08830088376998901,
|
|
"step": 8115,
|
|
"valid_targets_mean": 6354.9,
|
|
"valid_targets_min": 4525
|
|
},
|
|
{
|
|
"epoch": 4.253535882661079,
|
|
"grad_norm": 0.36501909862798587,
|
|
"learning_rate": 1.6009199742662804e-05,
|
|
"loss": 0.1724,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07458247244358063,
|
|
"step": 8120,
|
|
"valid_targets_mean": 7312.4,
|
|
"valid_targets_min": 4970
|
|
},
|
|
{
|
|
"epoch": 4.256155055002619,
|
|
"grad_norm": 0.3527712975682088,
|
|
"learning_rate": 1.5983605164737172e-05,
|
|
"loss": 0.1558,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06650740653276443,
|
|
"step": 8125,
|
|
"valid_targets_mean": 6569.9,
|
|
"valid_targets_min": 4552
|
|
},
|
|
{
|
|
"epoch": 4.258774227344159,
|
|
"grad_norm": 0.359774858131024,
|
|
"learning_rate": 1.595801743906932e-05,
|
|
"loss": 0.1703,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07978302240371704,
|
|
"step": 8130,
|
|
"valid_targets_mean": 7362.2,
|
|
"valid_targets_min": 3009
|
|
},
|
|
{
|
|
"epoch": 4.2613933996857,
|
|
"grad_norm": 0.35719891992865205,
|
|
"learning_rate": 1.5932436609313762e-05,
|
|
"loss": 0.1529,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.080568328499794,
|
|
"step": 8135,
|
|
"valid_targets_mean": 7558.4,
|
|
"valid_targets_min": 5304
|
|
},
|
|
{
|
|
"epoch": 4.264012572027239,
|
|
"grad_norm": 0.371634257534297,
|
|
"learning_rate": 1.5906862719113223e-05,
|
|
"loss": 0.1606,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07417374849319458,
|
|
"step": 8140,
|
|
"valid_targets_mean": 5596.0,
|
|
"valid_targets_min": 3815
|
|
},
|
|
{
|
|
"epoch": 4.266631744368779,
|
|
"grad_norm": 0.3557170988642468,
|
|
"learning_rate": 1.5881295812098584e-05,
|
|
"loss": 0.1447,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06920690834522247,
|
|
"step": 8145,
|
|
"valid_targets_mean": 6001.6,
|
|
"valid_targets_min": 4084
|
|
},
|
|
{
|
|
"epoch": 4.26925091671032,
|
|
"grad_norm": 0.4126654897990094,
|
|
"learning_rate": 1.5855735931888837e-05,
|
|
"loss": 0.147,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08121057599782944,
|
|
"step": 8150,
|
|
"valid_targets_mean": 8326.1,
|
|
"valid_targets_min": 4939
|
|
},
|
|
{
|
|
"epoch": 4.27187008905186,
|
|
"grad_norm": 0.35655938026525325,
|
|
"learning_rate": 1.583018312209096e-05,
|
|
"loss": 0.1612,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07522504031658173,
|
|
"step": 8155,
|
|
"valid_targets_mean": 6924.6,
|
|
"valid_targets_min": 5701
|
|
},
|
|
{
|
|
"epoch": 4.274489261393399,
|
|
"grad_norm": 0.34066722460043874,
|
|
"learning_rate": 1.5804637426299894e-05,
|
|
"loss": 0.1573,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05872965604066849,
|
|
"step": 8160,
|
|
"valid_targets_mean": 6086.9,
|
|
"valid_targets_min": 4384
|
|
},
|
|
{
|
|
"epoch": 4.27710843373494,
|
|
"grad_norm": 0.3492835418379819,
|
|
"learning_rate": 1.5779098888098413e-05,
|
|
"loss": 0.1438,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08275985717773438,
|
|
"step": 8165,
|
|
"valid_targets_mean": 7030.6,
|
|
"valid_targets_min": 4823
|
|
},
|
|
{
|
|
"epoch": 4.27972760607648,
|
|
"grad_norm": 0.3846431823338006,
|
|
"learning_rate": 1.5753567551057112e-05,
|
|
"loss": 0.1432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07181891798973083,
|
|
"step": 8170,
|
|
"valid_targets_mean": 6744.6,
|
|
"valid_targets_min": 5386
|
|
},
|
|
{
|
|
"epoch": 4.28234677841802,
|
|
"grad_norm": 0.3922439526902668,
|
|
"learning_rate": 1.5728043458734278e-05,
|
|
"loss": 0.1533,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08280803263187408,
|
|
"step": 8175,
|
|
"valid_targets_mean": 6975.9,
|
|
"valid_targets_min": 4868
|
|
},
|
|
{
|
|
"epoch": 4.28496595075956,
|
|
"grad_norm": 0.3705253592513052,
|
|
"learning_rate": 1.570252665467583e-05,
|
|
"loss": 0.1728,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09138161689043045,
|
|
"step": 8180,
|
|
"valid_targets_mean": 6363.0,
|
|
"valid_targets_min": 4898
|
|
},
|
|
{
|
|
"epoch": 4.2875851231011,
|
|
"grad_norm": 0.39750400903323196,
|
|
"learning_rate": 1.567701718241529e-05,
|
|
"loss": 0.1531,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08012625575065613,
|
|
"step": 8185,
|
|
"valid_targets_mean": 5625.8,
|
|
"valid_targets_min": 3834
|
|
},
|
|
{
|
|
"epoch": 4.29020429544264,
|
|
"grad_norm": 0.43953803462340063,
|
|
"learning_rate": 1.565151508547363e-05,
|
|
"loss": 0.157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10322760045528412,
|
|
"step": 8190,
|
|
"valid_targets_mean": 6519.0,
|
|
"valid_targets_min": 4551
|
|
},
|
|
{
|
|
"epoch": 4.292823467784181,
|
|
"grad_norm": 0.4045476476035004,
|
|
"learning_rate": 1.5626020407359266e-05,
|
|
"loss": 0.1648,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08422790467739105,
|
|
"step": 8195,
|
|
"valid_targets_mean": 5930.0,
|
|
"valid_targets_min": 4462
|
|
},
|
|
{
|
|
"epoch": 4.29544264012572,
|
|
"grad_norm": 0.37836438305764863,
|
|
"learning_rate": 1.560053319156795e-05,
|
|
"loss": 0.1835,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07530433684587479,
|
|
"step": 8200,
|
|
"valid_targets_mean": 5775.8,
|
|
"valid_targets_min": 4973
|
|
},
|
|
{
|
|
"epoch": 4.29806181246726,
|
|
"grad_norm": 0.37808126695418026,
|
|
"learning_rate": 1.5575053481582694e-05,
|
|
"loss": 0.1863,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08002137392759323,
|
|
"step": 8205,
|
|
"valid_targets_mean": 6459.4,
|
|
"valid_targets_min": 4619
|
|
},
|
|
{
|
|
"epoch": 4.300680984808801,
|
|
"grad_norm": 0.3833585919515166,
|
|
"learning_rate": 1.5549581320873715e-05,
|
|
"loss": 0.1578,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07882620394229889,
|
|
"step": 8210,
|
|
"valid_targets_mean": 6217.4,
|
|
"valid_targets_min": 5647
|
|
},
|
|
{
|
|
"epoch": 4.303300157150341,
|
|
"grad_norm": 0.36986935643126034,
|
|
"learning_rate": 1.5524116752898347e-05,
|
|
"loss": 0.1609,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07829003036022186,
|
|
"step": 8215,
|
|
"valid_targets_mean": 7183.4,
|
|
"valid_targets_min": 4398
|
|
},
|
|
{
|
|
"epoch": 4.30591932949188,
|
|
"grad_norm": 0.4002580775962094,
|
|
"learning_rate": 1.5498659821100975e-05,
|
|
"loss": 0.1619,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09484715759754181,
|
|
"step": 8220,
|
|
"valid_targets_mean": 6605.8,
|
|
"valid_targets_min": 4418
|
|
},
|
|
{
|
|
"epoch": 4.308538501833421,
|
|
"grad_norm": 0.3925185745729677,
|
|
"learning_rate": 1.5473210568912935e-05,
|
|
"loss": 0.1672,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07751842588186264,
|
|
"step": 8225,
|
|
"valid_targets_mean": 5922.4,
|
|
"valid_targets_min": 4969
|
|
},
|
|
{
|
|
"epoch": 4.311157674174961,
|
|
"grad_norm": 0.37590531017000645,
|
|
"learning_rate": 1.54477690397525e-05,
|
|
"loss": 0.1545,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08296653628349304,
|
|
"step": 8230,
|
|
"valid_targets_mean": 7903.5,
|
|
"valid_targets_min": 5042
|
|
},
|
|
{
|
|
"epoch": 4.313776846516501,
|
|
"grad_norm": 0.44387549334873155,
|
|
"learning_rate": 1.5422335277024733e-05,
|
|
"loss": 0.1693,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08491471409797668,
|
|
"step": 8235,
|
|
"valid_targets_mean": 7043.1,
|
|
"valid_targets_min": 4625
|
|
},
|
|
{
|
|
"epoch": 4.316396018858041,
|
|
"grad_norm": 0.4238226569876959,
|
|
"learning_rate": 1.5396909324121446e-05,
|
|
"loss": 0.1657,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0797121524810791,
|
|
"step": 8240,
|
|
"valid_targets_mean": 6242.0,
|
|
"valid_targets_min": 4820
|
|
},
|
|
{
|
|
"epoch": 4.319015191199581,
|
|
"grad_norm": 0.3864608870176148,
|
|
"learning_rate": 1.5371491224421165e-05,
|
|
"loss": 0.1532,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07022988796234131,
|
|
"step": 8245,
|
|
"valid_targets_mean": 5716.8,
|
|
"valid_targets_min": 4337
|
|
},
|
|
{
|
|
"epoch": 4.321634363541121,
|
|
"grad_norm": 0.3244819918244311,
|
|
"learning_rate": 1.5346081021288967e-05,
|
|
"loss": 0.1416,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07506868243217468,
|
|
"step": 8250,
|
|
"valid_targets_mean": 7500.8,
|
|
"valid_targets_min": 3794
|
|
},
|
|
{
|
|
"epoch": 4.3242535358826615,
|
|
"grad_norm": 0.4429870159176524,
|
|
"learning_rate": 1.5320678758076504e-05,
|
|
"loss": 0.1575,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07346892356872559,
|
|
"step": 8255,
|
|
"valid_targets_mean": 6387.5,
|
|
"valid_targets_min": 5298
|
|
},
|
|
{
|
|
"epoch": 4.326872708224201,
|
|
"grad_norm": 0.40158137143360445,
|
|
"learning_rate": 1.5295284478121847e-05,
|
|
"loss": 0.169,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08658412098884583,
|
|
"step": 8260,
|
|
"valid_targets_mean": 6235.0,
|
|
"valid_targets_min": 4543
|
|
},
|
|
{
|
|
"epoch": 4.329491880565741,
|
|
"grad_norm": 0.4174231059537175,
|
|
"learning_rate": 1.5269898224749454e-05,
|
|
"loss": 0.1645,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08053864538669586,
|
|
"step": 8265,
|
|
"valid_targets_mean": 6094.1,
|
|
"valid_targets_min": 4713
|
|
},
|
|
{
|
|
"epoch": 4.3321110529072815,
|
|
"grad_norm": 0.3911365100150199,
|
|
"learning_rate": 1.5244520041270114e-05,
|
|
"loss": 0.1503,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07582774758338928,
|
|
"step": 8270,
|
|
"valid_targets_mean": 6039.5,
|
|
"valid_targets_min": 5269
|
|
},
|
|
{
|
|
"epoch": 4.334730225248821,
|
|
"grad_norm": 0.38473495145801856,
|
|
"learning_rate": 1.5219149970980809e-05,
|
|
"loss": 0.1504,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08208172768354416,
|
|
"step": 8275,
|
|
"valid_targets_mean": 6892.6,
|
|
"valid_targets_min": 3450
|
|
},
|
|
{
|
|
"epoch": 4.337349397590361,
|
|
"grad_norm": 0.37531303252658804,
|
|
"learning_rate": 1.5193788057164716e-05,
|
|
"loss": 0.1592,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07927361130714417,
|
|
"step": 8280,
|
|
"valid_targets_mean": 6732.4,
|
|
"valid_targets_min": 4549
|
|
},
|
|
{
|
|
"epoch": 4.3399685699319015,
|
|
"grad_norm": 0.3707714784701887,
|
|
"learning_rate": 1.5168434343091061e-05,
|
|
"loss": 0.1629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0851498544216156,
|
|
"step": 8285,
|
|
"valid_targets_mean": 6249.5,
|
|
"valid_targets_min": 4838
|
|
},
|
|
{
|
|
"epoch": 4.342587742273442,
|
|
"grad_norm": 0.38722948193703377,
|
|
"learning_rate": 1.5143088872015121e-05,
|
|
"loss": 0.1525,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08623776584863663,
|
|
"step": 8290,
|
|
"valid_targets_mean": 6519.8,
|
|
"valid_targets_min": 4716
|
|
},
|
|
{
|
|
"epoch": 4.345206914614982,
|
|
"grad_norm": 0.44394643149366275,
|
|
"learning_rate": 1.511775168717808e-05,
|
|
"loss": 0.1567,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08702689409255981,
|
|
"step": 8295,
|
|
"valid_targets_mean": 5549.1,
|
|
"valid_targets_min": 4309
|
|
},
|
|
{
|
|
"epoch": 4.3478260869565215,
|
|
"grad_norm": 0.38937467723302926,
|
|
"learning_rate": 1.5092422831806981e-05,
|
|
"loss": 0.1636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0843573585152626,
|
|
"step": 8300,
|
|
"valid_targets_mean": 6353.5,
|
|
"valid_targets_min": 4923
|
|
},
|
|
{
|
|
"epoch": 4.350445259298062,
|
|
"grad_norm": 0.35874973212622546,
|
|
"learning_rate": 1.5067102349114686e-05,
|
|
"loss": 0.1355,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06433399021625519,
|
|
"step": 8305,
|
|
"valid_targets_mean": 6128.0,
|
|
"valid_targets_min": 4112
|
|
},
|
|
{
|
|
"epoch": 4.353064431639602,
|
|
"grad_norm": 0.37073705519190303,
|
|
"learning_rate": 1.5041790282299742e-05,
|
|
"loss": 0.1363,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0748823881149292,
|
|
"step": 8310,
|
|
"valid_targets_mean": 5755.4,
|
|
"valid_targets_min": 3761
|
|
},
|
|
{
|
|
"epoch": 4.3556836039811415,
|
|
"grad_norm": 0.3665699262275065,
|
|
"learning_rate": 1.5016486674546367e-05,
|
|
"loss": 0.1445,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06569778919219971,
|
|
"step": 8315,
|
|
"valid_targets_mean": 6331.2,
|
|
"valid_targets_min": 4928
|
|
},
|
|
{
|
|
"epoch": 4.358302776322682,
|
|
"grad_norm": 0.4175247424390106,
|
|
"learning_rate": 1.499119156902432e-05,
|
|
"loss": 0.151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07225983589887619,
|
|
"step": 8320,
|
|
"valid_targets_mean": 5321.1,
|
|
"valid_targets_min": 3853
|
|
},
|
|
{
|
|
"epoch": 4.360921948664222,
|
|
"grad_norm": 0.5019380728771753,
|
|
"learning_rate": 1.4965905008888864e-05,
|
|
"loss": 0.1717,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09124556183815002,
|
|
"step": 8325,
|
|
"valid_targets_mean": 6596.5,
|
|
"valid_targets_min": 4989
|
|
},
|
|
{
|
|
"epoch": 4.363541121005762,
|
|
"grad_norm": 0.4131044204920539,
|
|
"learning_rate": 1.4940627037280693e-05,
|
|
"loss": 0.1906,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08972490578889847,
|
|
"step": 8330,
|
|
"valid_targets_mean": 6367.1,
|
|
"valid_targets_min": 4487
|
|
},
|
|
{
|
|
"epoch": 4.366160293347303,
|
|
"grad_norm": 0.3813952818346332,
|
|
"learning_rate": 1.4915357697325834e-05,
|
|
"loss": 0.1512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06765107810497284,
|
|
"step": 8335,
|
|
"valid_targets_mean": 5445.4,
|
|
"valid_targets_min": 4773
|
|
},
|
|
{
|
|
"epoch": 4.368779465688842,
|
|
"grad_norm": 0.3979086901828073,
|
|
"learning_rate": 1.48900970321356e-05,
|
|
"loss": 0.1583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08942441642284393,
|
|
"step": 8340,
|
|
"valid_targets_mean": 6128.0,
|
|
"valid_targets_min": 4608
|
|
},
|
|
{
|
|
"epoch": 4.371398638030382,
|
|
"grad_norm": 0.3792761939688784,
|
|
"learning_rate": 1.4864845084806492e-05,
|
|
"loss": 0.1548,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07978663593530655,
|
|
"step": 8345,
|
|
"valid_targets_mean": 6867.9,
|
|
"valid_targets_min": 4896
|
|
},
|
|
{
|
|
"epoch": 4.374017810371923,
|
|
"grad_norm": 0.3965678390767401,
|
|
"learning_rate": 1.4839601898420153e-05,
|
|
"loss": 0.1772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08483615517616272,
|
|
"step": 8350,
|
|
"valid_targets_mean": 6932.2,
|
|
"valid_targets_min": 4949
|
|
},
|
|
{
|
|
"epoch": 4.376636982713462,
|
|
"grad_norm": 0.44849932550494337,
|
|
"learning_rate": 1.4814367516043263e-05,
|
|
"loss": 0.1638,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08830994367599487,
|
|
"step": 8355,
|
|
"valid_targets_mean": 6357.8,
|
|
"valid_targets_min": 4489
|
|
},
|
|
{
|
|
"epoch": 4.379256155055002,
|
|
"grad_norm": 0.4033732904904218,
|
|
"learning_rate": 1.4789141980727488e-05,
|
|
"loss": 0.1576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07515868544578552,
|
|
"step": 8360,
|
|
"valid_targets_mean": 5858.9,
|
|
"valid_targets_min": 4865
|
|
},
|
|
{
|
|
"epoch": 4.381875327396543,
|
|
"grad_norm": 0.3966493629251246,
|
|
"learning_rate": 1.4763925335509411e-05,
|
|
"loss": 0.1625,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07647809386253357,
|
|
"step": 8365,
|
|
"valid_targets_mean": 6141.9,
|
|
"valid_targets_min": 4452
|
|
},
|
|
{
|
|
"epoch": 4.384494499738083,
|
|
"grad_norm": 0.3614593249290682,
|
|
"learning_rate": 1.473871762341042e-05,
|
|
"loss": 0.1414,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07073986530303955,
|
|
"step": 8370,
|
|
"valid_targets_mean": 6311.6,
|
|
"valid_targets_min": 4915
|
|
},
|
|
{
|
|
"epoch": 4.387113672079622,
|
|
"grad_norm": 0.35566488596181045,
|
|
"learning_rate": 1.4713518887436702e-05,
|
|
"loss": 0.1564,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06397022306919098,
|
|
"step": 8375,
|
|
"valid_targets_mean": 5685.0,
|
|
"valid_targets_min": 3727
|
|
},
|
|
{
|
|
"epoch": 4.389732844421163,
|
|
"grad_norm": 0.38131493278440376,
|
|
"learning_rate": 1.46883291705791e-05,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07835820317268372,
|
|
"step": 8380,
|
|
"valid_targets_mean": 6329.0,
|
|
"valid_targets_min": 4482
|
|
},
|
|
{
|
|
"epoch": 4.392352016762703,
|
|
"grad_norm": 0.38581415170959144,
|
|
"learning_rate": 1.4663148515813066e-05,
|
|
"loss": 0.1488,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07043133676052094,
|
|
"step": 8385,
|
|
"valid_targets_mean": 6156.1,
|
|
"valid_targets_min": 3362
|
|
},
|
|
{
|
|
"epoch": 4.394971189104243,
|
|
"grad_norm": 0.4008182160511644,
|
|
"learning_rate": 1.4637976966098628e-05,
|
|
"loss": 0.1612,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07270411401987076,
|
|
"step": 8390,
|
|
"valid_targets_mean": 5596.8,
|
|
"valid_targets_min": 4395
|
|
},
|
|
{
|
|
"epoch": 4.397590361445783,
|
|
"grad_norm": 0.40338277115891336,
|
|
"learning_rate": 1.4612814564380231e-05,
|
|
"loss": 0.1608,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07577487826347351,
|
|
"step": 8395,
|
|
"valid_targets_mean": 5501.1,
|
|
"valid_targets_min": 4135
|
|
},
|
|
{
|
|
"epoch": 4.400209533787323,
|
|
"grad_norm": 0.37926452431990737,
|
|
"learning_rate": 1.4587661353586761e-05,
|
|
"loss": 0.1508,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06744927912950516,
|
|
"step": 8400,
|
|
"valid_targets_mean": 5784.1,
|
|
"valid_targets_min": 4262
|
|
},
|
|
{
|
|
"epoch": 4.402828706128863,
|
|
"grad_norm": 0.3872729945282216,
|
|
"learning_rate": 1.456251737663138e-05,
|
|
"loss": 0.1416,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08369916677474976,
|
|
"step": 8405,
|
|
"valid_targets_mean": 7510.2,
|
|
"valid_targets_min": 5149
|
|
},
|
|
{
|
|
"epoch": 4.405447878470404,
|
|
"grad_norm": 0.38975460785350247,
|
|
"learning_rate": 1.4537382676411536e-05,
|
|
"loss": 0.1476,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06407198309898376,
|
|
"step": 8410,
|
|
"valid_targets_mean": 5908.6,
|
|
"valid_targets_min": 3437
|
|
},
|
|
{
|
|
"epoch": 4.408067050811943,
|
|
"grad_norm": 0.394670282237162,
|
|
"learning_rate": 1.451225729580882e-05,
|
|
"loss": 0.1535,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07432819157838821,
|
|
"step": 8415,
|
|
"valid_targets_mean": 6852.4,
|
|
"valid_targets_min": 3535
|
|
},
|
|
{
|
|
"epoch": 4.410686223153483,
|
|
"grad_norm": 0.3901780229884973,
|
|
"learning_rate": 1.448714127768893e-05,
|
|
"loss": 0.1484,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08478435128927231,
|
|
"step": 8420,
|
|
"valid_targets_mean": 6686.4,
|
|
"valid_targets_min": 4712
|
|
},
|
|
{
|
|
"epoch": 4.413305395495024,
|
|
"grad_norm": 0.3489415062198455,
|
|
"learning_rate": 1.4462034664901608e-05,
|
|
"loss": 0.1505,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07812289893627167,
|
|
"step": 8425,
|
|
"valid_targets_mean": 7596.2,
|
|
"valid_targets_min": 4987
|
|
},
|
|
{
|
|
"epoch": 4.415924567836564,
|
|
"grad_norm": 0.376974197859487,
|
|
"learning_rate": 1.4436937500280522e-05,
|
|
"loss": 0.1583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07755880057811737,
|
|
"step": 8430,
|
|
"valid_targets_mean": 6937.0,
|
|
"valid_targets_min": 3577
|
|
},
|
|
{
|
|
"epoch": 4.418543740178103,
|
|
"grad_norm": 0.40396905556405893,
|
|
"learning_rate": 1.4411849826643252e-05,
|
|
"loss": 0.1383,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07415942847728729,
|
|
"step": 8435,
|
|
"valid_targets_mean": 5414.4,
|
|
"valid_targets_min": 3665
|
|
},
|
|
{
|
|
"epoch": 4.421162912519644,
|
|
"grad_norm": 0.4660682122084262,
|
|
"learning_rate": 1.4386771686791161e-05,
|
|
"loss": 0.1597,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07723142951726913,
|
|
"step": 8440,
|
|
"valid_targets_mean": 6955.4,
|
|
"valid_targets_min": 5409
|
|
},
|
|
{
|
|
"epoch": 4.423782084861184,
|
|
"grad_norm": 0.40122313872101795,
|
|
"learning_rate": 1.4361703123509347e-05,
|
|
"loss": 0.1606,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08311334252357483,
|
|
"step": 8445,
|
|
"valid_targets_mean": 6058.8,
|
|
"valid_targets_min": 3545
|
|
},
|
|
{
|
|
"epoch": 4.426401257202724,
|
|
"grad_norm": 0.6657607597510616,
|
|
"learning_rate": 1.4336644179566595e-05,
|
|
"loss": 0.186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1686384677886963,
|
|
"step": 8450,
|
|
"valid_targets_mean": 5807.8,
|
|
"valid_targets_min": 2317
|
|
},
|
|
{
|
|
"epoch": 4.429020429544264,
|
|
"grad_norm": 0.531777753535127,
|
|
"learning_rate": 1.4311594897715249e-05,
|
|
"loss": 0.2296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1320067048072815,
|
|
"step": 8455,
|
|
"valid_targets_mean": 5801.6,
|
|
"valid_targets_min": 2453
|
|
},
|
|
{
|
|
"epoch": 4.431639601885804,
|
|
"grad_norm": 0.5398361384530933,
|
|
"learning_rate": 1.4286555320691182e-05,
|
|
"loss": 0.2268,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12875959277153015,
|
|
"step": 8460,
|
|
"valid_targets_mean": 5075.4,
|
|
"valid_targets_min": 835
|
|
},
|
|
{
|
|
"epoch": 4.434258774227344,
|
|
"grad_norm": 0.5868007605547965,
|
|
"learning_rate": 1.4261525491213712e-05,
|
|
"loss": 0.2241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13063159584999084,
|
|
"step": 8465,
|
|
"valid_targets_mean": 4082.1,
|
|
"valid_targets_min": 1201
|
|
},
|
|
{
|
|
"epoch": 4.436877946568885,
|
|
"grad_norm": 0.5381634816077148,
|
|
"learning_rate": 1.4236505451985533e-05,
|
|
"loss": 0.2138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1091819629073143,
|
|
"step": 8470,
|
|
"valid_targets_mean": 4560.0,
|
|
"valid_targets_min": 1880
|
|
},
|
|
{
|
|
"epoch": 4.439497118910424,
|
|
"grad_norm": 0.5310120711737825,
|
|
"learning_rate": 1.4211495245692608e-05,
|
|
"loss": 0.2381,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16839918494224548,
|
|
"step": 8475,
|
|
"valid_targets_mean": 6637.9,
|
|
"valid_targets_min": 2171
|
|
},
|
|
{
|
|
"epoch": 4.442116291251964,
|
|
"grad_norm": 0.5190278146440941,
|
|
"learning_rate": 1.4186494915004159e-05,
|
|
"loss": 0.216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1253286898136139,
|
|
"step": 8480,
|
|
"valid_targets_mean": 6562.5,
|
|
"valid_targets_min": 1789
|
|
},
|
|
{
|
|
"epoch": 4.444735463593505,
|
|
"grad_norm": 0.4867697495156829,
|
|
"learning_rate": 1.4161504502572541e-05,
|
|
"loss": 0.2226,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09787391126155853,
|
|
"step": 8485,
|
|
"valid_targets_mean": 5572.5,
|
|
"valid_targets_min": 1855
|
|
},
|
|
{
|
|
"epoch": 4.447354635935045,
|
|
"grad_norm": 0.5822314957768182,
|
|
"learning_rate": 1.4136524051033181e-05,
|
|
"loss": 0.2232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11792335659265518,
|
|
"step": 8490,
|
|
"valid_targets_mean": 3913.8,
|
|
"valid_targets_min": 1024
|
|
},
|
|
{
|
|
"epoch": 4.449973808276584,
|
|
"grad_norm": 0.5158227550646646,
|
|
"learning_rate": 1.4111553603004538e-05,
|
|
"loss": 0.2314,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11358408629894257,
|
|
"step": 8495,
|
|
"valid_targets_mean": 4908.0,
|
|
"valid_targets_min": 2107
|
|
},
|
|
{
|
|
"epoch": 4.452592980618125,
|
|
"grad_norm": 0.5869961539965604,
|
|
"learning_rate": 1.4086593201087979e-05,
|
|
"loss": 0.2274,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12451714277267456,
|
|
"step": 8500,
|
|
"valid_targets_mean": 4423.2,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 4.455212152959665,
|
|
"grad_norm": 0.4611985891917862,
|
|
"learning_rate": 1.4061642887867732e-05,
|
|
"loss": 0.2133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08776941895484924,
|
|
"step": 8505,
|
|
"valid_targets_mean": 4354.5,
|
|
"valid_targets_min": 1597
|
|
},
|
|
{
|
|
"epoch": 4.457831325301205,
|
|
"grad_norm": 0.5688775918226212,
|
|
"learning_rate": 1.4036702705910842e-05,
|
|
"loss": 0.2269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11041654646396637,
|
|
"step": 8510,
|
|
"valid_targets_mean": 3518.2,
|
|
"valid_targets_min": 1650
|
|
},
|
|
{
|
|
"epoch": 4.460450497642745,
|
|
"grad_norm": 0.5518642875567444,
|
|
"learning_rate": 1.401177269776703e-05,
|
|
"loss": 0.2216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1132279559969902,
|
|
"step": 8515,
|
|
"valid_targets_mean": 4340.0,
|
|
"valid_targets_min": 1385
|
|
},
|
|
{
|
|
"epoch": 4.463069669984285,
|
|
"grad_norm": 0.6094475008163417,
|
|
"learning_rate": 1.3986852905968694e-05,
|
|
"loss": 0.2154,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1261036992073059,
|
|
"step": 8520,
|
|
"valid_targets_mean": 4061.2,
|
|
"valid_targets_min": 3087
|
|
},
|
|
{
|
|
"epoch": 4.465688842325825,
|
|
"grad_norm": 0.5162864921806446,
|
|
"learning_rate": 1.396194337303078e-05,
|
|
"loss": 0.2208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10235683619976044,
|
|
"step": 8525,
|
|
"valid_targets_mean": 5073.4,
|
|
"valid_targets_min": 2785
|
|
},
|
|
{
|
|
"epoch": 4.4683080146673655,
|
|
"grad_norm": 0.6088077594799497,
|
|
"learning_rate": 1.393704414145073e-05,
|
|
"loss": 0.2158,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11759968847036362,
|
|
"step": 8530,
|
|
"valid_targets_mean": 4354.0,
|
|
"valid_targets_min": 1210
|
|
},
|
|
{
|
|
"epoch": 4.470927187008905,
|
|
"grad_norm": 0.6151449199270609,
|
|
"learning_rate": 1.3912155253708437e-05,
|
|
"loss": 0.2196,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11967704445123672,
|
|
"step": 8535,
|
|
"valid_targets_mean": 3266.6,
|
|
"valid_targets_min": 1064
|
|
},
|
|
{
|
|
"epoch": 4.473546359350445,
|
|
"grad_norm": 0.6436183746478787,
|
|
"learning_rate": 1.3887276752266113e-05,
|
|
"loss": 0.2281,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11920354515314102,
|
|
"step": 8540,
|
|
"valid_targets_mean": 3360.6,
|
|
"valid_targets_min": 1678
|
|
},
|
|
{
|
|
"epoch": 4.4761655316919855,
|
|
"grad_norm": 0.589124862126253,
|
|
"learning_rate": 1.3862408679568282e-05,
|
|
"loss": 0.2215,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10800355672836304,
|
|
"step": 8545,
|
|
"valid_targets_mean": 3776.0,
|
|
"valid_targets_min": 671
|
|
},
|
|
{
|
|
"epoch": 4.478784704033526,
|
|
"grad_norm": 0.6317737872398538,
|
|
"learning_rate": 1.3837551078041644e-05,
|
|
"loss": 0.2233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11351955682039261,
|
|
"step": 8550,
|
|
"valid_targets_mean": 3642.9,
|
|
"valid_targets_min": 2000
|
|
},
|
|
{
|
|
"epoch": 4.481403876375065,
|
|
"grad_norm": 0.6298128571652256,
|
|
"learning_rate": 1.381270399009507e-05,
|
|
"loss": 0.2096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13147234916687012,
|
|
"step": 8555,
|
|
"valid_targets_mean": 4261.9,
|
|
"valid_targets_min": 1532
|
|
},
|
|
{
|
|
"epoch": 4.4840230487166055,
|
|
"grad_norm": 0.5943111027501365,
|
|
"learning_rate": 1.3787867458119464e-05,
|
|
"loss": 0.2116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07439844310283661,
|
|
"step": 8560,
|
|
"valid_targets_mean": 3298.9,
|
|
"valid_targets_min": 1051
|
|
},
|
|
{
|
|
"epoch": 4.486642221058146,
|
|
"grad_norm": 0.5826298433530936,
|
|
"learning_rate": 1.3763041524487721e-05,
|
|
"loss": 0.2047,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09946641325950623,
|
|
"step": 8565,
|
|
"valid_targets_mean": 3058.1,
|
|
"valid_targets_min": 1345
|
|
},
|
|
{
|
|
"epoch": 4.489261393399686,
|
|
"grad_norm": 0.5620343902348789,
|
|
"learning_rate": 1.3738226231554687e-05,
|
|
"loss": 0.2191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09985310584306717,
|
|
"step": 8570,
|
|
"valid_targets_mean": 4157.6,
|
|
"valid_targets_min": 2231
|
|
},
|
|
{
|
|
"epoch": 4.4918805657412255,
|
|
"grad_norm": 0.6195213507564987,
|
|
"learning_rate": 1.3713421621657015e-05,
|
|
"loss": 0.2231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08839579671621323,
|
|
"step": 8575,
|
|
"valid_targets_mean": 3561.4,
|
|
"valid_targets_min": 2127
|
|
},
|
|
{
|
|
"epoch": 4.494499738082766,
|
|
"grad_norm": 0.6569017535266028,
|
|
"learning_rate": 1.3688627737113163e-05,
|
|
"loss": 0.2134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11849110573530197,
|
|
"step": 8580,
|
|
"valid_targets_mean": 3363.4,
|
|
"valid_targets_min": 1265
|
|
},
|
|
{
|
|
"epoch": 4.497118910424306,
|
|
"grad_norm": 0.564846860750686,
|
|
"learning_rate": 1.3663844620223272e-05,
|
|
"loss": 0.2114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08929997682571411,
|
|
"step": 8585,
|
|
"valid_targets_mean": 3700.0,
|
|
"valid_targets_min": 908
|
|
},
|
|
{
|
|
"epoch": 4.499738082765846,
|
|
"grad_norm": 0.6512149054655086,
|
|
"learning_rate": 1.3639072313269106e-05,
|
|
"loss": 0.2141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11318524181842804,
|
|
"step": 8590,
|
|
"valid_targets_mean": 2910.9,
|
|
"valid_targets_min": 1264
|
|
},
|
|
{
|
|
"epoch": 4.502357255107386,
|
|
"grad_norm": 0.6465709019919834,
|
|
"learning_rate": 1.3614310858514016e-05,
|
|
"loss": 0.2102,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1102558895945549,
|
|
"step": 8595,
|
|
"valid_targets_mean": 2855.2,
|
|
"valid_targets_min": 926
|
|
},
|
|
{
|
|
"epoch": 4.504976427448926,
|
|
"grad_norm": 0.6563597261559485,
|
|
"learning_rate": 1.3589560298202811e-05,
|
|
"loss": 0.2115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08837442845106125,
|
|
"step": 8600,
|
|
"valid_targets_mean": 2905.6,
|
|
"valid_targets_min": 1016
|
|
},
|
|
{
|
|
"epoch": 4.507595599790466,
|
|
"grad_norm": 0.6180672798730263,
|
|
"learning_rate": 1.3564820674561726e-05,
|
|
"loss": 0.2081,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09526603668928146,
|
|
"step": 8605,
|
|
"valid_targets_mean": 3499.9,
|
|
"valid_targets_min": 2019
|
|
},
|
|
{
|
|
"epoch": 4.510214772132006,
|
|
"grad_norm": 0.602696371537178,
|
|
"learning_rate": 1.354009202979833e-05,
|
|
"loss": 0.2165,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13709108531475067,
|
|
"step": 8610,
|
|
"valid_targets_mean": 4911.0,
|
|
"valid_targets_min": 3583
|
|
},
|
|
{
|
|
"epoch": 4.512833944473546,
|
|
"grad_norm": 0.5830142516755014,
|
|
"learning_rate": 1.3515374406101476e-05,
|
|
"loss": 0.2088,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09028695523738861,
|
|
"step": 8615,
|
|
"valid_targets_mean": 3381.5,
|
|
"valid_targets_min": 1110
|
|
},
|
|
{
|
|
"epoch": 4.515453116815086,
|
|
"grad_norm": 0.6140449216611795,
|
|
"learning_rate": 1.3490667845641188e-05,
|
|
"loss": 0.2076,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12959925830364227,
|
|
"step": 8620,
|
|
"valid_targets_mean": 4497.9,
|
|
"valid_targets_min": 1460
|
|
},
|
|
{
|
|
"epoch": 4.518072289156627,
|
|
"grad_norm": 0.6397151511956849,
|
|
"learning_rate": 1.3465972390568643e-05,
|
|
"loss": 0.2057,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12052600830793381,
|
|
"step": 8625,
|
|
"valid_targets_mean": 3962.4,
|
|
"valid_targets_min": 2719
|
|
},
|
|
{
|
|
"epoch": 4.520691461498167,
|
|
"grad_norm": 0.6190465441758102,
|
|
"learning_rate": 1.3441288083016058e-05,
|
|
"loss": 0.2182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09589071571826935,
|
|
"step": 8630,
|
|
"valid_targets_mean": 3175.9,
|
|
"valid_targets_min": 1673
|
|
},
|
|
{
|
|
"epoch": 4.523310633839706,
|
|
"grad_norm": 0.5992094789548759,
|
|
"learning_rate": 1.3416614965096622e-05,
|
|
"loss": 0.212,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09990492463111877,
|
|
"step": 8635,
|
|
"valid_targets_mean": 3482.6,
|
|
"valid_targets_min": 1860
|
|
},
|
|
{
|
|
"epoch": 4.525929806181247,
|
|
"grad_norm": 0.5499261289440762,
|
|
"learning_rate": 1.3391953078904458e-05,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1030711829662323,
|
|
"step": 8640,
|
|
"valid_targets_mean": 4597.0,
|
|
"valid_targets_min": 2934
|
|
},
|
|
{
|
|
"epoch": 4.528548978522787,
|
|
"grad_norm": 0.6006706484609484,
|
|
"learning_rate": 1.3367302466514514e-05,
|
|
"loss": 0.2055,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10380536317825317,
|
|
"step": 8645,
|
|
"valid_targets_mean": 3667.6,
|
|
"valid_targets_min": 1945
|
|
},
|
|
{
|
|
"epoch": 4.5311681508643264,
|
|
"grad_norm": 0.5931622596203528,
|
|
"learning_rate": 1.3342663169982486e-05,
|
|
"loss": 0.2162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10778256505727768,
|
|
"step": 8650,
|
|
"valid_targets_mean": 4618.9,
|
|
"valid_targets_min": 1908
|
|
},
|
|
{
|
|
"epoch": 4.533787323205867,
|
|
"grad_norm": 0.6151803569221982,
|
|
"learning_rate": 1.3318035231344806e-05,
|
|
"loss": 0.2108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1142893061041832,
|
|
"step": 8655,
|
|
"valid_targets_mean": 3561.0,
|
|
"valid_targets_min": 2250
|
|
},
|
|
{
|
|
"epoch": 4.536406495547407,
|
|
"grad_norm": 0.583305321099175,
|
|
"learning_rate": 1.3293418692618483e-05,
|
|
"loss": 0.2099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08924940228462219,
|
|
"step": 8660,
|
|
"valid_targets_mean": 4291.5,
|
|
"valid_targets_min": 2492
|
|
},
|
|
{
|
|
"epoch": 4.539025667888947,
|
|
"grad_norm": 0.7014814716734583,
|
|
"learning_rate": 1.3268813595801113e-05,
|
|
"loss": 0.2126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10996967554092407,
|
|
"step": 8665,
|
|
"valid_targets_mean": 4467.1,
|
|
"valid_targets_min": 3059
|
|
},
|
|
{
|
|
"epoch": 4.541644840230488,
|
|
"grad_norm": 0.6206863740286671,
|
|
"learning_rate": 1.3244219982870748e-05,
|
|
"loss": 0.2049,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1131991371512413,
|
|
"step": 8670,
|
|
"valid_targets_mean": 4255.2,
|
|
"valid_targets_min": 2283
|
|
},
|
|
{
|
|
"epoch": 4.544264012572027,
|
|
"grad_norm": 0.6771063630964453,
|
|
"learning_rate": 1.3219637895785864e-05,
|
|
"loss": 0.2082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11287044733762741,
|
|
"step": 8675,
|
|
"valid_targets_mean": 3564.6,
|
|
"valid_targets_min": 1182
|
|
},
|
|
{
|
|
"epoch": 4.546883184913567,
|
|
"grad_norm": 0.5413704303329966,
|
|
"learning_rate": 1.3195067376485264e-05,
|
|
"loss": 0.2078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09840285032987595,
|
|
"step": 8680,
|
|
"valid_targets_mean": 4895.6,
|
|
"valid_targets_min": 1869
|
|
},
|
|
{
|
|
"epoch": 4.549502357255108,
|
|
"grad_norm": 0.5874254960826278,
|
|
"learning_rate": 1.3170508466888009e-05,
|
|
"loss": 0.2142,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12842625379562378,
|
|
"step": 8685,
|
|
"valid_targets_mean": 4721.1,
|
|
"valid_targets_min": 2498
|
|
},
|
|
{
|
|
"epoch": 4.552121529596647,
|
|
"grad_norm": 0.6533078396292977,
|
|
"learning_rate": 1.3145961208893372e-05,
|
|
"loss": 0.2198,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1062917560338974,
|
|
"step": 8690,
|
|
"valid_targets_mean": 2935.6,
|
|
"valid_targets_min": 988
|
|
},
|
|
{
|
|
"epoch": 4.554740701938187,
|
|
"grad_norm": 0.6391898982211924,
|
|
"learning_rate": 1.312142564438072e-05,
|
|
"loss": 0.2062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11164413392543793,
|
|
"step": 8695,
|
|
"valid_targets_mean": 4191.4,
|
|
"valid_targets_min": 1534
|
|
},
|
|
{
|
|
"epoch": 4.557359874279728,
|
|
"grad_norm": 0.6046360407355982,
|
|
"learning_rate": 1.309690181520951e-05,
|
|
"loss": 0.2117,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10599304735660553,
|
|
"step": 8700,
|
|
"valid_targets_mean": 3546.4,
|
|
"valid_targets_min": 2252
|
|
},
|
|
{
|
|
"epoch": 4.559979046621268,
|
|
"grad_norm": 0.8632738414383352,
|
|
"learning_rate": 1.3072389763219144e-05,
|
|
"loss": 0.2167,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1250210702419281,
|
|
"step": 8705,
|
|
"valid_targets_mean": 3548.0,
|
|
"valid_targets_min": 1436
|
|
},
|
|
{
|
|
"epoch": 4.562598218962807,
|
|
"grad_norm": 0.6118169461610888,
|
|
"learning_rate": 1.3047889530228928e-05,
|
|
"loss": 0.2099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1000833511352539,
|
|
"step": 8710,
|
|
"valid_targets_mean": 4002.1,
|
|
"valid_targets_min": 2333
|
|
},
|
|
{
|
|
"epoch": 4.565217391304348,
|
|
"grad_norm": 0.6094791048918677,
|
|
"learning_rate": 1.3023401158038038e-05,
|
|
"loss": 0.204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0900658667087555,
|
|
"step": 8715,
|
|
"valid_targets_mean": 3507.9,
|
|
"valid_targets_min": 1259
|
|
},
|
|
{
|
|
"epoch": 4.567836563645888,
|
|
"grad_norm": 0.6202992915362916,
|
|
"learning_rate": 1.2998924688425377e-05,
|
|
"loss": 0.2032,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08818884193897247,
|
|
"step": 8720,
|
|
"valid_targets_mean": 2891.5,
|
|
"valid_targets_min": 1319
|
|
},
|
|
{
|
|
"epoch": 4.570455735987428,
|
|
"grad_norm": 0.6106657583472571,
|
|
"learning_rate": 1.2974460163149575e-05,
|
|
"loss": 0.1924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09374313056468964,
|
|
"step": 8725,
|
|
"valid_targets_mean": 3002.8,
|
|
"valid_targets_min": 1194
|
|
},
|
|
{
|
|
"epoch": 4.573074908328968,
|
|
"grad_norm": 0.6611859041510075,
|
|
"learning_rate": 1.2950007623948857e-05,
|
|
"loss": 0.2274,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0970490351319313,
|
|
"step": 8730,
|
|
"valid_targets_mean": 2704.0,
|
|
"valid_targets_min": 1453
|
|
},
|
|
{
|
|
"epoch": 4.575694080670508,
|
|
"grad_norm": 0.5176574483366256,
|
|
"learning_rate": 1.292556711254101e-05,
|
|
"loss": 0.2018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09145693480968475,
|
|
"step": 8735,
|
|
"valid_targets_mean": 5175.9,
|
|
"valid_targets_min": 3011
|
|
},
|
|
{
|
|
"epoch": 4.578313253012048,
|
|
"grad_norm": 0.5876009983134065,
|
|
"learning_rate": 1.2901138670623308e-05,
|
|
"loss": 0.2178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0815117210149765,
|
|
"step": 8740,
|
|
"valid_targets_mean": 2915.9,
|
|
"valid_targets_min": 1256
|
|
},
|
|
{
|
|
"epoch": 4.580932425353589,
|
|
"grad_norm": 0.6349592745954749,
|
|
"learning_rate": 1.2876722339872414e-05,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1345708966255188,
|
|
"step": 8745,
|
|
"valid_targets_mean": 4661.6,
|
|
"valid_targets_min": 1336
|
|
},
|
|
{
|
|
"epoch": 4.583551597695128,
|
|
"grad_norm": 0.6521800380858361,
|
|
"learning_rate": 1.2852318161944348e-05,
|
|
"loss": 0.2209,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10511691123247147,
|
|
"step": 8750,
|
|
"valid_targets_mean": 3390.4,
|
|
"valid_targets_min": 1275
|
|
},
|
|
{
|
|
"epoch": 4.586170770036668,
|
|
"grad_norm": 0.6147533075586407,
|
|
"learning_rate": 1.2827926178474388e-05,
|
|
"loss": 0.2079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11512017250061035,
|
|
"step": 8755,
|
|
"valid_targets_mean": 3660.0,
|
|
"valid_targets_min": 2202
|
|
},
|
|
{
|
|
"epoch": 4.588789942378209,
|
|
"grad_norm": 0.605091604756251,
|
|
"learning_rate": 1.2803546431077008e-05,
|
|
"loss": 0.2036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11255483329296112,
|
|
"step": 8760,
|
|
"valid_targets_mean": 4094.2,
|
|
"valid_targets_min": 3236
|
|
},
|
|
{
|
|
"epoch": 4.591409114719749,
|
|
"grad_norm": 0.6660110056330647,
|
|
"learning_rate": 1.2779178961345804e-05,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10259729623794556,
|
|
"step": 8765,
|
|
"valid_targets_mean": 3057.6,
|
|
"valid_targets_min": 1573
|
|
},
|
|
{
|
|
"epoch": 4.594028287061288,
|
|
"grad_norm": 0.6149358237464726,
|
|
"learning_rate": 1.2754823810853427e-05,
|
|
"loss": 0.1977,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0949232429265976,
|
|
"step": 8770,
|
|
"valid_targets_mean": 3780.9,
|
|
"valid_targets_min": 2343
|
|
},
|
|
{
|
|
"epoch": 4.596647459402829,
|
|
"grad_norm": 0.5747609813308386,
|
|
"learning_rate": 1.2730481021151518e-05,
|
|
"loss": 0.193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08529553562402725,
|
|
"step": 8775,
|
|
"valid_targets_mean": 2815.8,
|
|
"valid_targets_min": 1026
|
|
},
|
|
{
|
|
"epoch": 4.599266631744369,
|
|
"grad_norm": 0.6395929759195625,
|
|
"learning_rate": 1.2706150633770612e-05,
|
|
"loss": 0.1972,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09475855529308319,
|
|
"step": 8780,
|
|
"valid_targets_mean": 3460.9,
|
|
"valid_targets_min": 1042
|
|
},
|
|
{
|
|
"epoch": 4.601885804085909,
|
|
"grad_norm": 0.6147297344304274,
|
|
"learning_rate": 1.268183269022011e-05,
|
|
"loss": 0.1841,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08383920788764954,
|
|
"step": 8785,
|
|
"valid_targets_mean": 3172.8,
|
|
"valid_targets_min": 1274
|
|
},
|
|
{
|
|
"epoch": 4.604504976427449,
|
|
"grad_norm": 0.6828514781586573,
|
|
"learning_rate": 1.2657527231988151e-05,
|
|
"loss": 0.2002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10674040764570236,
|
|
"step": 8790,
|
|
"valid_targets_mean": 3939.5,
|
|
"valid_targets_min": 1575
|
|
},
|
|
{
|
|
"epoch": 4.607124148768989,
|
|
"grad_norm": 0.5924182910795877,
|
|
"learning_rate": 1.2633234300541612e-05,
|
|
"loss": 0.2237,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07903445512056351,
|
|
"step": 8795,
|
|
"valid_targets_mean": 3739.0,
|
|
"valid_targets_min": 1312
|
|
},
|
|
{
|
|
"epoch": 4.609743321110529,
|
|
"grad_norm": 0.614605999726541,
|
|
"learning_rate": 1.2608953937325965e-05,
|
|
"loss": 0.2019,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09165416657924652,
|
|
"step": 8800,
|
|
"valid_targets_mean": 3187.8,
|
|
"valid_targets_min": 1683
|
|
},
|
|
{
|
|
"epoch": 4.6123624934520695,
|
|
"grad_norm": 0.5661510792836556,
|
|
"learning_rate": 1.2584686183765245e-05,
|
|
"loss": 0.2065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09626249969005585,
|
|
"step": 8805,
|
|
"valid_targets_mean": 3945.8,
|
|
"valid_targets_min": 1743
|
|
},
|
|
{
|
|
"epoch": 4.614981665793609,
|
|
"grad_norm": 0.6096210195459415,
|
|
"learning_rate": 1.2560431081261997e-05,
|
|
"loss": 0.2033,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10719242691993713,
|
|
"step": 8810,
|
|
"valid_targets_mean": 3953.8,
|
|
"valid_targets_min": 2213
|
|
},
|
|
{
|
|
"epoch": 4.617600838135149,
|
|
"grad_norm": 0.6494948532806525,
|
|
"learning_rate": 1.253618867119715e-05,
|
|
"loss": 0.2054,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10217788815498352,
|
|
"step": 8815,
|
|
"valid_targets_mean": 3690.4,
|
|
"valid_targets_min": 1586
|
|
},
|
|
{
|
|
"epoch": 4.6202200104766895,
|
|
"grad_norm": 0.6640530052773205,
|
|
"learning_rate": 1.251195899493001e-05,
|
|
"loss": 0.2026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10520057380199432,
|
|
"step": 8820,
|
|
"valid_targets_mean": 3381.4,
|
|
"valid_targets_min": 2033
|
|
},
|
|
{
|
|
"epoch": 4.62283918281823,
|
|
"grad_norm": 0.5569491511306094,
|
|
"learning_rate": 1.248774209379814e-05,
|
|
"loss": 0.1914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08458670973777771,
|
|
"step": 8825,
|
|
"valid_targets_mean": 3651.2,
|
|
"valid_targets_min": 1616
|
|
},
|
|
{
|
|
"epoch": 4.625458355159769,
|
|
"grad_norm": 0.6330676503326945,
|
|
"learning_rate": 1.2463538009117296e-05,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09859472513198853,
|
|
"step": 8830,
|
|
"valid_targets_mean": 3462.6,
|
|
"valid_targets_min": 2000
|
|
},
|
|
{
|
|
"epoch": 4.6280775275013095,
|
|
"grad_norm": 0.5956085487061661,
|
|
"learning_rate": 1.2439346782181407e-05,
|
|
"loss": 0.1984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10616499185562134,
|
|
"step": 8835,
|
|
"valid_targets_mean": 3591.0,
|
|
"valid_targets_min": 1750
|
|
},
|
|
{
|
|
"epoch": 4.63069669984285,
|
|
"grad_norm": 0.6329298549757268,
|
|
"learning_rate": 1.2415168454262418e-05,
|
|
"loss": 0.2061,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12511585652828217,
|
|
"step": 8840,
|
|
"valid_targets_mean": 4291.8,
|
|
"valid_targets_min": 2429
|
|
},
|
|
{
|
|
"epoch": 4.63331587218439,
|
|
"grad_norm": 0.6358613876406229,
|
|
"learning_rate": 1.2391003066610314e-05,
|
|
"loss": 0.2131,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10768921673297882,
|
|
"step": 8845,
|
|
"valid_targets_mean": 3645.9,
|
|
"valid_targets_min": 2342
|
|
},
|
|
{
|
|
"epoch": 4.6359350445259295,
|
|
"grad_norm": 0.650281533288699,
|
|
"learning_rate": 1.2366850660452965e-05,
|
|
"loss": 0.2068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09575870633125305,
|
|
"step": 8850,
|
|
"valid_targets_mean": 3556.8,
|
|
"valid_targets_min": 1571
|
|
},
|
|
{
|
|
"epoch": 4.63855421686747,
|
|
"grad_norm": 0.6685805031293942,
|
|
"learning_rate": 1.2342711276996113e-05,
|
|
"loss": 0.2146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0916060209274292,
|
|
"step": 8855,
|
|
"valid_targets_mean": 2768.0,
|
|
"valid_targets_min": 1487
|
|
},
|
|
{
|
|
"epoch": 4.64117338920901,
|
|
"grad_norm": 0.5820326114744039,
|
|
"learning_rate": 1.2318584957423286e-05,
|
|
"loss": 0.2056,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08970531821250916,
|
|
"step": 8860,
|
|
"valid_targets_mean": 3589.1,
|
|
"valid_targets_min": 1970
|
|
},
|
|
{
|
|
"epoch": 4.6437925615505495,
|
|
"grad_norm": 0.5992598789489589,
|
|
"learning_rate": 1.22944717428957e-05,
|
|
"loss": 0.218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0977601408958435,
|
|
"step": 8865,
|
|
"valid_targets_mean": 4087.2,
|
|
"valid_targets_min": 2794
|
|
},
|
|
{
|
|
"epoch": 4.64641173389209,
|
|
"grad_norm": 0.6474027107259887,
|
|
"learning_rate": 1.2270371674552242e-05,
|
|
"loss": 0.2122,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0806116908788681,
|
|
"step": 8870,
|
|
"valid_targets_mean": 2785.5,
|
|
"valid_targets_min": 1066
|
|
},
|
|
{
|
|
"epoch": 4.64903090623363,
|
|
"grad_norm": 0.6027629773312335,
|
|
"learning_rate": 1.2246284793509355e-05,
|
|
"loss": 0.2126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09002608805894852,
|
|
"step": 8875,
|
|
"valid_targets_mean": 3409.9,
|
|
"valid_targets_min": 2281
|
|
},
|
|
{
|
|
"epoch": 4.65165007857517,
|
|
"grad_norm": 0.6146930851839235,
|
|
"learning_rate": 1.222221114086099e-05,
|
|
"loss": 0.2093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10688142478466034,
|
|
"step": 8880,
|
|
"valid_targets_mean": 4227.4,
|
|
"valid_targets_min": 1969
|
|
},
|
|
{
|
|
"epoch": 4.654269250916711,
|
|
"grad_norm": 0.637652856281438,
|
|
"learning_rate": 1.2198150757678518e-05,
|
|
"loss": 0.2052,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08704520016908646,
|
|
"step": 8885,
|
|
"valid_targets_mean": 2942.6,
|
|
"valid_targets_min": 1005
|
|
},
|
|
{
|
|
"epoch": 4.65688842325825,
|
|
"grad_norm": 0.6166255110417656,
|
|
"learning_rate": 1.2174103685010682e-05,
|
|
"loss": 0.2052,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09632515907287598,
|
|
"step": 8890,
|
|
"valid_targets_mean": 3895.9,
|
|
"valid_targets_min": 1980
|
|
},
|
|
{
|
|
"epoch": 4.65950759559979,
|
|
"grad_norm": 0.6358729645012968,
|
|
"learning_rate": 1.2150069963883517e-05,
|
|
"loss": 0.2113,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12068484723567963,
|
|
"step": 8895,
|
|
"valid_targets_mean": 4768.2,
|
|
"valid_targets_min": 1257
|
|
},
|
|
{
|
|
"epoch": 4.662126767941331,
|
|
"grad_norm": 0.6516633702770698,
|
|
"learning_rate": 1.2126049635300268e-05,
|
|
"loss": 0.1997,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1033945083618164,
|
|
"step": 8900,
|
|
"valid_targets_mean": 3240.4,
|
|
"valid_targets_min": 1357
|
|
},
|
|
{
|
|
"epoch": 4.66474594028287,
|
|
"grad_norm": 0.6050422744445452,
|
|
"learning_rate": 1.2102042740241351e-05,
|
|
"loss": 0.191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08464160561561584,
|
|
"step": 8905,
|
|
"valid_targets_mean": 4587.8,
|
|
"valid_targets_min": 1410
|
|
},
|
|
{
|
|
"epoch": 4.6673651126244105,
|
|
"grad_norm": 0.5991597385320386,
|
|
"learning_rate": 1.207804931966424e-05,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1021055206656456,
|
|
"step": 8910,
|
|
"valid_targets_mean": 3696.5,
|
|
"valid_targets_min": 1390
|
|
},
|
|
{
|
|
"epoch": 4.669984284965951,
|
|
"grad_norm": 0.6664424300772166,
|
|
"learning_rate": 1.2054069414503442e-05,
|
|
"loss": 0.1948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08218555897474289,
|
|
"step": 8915,
|
|
"valid_targets_mean": 2997.4,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 4.672603457307491,
|
|
"grad_norm": 0.5828292771920752,
|
|
"learning_rate": 1.2030103065670393e-05,
|
|
"loss": 0.2097,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10508102178573608,
|
|
"step": 8920,
|
|
"valid_targets_mean": 4428.5,
|
|
"valid_targets_min": 1901
|
|
},
|
|
{
|
|
"epoch": 4.675222629649031,
|
|
"grad_norm": 0.6601174887657705,
|
|
"learning_rate": 1.2006150314053393e-05,
|
|
"loss": 0.207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0941559448838234,
|
|
"step": 8925,
|
|
"valid_targets_mean": 2790.1,
|
|
"valid_targets_min": 1460
|
|
},
|
|
{
|
|
"epoch": 4.677841801990571,
|
|
"grad_norm": 0.6374464539298814,
|
|
"learning_rate": 1.1982211200517573e-05,
|
|
"loss": 0.2066,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11535796523094177,
|
|
"step": 8930,
|
|
"valid_targets_mean": 4607.4,
|
|
"valid_targets_min": 952
|
|
},
|
|
{
|
|
"epoch": 4.680460974332111,
|
|
"grad_norm": 0.7119024765620463,
|
|
"learning_rate": 1.195828576590476e-05,
|
|
"loss": 0.2022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09456755220890045,
|
|
"step": 8935,
|
|
"valid_targets_mean": 2605.1,
|
|
"valid_targets_min": 1551
|
|
},
|
|
{
|
|
"epoch": 4.683080146673651,
|
|
"grad_norm": 0.5781354468810069,
|
|
"learning_rate": 1.1934374051033482e-05,
|
|
"loss": 0.2022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08129958063364029,
|
|
"step": 8940,
|
|
"valid_targets_mean": 3755.8,
|
|
"valid_targets_min": 2267
|
|
},
|
|
{
|
|
"epoch": 4.685699319015191,
|
|
"grad_norm": 0.5551995108738447,
|
|
"learning_rate": 1.1910476096698828e-05,
|
|
"loss": 0.2036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08932093530893326,
|
|
"step": 8945,
|
|
"valid_targets_mean": 5198.8,
|
|
"valid_targets_min": 1233
|
|
},
|
|
{
|
|
"epoch": 4.688318491356731,
|
|
"grad_norm": 0.5508827780675954,
|
|
"learning_rate": 1.1886591943672418e-05,
|
|
"loss": 0.1925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0868104100227356,
|
|
"step": 8950,
|
|
"valid_targets_mean": 4899.5,
|
|
"valid_targets_min": 1741
|
|
},
|
|
{
|
|
"epoch": 4.690937663698271,
|
|
"grad_norm": 0.621708285770408,
|
|
"learning_rate": 1.1862721632702347e-05,
|
|
"loss": 0.2026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10903164744377136,
|
|
"step": 8955,
|
|
"valid_targets_mean": 4473.5,
|
|
"valid_targets_min": 856
|
|
},
|
|
{
|
|
"epoch": 4.693556836039812,
|
|
"grad_norm": 0.616603535846322,
|
|
"learning_rate": 1.1838865204513061e-05,
|
|
"loss": 0.1989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11983565986156464,
|
|
"step": 8960,
|
|
"valid_targets_mean": 4047.8,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 4.696176008381352,
|
|
"grad_norm": 0.559030691721046,
|
|
"learning_rate": 1.1815022699805355e-05,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08906766027212143,
|
|
"step": 8965,
|
|
"valid_targets_mean": 4106.0,
|
|
"valid_targets_min": 974
|
|
},
|
|
{
|
|
"epoch": 4.698795180722891,
|
|
"grad_norm": 0.631272737359263,
|
|
"learning_rate": 1.1791194159256235e-05,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09801031649112701,
|
|
"step": 8970,
|
|
"valid_targets_mean": 3092.2,
|
|
"valid_targets_min": 1447
|
|
},
|
|
{
|
|
"epoch": 4.701414353064432,
|
|
"grad_norm": 0.6340130444613701,
|
|
"learning_rate": 1.1767379623518919e-05,
|
|
"loss": 0.1898,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.106422558426857,
|
|
"step": 8975,
|
|
"valid_targets_mean": 4404.8,
|
|
"valid_targets_min": 1744
|
|
},
|
|
{
|
|
"epoch": 4.704033525405972,
|
|
"grad_norm": 0.6797587465450963,
|
|
"learning_rate": 1.1743579133222707e-05,
|
|
"loss": 0.1938,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11266933381557465,
|
|
"step": 8980,
|
|
"valid_targets_mean": 4678.9,
|
|
"valid_targets_min": 1515
|
|
},
|
|
{
|
|
"epoch": 4.706652697747511,
|
|
"grad_norm": 0.7201255929766432,
|
|
"learning_rate": 1.1719792728972932e-05,
|
|
"loss": 0.2118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12300004810094833,
|
|
"step": 8985,
|
|
"valid_targets_mean": 3396.1,
|
|
"valid_targets_min": 1349
|
|
},
|
|
{
|
|
"epoch": 4.709271870089052,
|
|
"grad_norm": 0.6591457046315632,
|
|
"learning_rate": 1.1696020451350924e-05,
|
|
"loss": 0.2008,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0994521751999855,
|
|
"step": 8990,
|
|
"valid_targets_mean": 3056.6,
|
|
"valid_targets_min": 1156
|
|
},
|
|
{
|
|
"epoch": 4.711891042430592,
|
|
"grad_norm": 0.6613931388079063,
|
|
"learning_rate": 1.1672262340913884e-05,
|
|
"loss": 0.194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1038329005241394,
|
|
"step": 8995,
|
|
"valid_targets_mean": 3856.4,
|
|
"valid_targets_min": 2076
|
|
},
|
|
{
|
|
"epoch": 4.714510214772132,
|
|
"grad_norm": 0.594735974520261,
|
|
"learning_rate": 1.1648518438194853e-05,
|
|
"loss": 0.1896,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09430505335330963,
|
|
"step": 9000,
|
|
"valid_targets_mean": 3352.8,
|
|
"valid_targets_min": 1021
|
|
},
|
|
{
|
|
"epoch": 4.717129387113673,
|
|
"grad_norm": 0.681225593238549,
|
|
"learning_rate": 1.1624788783702647e-05,
|
|
"loss": 0.1989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10264322906732559,
|
|
"step": 9005,
|
|
"valid_targets_mean": 3367.0,
|
|
"valid_targets_min": 654
|
|
},
|
|
{
|
|
"epoch": 4.719748559455212,
|
|
"grad_norm": 0.6283800600755348,
|
|
"learning_rate": 1.1601073417921729e-05,
|
|
"loss": 0.1931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10935042798519135,
|
|
"step": 9010,
|
|
"valid_targets_mean": 4698.5,
|
|
"valid_targets_min": 2336
|
|
},
|
|
{
|
|
"epoch": 4.722367731796752,
|
|
"grad_norm": 0.7413035203402667,
|
|
"learning_rate": 1.1577372381312251e-05,
|
|
"loss": 0.2067,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09140511602163315,
|
|
"step": 9015,
|
|
"valid_targets_mean": 2237.0,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 4.724986904138293,
|
|
"grad_norm": 0.5940483417790948,
|
|
"learning_rate": 1.155368571430985e-05,
|
|
"loss": 0.1994,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08947218954563141,
|
|
"step": 9020,
|
|
"valid_targets_mean": 3623.5,
|
|
"valid_targets_min": 2389
|
|
},
|
|
{
|
|
"epoch": 4.727606076479832,
|
|
"grad_norm": 0.6492608693015333,
|
|
"learning_rate": 1.1530013457325708e-05,
|
|
"loss": 0.2028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10537391901016235,
|
|
"step": 9025,
|
|
"valid_targets_mean": 3766.9,
|
|
"valid_targets_min": 1793
|
|
},
|
|
{
|
|
"epoch": 4.730225248821372,
|
|
"grad_norm": 0.6693284764498884,
|
|
"learning_rate": 1.1506355650746364e-05,
|
|
"loss": 0.2031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0952562540769577,
|
|
"step": 9030,
|
|
"valid_targets_mean": 3133.1,
|
|
"valid_targets_min": 1217
|
|
},
|
|
{
|
|
"epoch": 4.732844421162913,
|
|
"grad_norm": 0.6464798992791625,
|
|
"learning_rate": 1.1482712334933769e-05,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06148266792297363,
|
|
"step": 9035,
|
|
"valid_targets_mean": 1525.2,
|
|
"valid_targets_min": 534
|
|
},
|
|
{
|
|
"epoch": 4.735463593504453,
|
|
"grad_norm": 0.3997519225286325,
|
|
"learning_rate": 1.1459083550225088e-05,
|
|
"loss": 0.1007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03425814211368561,
|
|
"step": 9040,
|
|
"valid_targets_mean": 2735.4,
|
|
"valid_targets_min": 834
|
|
},
|
|
{
|
|
"epoch": 4.738082765845992,
|
|
"grad_norm": 0.3805866264698184,
|
|
"learning_rate": 1.1435469336932731e-05,
|
|
"loss": 0.0832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03802028298377991,
|
|
"step": 9045,
|
|
"valid_targets_mean": 3253.9,
|
|
"valid_targets_min": 2790
|
|
},
|
|
{
|
|
"epoch": 4.740701938187533,
|
|
"grad_norm": 0.5271227522994802,
|
|
"learning_rate": 1.141186973534424e-05,
|
|
"loss": 0.0779,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038571856915950775,
|
|
"step": 9050,
|
|
"valid_targets_mean": 3383.1,
|
|
"valid_targets_min": 2133
|
|
},
|
|
{
|
|
"epoch": 4.743321110529073,
|
|
"grad_norm": 0.42506608635375265,
|
|
"learning_rate": 1.1388284785722235e-05,
|
|
"loss": 0.0968,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04349125921726227,
|
|
"step": 9055,
|
|
"valid_targets_mean": 2954.8,
|
|
"valid_targets_min": 2338
|
|
},
|
|
{
|
|
"epoch": 4.745940282870613,
|
|
"grad_norm": 0.4400398864057771,
|
|
"learning_rate": 1.1364714528304325e-05,
|
|
"loss": 0.0762,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039018213748931885,
|
|
"step": 9060,
|
|
"valid_targets_mean": 2680.4,
|
|
"valid_targets_min": 662
|
|
},
|
|
{
|
|
"epoch": 4.748559455212153,
|
|
"grad_norm": 1.0406638024773125,
|
|
"learning_rate": 1.1341159003303075e-05,
|
|
"loss": 0.1052,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07803843915462494,
|
|
"step": 9065,
|
|
"valid_targets_mean": 1101.0,
|
|
"valid_targets_min": 559
|
|
},
|
|
{
|
|
"epoch": 4.751178627553693,
|
|
"grad_norm": 0.4234808201914869,
|
|
"learning_rate": 1.131761825090588e-05,
|
|
"loss": 0.1146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035465411841869354,
|
|
"step": 9070,
|
|
"valid_targets_mean": 3490.6,
|
|
"valid_targets_min": 1091
|
|
},
|
|
{
|
|
"epoch": 4.753797799895233,
|
|
"grad_norm": 0.4052125616024461,
|
|
"learning_rate": 1.129409231127496e-05,
|
|
"loss": 0.0871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04046661779284477,
|
|
"step": 9075,
|
|
"valid_targets_mean": 3061.8,
|
|
"valid_targets_min": 947
|
|
},
|
|
{
|
|
"epoch": 4.7564169722367735,
|
|
"grad_norm": 0.5151153926725468,
|
|
"learning_rate": 1.1270581224547258e-05,
|
|
"loss": 0.1124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09529449045658112,
|
|
"step": 9080,
|
|
"valid_targets_mean": 2125.4,
|
|
"valid_targets_min": 859
|
|
},
|
|
{
|
|
"epoch": 4.759036144578313,
|
|
"grad_norm": 0.30622345996440775,
|
|
"learning_rate": 1.1247085030834378e-05,
|
|
"loss": 0.0798,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043526582419872284,
|
|
"step": 9085,
|
|
"valid_targets_mean": 5837.8,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 4.761655316919853,
|
|
"grad_norm": 0.3911448819762176,
|
|
"learning_rate": 1.1223603770222509e-05,
|
|
"loss": 0.0785,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043331749737262726,
|
|
"step": 9090,
|
|
"valid_targets_mean": 2185.0,
|
|
"valid_targets_min": 826
|
|
},
|
|
{
|
|
"epoch": 4.7642744892613935,
|
|
"grad_norm": 0.3070158190135269,
|
|
"learning_rate": 1.1200137482772369e-05,
|
|
"loss": 0.0778,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03534020483493805,
|
|
"step": 9095,
|
|
"valid_targets_mean": 3446.5,
|
|
"valid_targets_min": 1096
|
|
},
|
|
{
|
|
"epoch": 4.766893661602934,
|
|
"grad_norm": 0.4828342763934001,
|
|
"learning_rate": 1.1176686208519131e-05,
|
|
"loss": 0.096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04079655930399895,
|
|
"step": 9100,
|
|
"valid_targets_mean": 2343.1,
|
|
"valid_targets_min": 688
|
|
},
|
|
{
|
|
"epoch": 4.769512833944473,
|
|
"grad_norm": 0.3721916673928446,
|
|
"learning_rate": 1.1153249987472332e-05,
|
|
"loss": 0.097,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028864160180091858,
|
|
"step": 9105,
|
|
"valid_targets_mean": 2938.5,
|
|
"valid_targets_min": 553
|
|
},
|
|
{
|
|
"epoch": 4.7721320062860135,
|
|
"grad_norm": 0.4579098193653015,
|
|
"learning_rate": 1.1129828859615879e-05,
|
|
"loss": 0.0802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04318414255976677,
|
|
"step": 9110,
|
|
"valid_targets_mean": 2988.9,
|
|
"valid_targets_min": 566
|
|
},
|
|
{
|
|
"epoch": 4.774751178627554,
|
|
"grad_norm": 0.38923712166486607,
|
|
"learning_rate": 1.1106422864907863e-05,
|
|
"loss": 0.0739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029954858124256134,
|
|
"step": 9115,
|
|
"valid_targets_mean": 3230.2,
|
|
"valid_targets_min": 1111
|
|
},
|
|
{
|
|
"epoch": 4.777370350969094,
|
|
"grad_norm": 0.4561912850348876,
|
|
"learning_rate": 1.1083032043280624e-05,
|
|
"loss": 0.0786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05434397608041763,
|
|
"step": 9120,
|
|
"valid_targets_mean": 1232.1,
|
|
"valid_targets_min": 622
|
|
},
|
|
{
|
|
"epoch": 4.7799895233106335,
|
|
"grad_norm": 0.5383278388405104,
|
|
"learning_rate": 1.1059656434640557e-05,
|
|
"loss": 0.09,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042419224977493286,
|
|
"step": 9125,
|
|
"valid_targets_mean": 2132.5,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 4.782608695652174,
|
|
"grad_norm": 0.46478132442947867,
|
|
"learning_rate": 1.1036296078868136e-05,
|
|
"loss": 0.0881,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0451333187520504,
|
|
"step": 9130,
|
|
"valid_targets_mean": 3225.2,
|
|
"valid_targets_min": 1922
|
|
},
|
|
{
|
|
"epoch": 4.785227867993714,
|
|
"grad_norm": 0.4698233923807494,
|
|
"learning_rate": 1.1012951015817804e-05,
|
|
"loss": 0.0891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032914064824581146,
|
|
"step": 9135,
|
|
"valid_targets_mean": 2008.5,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 4.787847040335254,
|
|
"grad_norm": 0.5308778820315359,
|
|
"learning_rate": 1.0989621285317913e-05,
|
|
"loss": 0.0962,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06775735318660736,
|
|
"step": 9140,
|
|
"valid_targets_mean": 1754.1,
|
|
"valid_targets_min": 957
|
|
},
|
|
{
|
|
"epoch": 4.790466212676794,
|
|
"grad_norm": 0.4084670385575034,
|
|
"learning_rate": 1.0966306927170658e-05,
|
|
"loss": 0.1154,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043514907360076904,
|
|
"step": 9145,
|
|
"valid_targets_mean": 2498.5,
|
|
"valid_targets_min": 641
|
|
},
|
|
{
|
|
"epoch": 4.793085385018334,
|
|
"grad_norm": 0.4279232378741644,
|
|
"learning_rate": 1.094300798115201e-05,
|
|
"loss": 0.0756,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04977592080831528,
|
|
"step": 9150,
|
|
"valid_targets_mean": 2364.6,
|
|
"valid_targets_min": 924
|
|
},
|
|
{
|
|
"epoch": 4.795704557359874,
|
|
"grad_norm": 0.5275163178993386,
|
|
"learning_rate": 1.0919724487011636e-05,
|
|
"loss": 0.1134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08120264858007431,
|
|
"step": 9155,
|
|
"valid_targets_mean": 2385.5,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 4.798323729701415,
|
|
"grad_norm": 0.49257488007398736,
|
|
"learning_rate": 1.0896456484472862e-05,
|
|
"loss": 0.0941,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07682208716869354,
|
|
"step": 9160,
|
|
"valid_targets_mean": 3363.2,
|
|
"valid_targets_min": 2731
|
|
},
|
|
{
|
|
"epoch": 4.800942902042954,
|
|
"grad_norm": 0.5415718875503004,
|
|
"learning_rate": 1.0873204013232544e-05,
|
|
"loss": 0.084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07005016505718231,
|
|
"step": 9165,
|
|
"valid_targets_mean": 2615.0,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 4.8035620743844945,
|
|
"grad_norm": 0.6949620057147603,
|
|
"learning_rate": 1.0849967112961097e-05,
|
|
"loss": 0.2478,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.26949775218963623,
|
|
"step": 9170,
|
|
"valid_targets_mean": 2615.9,
|
|
"valid_targets_min": 922
|
|
},
|
|
{
|
|
"epoch": 4.806181246726035,
|
|
"grad_norm": 0.4735129998424099,
|
|
"learning_rate": 1.0826745823302312e-05,
|
|
"loss": 0.1188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03111286647617817,
|
|
"step": 9175,
|
|
"valid_targets_mean": 1357.4,
|
|
"valid_targets_min": 507
|
|
},
|
|
{
|
|
"epoch": 4.808800419067575,
|
|
"grad_norm": 0.41949657373235166,
|
|
"learning_rate": 1.0803540183873399e-05,
|
|
"loss": 0.0969,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04782308638095856,
|
|
"step": 9180,
|
|
"valid_targets_mean": 3236.9,
|
|
"valid_targets_min": 615
|
|
},
|
|
{
|
|
"epoch": 4.8114195914091145,
|
|
"grad_norm": 0.4706428085338741,
|
|
"learning_rate": 1.078035023426483e-05,
|
|
"loss": 0.1776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06388384103775024,
|
|
"step": 9185,
|
|
"valid_targets_mean": 3023.1,
|
|
"valid_targets_min": 826
|
|
},
|
|
{
|
|
"epoch": 4.814038763750655,
|
|
"grad_norm": 0.4592019713471492,
|
|
"learning_rate": 1.0757176014040325e-05,
|
|
"loss": 0.0824,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04213862866163254,
|
|
"step": 9190,
|
|
"valid_targets_mean": 2924.4,
|
|
"valid_targets_min": 851
|
|
},
|
|
{
|
|
"epoch": 4.816657936092195,
|
|
"grad_norm": 0.4632823742735354,
|
|
"learning_rate": 1.073401756273677e-05,
|
|
"loss": 0.0831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05453056842088699,
|
|
"step": 9195,
|
|
"valid_targets_mean": 4528.4,
|
|
"valid_targets_min": 2665
|
|
},
|
|
{
|
|
"epoch": 4.8192771084337345,
|
|
"grad_norm": 0.3609215644783444,
|
|
"learning_rate": 1.0710874919864143e-05,
|
|
"loss": 0.0856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02940385788679123,
|
|
"step": 9200,
|
|
"valid_targets_mean": 2820.6,
|
|
"valid_targets_min": 595
|
|
},
|
|
{
|
|
"epoch": 4.821896280775275,
|
|
"grad_norm": 0.44209632448122455,
|
|
"learning_rate": 1.0687748124905451e-05,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05022032558917999,
|
|
"step": 9205,
|
|
"valid_targets_mean": 2583.1,
|
|
"valid_targets_min": 825
|
|
},
|
|
{
|
|
"epoch": 4.824515453116815,
|
|
"grad_norm": 0.44249999142815694,
|
|
"learning_rate": 1.066463721731667e-05,
|
|
"loss": 0.0786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03852751478552818,
|
|
"step": 9210,
|
|
"valid_targets_mean": 3771.5,
|
|
"valid_targets_min": 672
|
|
},
|
|
{
|
|
"epoch": 4.827134625458355,
|
|
"grad_norm": 0.3611442497021939,
|
|
"learning_rate": 1.0641542236526672e-05,
|
|
"loss": 0.0729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029995795339345932,
|
|
"step": 9215,
|
|
"valid_targets_mean": 3165.0,
|
|
"valid_targets_min": 1956
|
|
},
|
|
{
|
|
"epoch": 4.829753797799896,
|
|
"grad_norm": 0.6706741443758514,
|
|
"learning_rate": 1.0618463221937136e-05,
|
|
"loss": 0.0802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03590845689177513,
|
|
"step": 9220,
|
|
"valid_targets_mean": 3719.8,
|
|
"valid_targets_min": 2540
|
|
},
|
|
{
|
|
"epoch": 4.832372970141435,
|
|
"grad_norm": 0.45590454833112176,
|
|
"learning_rate": 1.0595400212922526e-05,
|
|
"loss": 0.0705,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04903195798397064,
|
|
"step": 9225,
|
|
"valid_targets_mean": 3447.5,
|
|
"valid_targets_min": 1240
|
|
},
|
|
{
|
|
"epoch": 4.834992142482975,
|
|
"grad_norm": 0.42236440086752497,
|
|
"learning_rate": 1.0572353248829988e-05,
|
|
"loss": 0.1105,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.022666532546281815,
|
|
"step": 9230,
|
|
"valid_targets_mean": 2975.2,
|
|
"valid_targets_min": 623
|
|
},
|
|
{
|
|
"epoch": 4.837611314824516,
|
|
"grad_norm": 0.3818881079962277,
|
|
"learning_rate": 1.0549322368979298e-05,
|
|
"loss": 0.0668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03372521325945854,
|
|
"step": 9235,
|
|
"valid_targets_mean": 2960.1,
|
|
"valid_targets_min": 583
|
|
},
|
|
{
|
|
"epoch": 4.840230487166055,
|
|
"grad_norm": 0.3751963558558432,
|
|
"learning_rate": 1.0526307612662792e-05,
|
|
"loss": 0.0715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03359278291463852,
|
|
"step": 9240,
|
|
"valid_targets_mean": 3723.5,
|
|
"valid_targets_min": 3230
|
|
},
|
|
{
|
|
"epoch": 4.842849659507595,
|
|
"grad_norm": 0.4184602801803382,
|
|
"learning_rate": 1.05033090191453e-05,
|
|
"loss": 0.0693,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028666600584983826,
|
|
"step": 9245,
|
|
"valid_targets_mean": 2520.9,
|
|
"valid_targets_min": 647
|
|
},
|
|
{
|
|
"epoch": 4.845468831849136,
|
|
"grad_norm": 0.7642504032655404,
|
|
"learning_rate": 1.048032662766405e-05,
|
|
"loss": 0.1154,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08870033919811249,
|
|
"step": 9250,
|
|
"valid_targets_mean": 1094.5,
|
|
"valid_targets_min": 195
|
|
},
|
|
{
|
|
"epoch": 4.848088004190676,
|
|
"grad_norm": 0.3969633343007265,
|
|
"learning_rate": 1.0457360477428682e-05,
|
|
"loss": 0.0782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05737834423780441,
|
|
"step": 9255,
|
|
"valid_targets_mean": 3635.9,
|
|
"valid_targets_min": 3050
|
|
},
|
|
{
|
|
"epoch": 4.850707176532216,
|
|
"grad_norm": 0.5763074321343137,
|
|
"learning_rate": 1.0434410607621065e-05,
|
|
"loss": 0.1106,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06752394139766693,
|
|
"step": 9260,
|
|
"valid_targets_mean": 2382.9,
|
|
"valid_targets_min": 1087
|
|
},
|
|
{
|
|
"epoch": 4.853326348873756,
|
|
"grad_norm": 0.4031647724187547,
|
|
"learning_rate": 1.0411477057395358e-05,
|
|
"loss": 0.0786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034291207790374756,
|
|
"step": 9265,
|
|
"valid_targets_mean": 3025.5,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 4.855945521215296,
|
|
"grad_norm": 0.5587627567342315,
|
|
"learning_rate": 1.0388559865877816e-05,
|
|
"loss": 0.1038,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05773116275668144,
|
|
"step": 9270,
|
|
"valid_targets_mean": 2263.1,
|
|
"valid_targets_min": 916
|
|
},
|
|
{
|
|
"epoch": 4.858564693556836,
|
|
"grad_norm": 0.46595019245947983,
|
|
"learning_rate": 1.0365659072166823e-05,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03987809270620346,
|
|
"step": 9275,
|
|
"valid_targets_mean": 2567.0,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 4.861183865898376,
|
|
"grad_norm": 0.6864424953181596,
|
|
"learning_rate": 1.0342774715332776e-05,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06730234622955322,
|
|
"step": 9280,
|
|
"valid_targets_mean": 1443.4,
|
|
"valid_targets_min": 771
|
|
},
|
|
{
|
|
"epoch": 4.863803038239916,
|
|
"grad_norm": 0.4289349949699643,
|
|
"learning_rate": 1.0319906834418027e-05,
|
|
"loss": 0.0791,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.036795586347579956,
|
|
"step": 9285,
|
|
"valid_targets_mean": 2753.2,
|
|
"valid_targets_min": 707
|
|
},
|
|
{
|
|
"epoch": 4.866422210581456,
|
|
"grad_norm": 0.38043262212418566,
|
|
"learning_rate": 1.0297055468436823e-05,
|
|
"loss": 0.0854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038165103644132614,
|
|
"step": 9290,
|
|
"valid_targets_mean": 3665.8,
|
|
"valid_targets_min": 2850
|
|
},
|
|
{
|
|
"epoch": 4.869041382922997,
|
|
"grad_norm": 0.43476779566750035,
|
|
"learning_rate": 1.0274220656375235e-05,
|
|
"loss": 0.0842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03128594160079956,
|
|
"step": 9295,
|
|
"valid_targets_mean": 2238.2,
|
|
"valid_targets_min": 749
|
|
},
|
|
{
|
|
"epoch": 4.871660555264537,
|
|
"grad_norm": 0.370313368646326,
|
|
"learning_rate": 1.0251402437191089e-05,
|
|
"loss": 0.0708,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038706958293914795,
|
|
"step": 9300,
|
|
"valid_targets_mean": 4488.2,
|
|
"valid_targets_min": 1052
|
|
},
|
|
{
|
|
"epoch": 4.874279727606076,
|
|
"grad_norm": 0.319750712033594,
|
|
"learning_rate": 1.0228600849813915e-05,
|
|
"loss": 0.0671,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.036063797771930695,
|
|
"step": 9305,
|
|
"valid_targets_mean": 4258.2,
|
|
"valid_targets_min": 2540
|
|
},
|
|
{
|
|
"epoch": 4.876898899947617,
|
|
"grad_norm": 0.33029579924082025,
|
|
"learning_rate": 1.0205815933144833e-05,
|
|
"loss": 0.0799,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029350923374295235,
|
|
"step": 9310,
|
|
"valid_targets_mean": 3782.5,
|
|
"valid_targets_min": 1859
|
|
},
|
|
{
|
|
"epoch": 4.879518072289157,
|
|
"grad_norm": 0.38085938971164035,
|
|
"learning_rate": 1.0183047726056574e-05,
|
|
"loss": 0.0938,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03152487426996231,
|
|
"step": 9315,
|
|
"valid_targets_mean": 2879.5,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 4.882137244630696,
|
|
"grad_norm": 0.4831945174173879,
|
|
"learning_rate": 1.0160296267393308e-05,
|
|
"loss": 0.0889,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04167293757200241,
|
|
"step": 9320,
|
|
"valid_targets_mean": 2072.6,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 4.884756416972237,
|
|
"grad_norm": 0.3146208676922227,
|
|
"learning_rate": 1.013756159597068e-05,
|
|
"loss": 0.0827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.024426061660051346,
|
|
"step": 9325,
|
|
"valid_targets_mean": 4205.6,
|
|
"valid_targets_min": 3934
|
|
},
|
|
{
|
|
"epoch": 4.887375589313777,
|
|
"grad_norm": 0.4519748356721149,
|
|
"learning_rate": 1.0114843750575654e-05,
|
|
"loss": 0.0738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04126317426562309,
|
|
"step": 9330,
|
|
"valid_targets_mean": 3174.9,
|
|
"valid_targets_min": 1576
|
|
},
|
|
{
|
|
"epoch": 4.889994761655317,
|
|
"grad_norm": 0.48719835907350145,
|
|
"learning_rate": 1.0092142769966506e-05,
|
|
"loss": 0.0862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04686902463436127,
|
|
"step": 9335,
|
|
"valid_targets_mean": 3241.8,
|
|
"valid_targets_min": 1006
|
|
},
|
|
{
|
|
"epoch": 4.892613933996857,
|
|
"grad_norm": 0.32010999479451724,
|
|
"learning_rate": 1.0069458692872742e-05,
|
|
"loss": 0.0714,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02505858615040779,
|
|
"step": 9340,
|
|
"valid_targets_mean": 3445.5,
|
|
"valid_targets_min": 980
|
|
},
|
|
{
|
|
"epoch": 4.895233106338397,
|
|
"grad_norm": 0.5725901656192572,
|
|
"learning_rate": 1.0046791557995023e-05,
|
|
"loss": 0.0804,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05188882350921631,
|
|
"step": 9345,
|
|
"valid_targets_mean": 2083.9,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 4.897852278679937,
|
|
"grad_norm": 0.5549702485809708,
|
|
"learning_rate": 1.002414140400511e-05,
|
|
"loss": 0.0788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050374835729599,
|
|
"step": 9350,
|
|
"valid_targets_mean": 2768.8,
|
|
"valid_targets_min": 1020
|
|
},
|
|
{
|
|
"epoch": 4.9004714510214775,
|
|
"grad_norm": 0.4836784282793694,
|
|
"learning_rate": 1.000150826954579e-05,
|
|
"loss": 0.0797,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051911644637584686,
|
|
"step": 9355,
|
|
"valid_targets_mean": 2314.8,
|
|
"valid_targets_min": 967
|
|
},
|
|
{
|
|
"epoch": 4.903090623363017,
|
|
"grad_norm": 0.4360657137850092,
|
|
"learning_rate": 9.978892193230821e-06,
|
|
"loss": 0.0971,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03503736853599548,
|
|
"step": 9360,
|
|
"valid_targets_mean": 2462.0,
|
|
"valid_targets_min": 1015
|
|
},
|
|
{
|
|
"epoch": 4.905709795704557,
|
|
"grad_norm": 0.5904512749187,
|
|
"learning_rate": 9.956293213644839e-06,
|
|
"loss": 0.0774,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044043950736522675,
|
|
"step": 9365,
|
|
"valid_targets_mean": 1214.9,
|
|
"valid_targets_min": 705
|
|
},
|
|
{
|
|
"epoch": 4.9083289680460975,
|
|
"grad_norm": 0.3881901935145882,
|
|
"learning_rate": 9.933711369343329e-06,
|
|
"loss": 0.0883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02765190787613392,
|
|
"step": 9370,
|
|
"valid_targets_mean": 2364.8,
|
|
"valid_targets_min": 622
|
|
},
|
|
{
|
|
"epoch": 4.910948140387638,
|
|
"grad_norm": 0.5521495497495682,
|
|
"learning_rate": 9.91114669885254e-06,
|
|
"loss": 0.1031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033989161252975464,
|
|
"step": 9375,
|
|
"valid_targets_mean": 3702.2,
|
|
"valid_targets_min": 1322
|
|
},
|
|
{
|
|
"epoch": 4.913567312729177,
|
|
"grad_norm": 0.41071973811584744,
|
|
"learning_rate": 9.888599240669419e-06,
|
|
"loss": 0.0804,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04003782570362091,
|
|
"step": 9380,
|
|
"valid_targets_mean": 2863.1,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 4.9161864850707175,
|
|
"grad_norm": 0.49161604591708735,
|
|
"learning_rate": 9.866069033261548e-06,
|
|
"loss": 0.0701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044861674308776855,
|
|
"step": 9385,
|
|
"valid_targets_mean": 3106.1,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 4.918805657412258,
|
|
"grad_norm": 0.583969785929338,
|
|
"learning_rate": 9.843556115067083e-06,
|
|
"loss": 0.0848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0578463077545166,
|
|
"step": 9390,
|
|
"valid_targets_mean": 3161.4,
|
|
"valid_targets_min": 1420
|
|
},
|
|
{
|
|
"epoch": 4.921424829753798,
|
|
"grad_norm": 0.4137817721617095,
|
|
"learning_rate": 9.821060524494658e-06,
|
|
"loss": 0.0959,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03279948607087135,
|
|
"step": 9395,
|
|
"valid_targets_mean": 2796.0,
|
|
"valid_targets_min": 1109
|
|
},
|
|
{
|
|
"epoch": 4.9240440020953375,
|
|
"grad_norm": 0.40270192516132824,
|
|
"learning_rate": 9.798582299923393e-06,
|
|
"loss": 0.0629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03196260333061218,
|
|
"step": 9400,
|
|
"valid_targets_mean": 1822.9,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 4.926663174436878,
|
|
"grad_norm": 0.46867222941763004,
|
|
"learning_rate": 9.776121479702722e-06,
|
|
"loss": 0.0825,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03990498185157776,
|
|
"step": 9405,
|
|
"valid_targets_mean": 2389.8,
|
|
"valid_targets_min": 671
|
|
},
|
|
{
|
|
"epoch": 4.929282346778418,
|
|
"grad_norm": 0.36127700554593084,
|
|
"learning_rate": 9.753678102152449e-06,
|
|
"loss": 0.0727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034535013139247894,
|
|
"step": 9410,
|
|
"valid_targets_mean": 4513.8,
|
|
"valid_targets_min": 1057
|
|
},
|
|
{
|
|
"epoch": 4.931901519119958,
|
|
"grad_norm": 0.3633748308439699,
|
|
"learning_rate": 9.731252205562551e-06,
|
|
"loss": 0.0643,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030670013278722763,
|
|
"step": 9415,
|
|
"valid_targets_mean": 3533.1,
|
|
"valid_targets_min": 1243
|
|
},
|
|
{
|
|
"epoch": 4.934520691461498,
|
|
"grad_norm": 0.3800339958338929,
|
|
"learning_rate": 9.708843828193251e-06,
|
|
"loss": 0.086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03751269355416298,
|
|
"step": 9420,
|
|
"valid_targets_mean": 3911.2,
|
|
"valid_targets_min": 3322
|
|
},
|
|
{
|
|
"epoch": 4.937139863803038,
|
|
"grad_norm": 0.5273963512085883,
|
|
"learning_rate": 9.686453008274825e-06,
|
|
"loss": 0.0728,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04546033591032028,
|
|
"step": 9425,
|
|
"valid_targets_mean": 1820.0,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 4.9397590361445785,
|
|
"grad_norm": 0.43278411601206196,
|
|
"learning_rate": 9.664079784007627e-06,
|
|
"loss": 0.1015,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043846093118190765,
|
|
"step": 9430,
|
|
"valid_targets_mean": 3371.9,
|
|
"valid_targets_min": 640
|
|
},
|
|
{
|
|
"epoch": 4.942378208486119,
|
|
"grad_norm": 0.27838898081111807,
|
|
"learning_rate": 9.64172419356198e-06,
|
|
"loss": 0.0881,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02142597734928131,
|
|
"step": 9435,
|
|
"valid_targets_mean": 3327.0,
|
|
"valid_targets_min": 929
|
|
},
|
|
{
|
|
"epoch": 4.944997380827658,
|
|
"grad_norm": 0.6142096335256088,
|
|
"learning_rate": 9.619386275078129e-06,
|
|
"loss": 0.0925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05631929636001587,
|
|
"step": 9440,
|
|
"valid_targets_mean": 3318.0,
|
|
"valid_targets_min": 2380
|
|
},
|
|
{
|
|
"epoch": 4.9476165531691985,
|
|
"grad_norm": 0.38334975674724514,
|
|
"learning_rate": 9.597066066666164e-06,
|
|
"loss": 0.0717,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03687911853194237,
|
|
"step": 9445,
|
|
"valid_targets_mean": 3709.4,
|
|
"valid_targets_min": 2619
|
|
},
|
|
{
|
|
"epoch": 4.950235725510739,
|
|
"grad_norm": 0.67861113279472,
|
|
"learning_rate": 9.574763606405968e-06,
|
|
"loss": 0.0815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049105964601039886,
|
|
"step": 9450,
|
|
"valid_targets_mean": 1574.6,
|
|
"valid_targets_min": 693
|
|
},
|
|
{
|
|
"epoch": 4.952854897852279,
|
|
"grad_norm": 0.3957068978774986,
|
|
"learning_rate": 9.55247893234712e-06,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041627075523138046,
|
|
"step": 9455,
|
|
"valid_targets_mean": 3885.1,
|
|
"valid_targets_min": 3129
|
|
},
|
|
{
|
|
"epoch": 4.9554740701938185,
|
|
"grad_norm": 0.7326783103878073,
|
|
"learning_rate": 9.5302120825089e-06,
|
|
"loss": 0.1037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04959501326084137,
|
|
"step": 9460,
|
|
"valid_targets_mean": 1611.0,
|
|
"valid_targets_min": 543
|
|
},
|
|
{
|
|
"epoch": 4.958093242535359,
|
|
"grad_norm": 0.44825197019491486,
|
|
"learning_rate": 9.507963094880124e-06,
|
|
"loss": 0.1051,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047432661056518555,
|
|
"step": 9465,
|
|
"valid_targets_mean": 2319.6,
|
|
"valid_targets_min": 503
|
|
},
|
|
{
|
|
"epoch": 4.960712414876899,
|
|
"grad_norm": 0.510071533666162,
|
|
"learning_rate": 9.485732007419188e-06,
|
|
"loss": 0.0827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053640030324459076,
|
|
"step": 9470,
|
|
"valid_targets_mean": 1987.9,
|
|
"valid_targets_min": 625
|
|
},
|
|
{
|
|
"epoch": 4.963331587218439,
|
|
"grad_norm": 0.43448292051944704,
|
|
"learning_rate": 9.463518858053904e-06,
|
|
"loss": 0.0786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03527310863137245,
|
|
"step": 9475,
|
|
"valid_targets_mean": 3822.8,
|
|
"valid_targets_min": 2822
|
|
},
|
|
{
|
|
"epoch": 4.965950759559979,
|
|
"grad_norm": 0.42248384226037244,
|
|
"learning_rate": 9.441323684681502e-06,
|
|
"loss": 0.0773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046227335929870605,
|
|
"step": 9480,
|
|
"valid_targets_mean": 3724.5,
|
|
"valid_targets_min": 2987
|
|
},
|
|
{
|
|
"epoch": 4.968569931901519,
|
|
"grad_norm": 0.382801073687794,
|
|
"learning_rate": 9.419146525168545e-06,
|
|
"loss": 0.0685,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028592493385076523,
|
|
"step": 9485,
|
|
"valid_targets_mean": 3022.0,
|
|
"valid_targets_min": 856
|
|
},
|
|
{
|
|
"epoch": 4.971189104243059,
|
|
"grad_norm": 0.9587364146007591,
|
|
"learning_rate": 9.396987417350856e-06,
|
|
"loss": 0.0807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0448029451072216,
|
|
"step": 9490,
|
|
"valid_targets_mean": 764.2,
|
|
"valid_targets_min": 563
|
|
},
|
|
{
|
|
"epoch": 4.973808276584599,
|
|
"grad_norm": 0.39898776336005193,
|
|
"learning_rate": 9.374846399033469e-06,
|
|
"loss": 0.0795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033013638108968735,
|
|
"step": 9495,
|
|
"valid_targets_mean": 3117.1,
|
|
"valid_targets_min": 751
|
|
},
|
|
{
|
|
"epoch": 4.976427448926139,
|
|
"grad_norm": 0.5255979997506766,
|
|
"learning_rate": 9.352723507990528e-06,
|
|
"loss": 0.0714,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046126432716846466,
|
|
"step": 9500,
|
|
"valid_targets_mean": 3566.1,
|
|
"valid_targets_min": 980
|
|
},
|
|
{
|
|
"epoch": 4.979046621267679,
|
|
"grad_norm": 0.446781363338972,
|
|
"learning_rate": 9.330618781965305e-06,
|
|
"loss": 0.0727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038315288722515106,
|
|
"step": 9505,
|
|
"valid_targets_mean": 3589.8,
|
|
"valid_targets_min": 2538
|
|
},
|
|
{
|
|
"epoch": 4.98166579360922,
|
|
"grad_norm": 0.3124386468514202,
|
|
"learning_rate": 9.308532258670025e-06,
|
|
"loss": 0.0806,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027780020609498024,
|
|
"step": 9510,
|
|
"valid_targets_mean": 4182.8,
|
|
"valid_targets_min": 485
|
|
},
|
|
{
|
|
"epoch": 4.98428496595076,
|
|
"grad_norm": 0.30054591798536434,
|
|
"learning_rate": 9.28646397578589e-06,
|
|
"loss": 0.069,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033520281314849854,
|
|
"step": 9515,
|
|
"valid_targets_mean": 4832.8,
|
|
"valid_targets_min": 2551
|
|
},
|
|
{
|
|
"epoch": 4.986904138292299,
|
|
"grad_norm": 0.33328379929575486,
|
|
"learning_rate": 9.264413970962969e-06,
|
|
"loss": 0.0742,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037390343844890594,
|
|
"step": 9520,
|
|
"valid_targets_mean": 4733.6,
|
|
"valid_targets_min": 3365
|
|
},
|
|
{
|
|
"epoch": 4.98952331063384,
|
|
"grad_norm": 0.44878278869138466,
|
|
"learning_rate": 9.24238228182016e-06,
|
|
"loss": 0.0791,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05923148989677429,
|
|
"step": 9525,
|
|
"valid_targets_mean": 2954.4,
|
|
"valid_targets_min": 872
|
|
},
|
|
{
|
|
"epoch": 4.99214248297538,
|
|
"grad_norm": 0.3811491511881796,
|
|
"learning_rate": 9.220368945945103e-06,
|
|
"loss": 0.1636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03844798728823662,
|
|
"step": 9530,
|
|
"valid_targets_mean": 2590.9,
|
|
"valid_targets_min": 1015
|
|
},
|
|
{
|
|
"epoch": 4.994761655316919,
|
|
"grad_norm": 0.4348334992325256,
|
|
"learning_rate": 9.198374000894123e-06,
|
|
"loss": 0.074,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05938860401511192,
|
|
"step": 9535,
|
|
"valid_targets_mean": 2342.4,
|
|
"valid_targets_min": 967
|
|
},
|
|
{
|
|
"epoch": 4.99738082765846,
|
|
"grad_norm": 0.39936657461229824,
|
|
"learning_rate": 9.176397484192184e-06,
|
|
"loss": 0.076,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03136672079563141,
|
|
"step": 9540,
|
|
"valid_targets_mean": 3335.1,
|
|
"valid_targets_min": 886
|
|
},
|
|
{
|
|
"epoch": 5.0,
|
|
"grad_norm": 0.3896753621417805,
|
|
"learning_rate": 9.1544394333328e-06,
|
|
"loss": 0.1058,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03392448276281357,
|
|
"step": 9545,
|
|
"valid_targets_mean": 2983.1,
|
|
"valid_targets_min": 791
|
|
},
|
|
{
|
|
"epoch": 5.00261917234154,
|
|
"grad_norm": 0.8291698685938063,
|
|
"learning_rate": 9.132499885777964e-06,
|
|
"loss": 0.1593,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06088561937212944,
|
|
"step": 9550,
|
|
"valid_targets_mean": 1236.5,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 5.00523834468308,
|
|
"grad_norm": 0.795133667121655,
|
|
"learning_rate": 9.110578878958145e-06,
|
|
"loss": 0.1432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07181090116500854,
|
|
"step": 9555,
|
|
"valid_targets_mean": 1550.8,
|
|
"valid_targets_min": 612
|
|
},
|
|
{
|
|
"epoch": 5.00785751702462,
|
|
"grad_norm": 0.7188523527435988,
|
|
"learning_rate": 9.088676450272124e-06,
|
|
"loss": 0.1396,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06605690717697144,
|
|
"step": 9560,
|
|
"valid_targets_mean": 1455.8,
|
|
"valid_targets_min": 904
|
|
},
|
|
{
|
|
"epoch": 5.01047668936616,
|
|
"grad_norm": 0.7499204984155243,
|
|
"learning_rate": 9.06679263708705e-06,
|
|
"loss": 0.1315,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06748449802398682,
|
|
"step": 9565,
|
|
"valid_targets_mean": 1514.4,
|
|
"valid_targets_min": 967
|
|
},
|
|
{
|
|
"epoch": 5.013095861707701,
|
|
"grad_norm": 0.8584249807542884,
|
|
"learning_rate": 9.044927476738252e-06,
|
|
"loss": 0.1438,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.069888174533844,
|
|
"step": 9570,
|
|
"valid_targets_mean": 1500.2,
|
|
"valid_targets_min": 718
|
|
},
|
|
{
|
|
"epoch": 5.01571503404924,
|
|
"grad_norm": 0.7662334735645583,
|
|
"learning_rate": 9.02308100652927e-06,
|
|
"loss": 0.1384,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07856231182813644,
|
|
"step": 9575,
|
|
"valid_targets_mean": 1689.9,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 5.01833420639078,
|
|
"grad_norm": 0.7289531629346846,
|
|
"learning_rate": 9.001253263731753e-06,
|
|
"loss": 0.1269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06529080867767334,
|
|
"step": 9580,
|
|
"valid_targets_mean": 1543.9,
|
|
"valid_targets_min": 725
|
|
},
|
|
{
|
|
"epoch": 5.020953378732321,
|
|
"grad_norm": 0.8227298254980042,
|
|
"learning_rate": 8.979444285585392e-06,
|
|
"loss": 0.1447,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05642779916524887,
|
|
"step": 9585,
|
|
"valid_targets_mean": 1290.1,
|
|
"valid_targets_min": 610
|
|
},
|
|
{
|
|
"epoch": 5.023572551073861,
|
|
"grad_norm": 0.7629736165857393,
|
|
"learning_rate": 8.957654109297874e-06,
|
|
"loss": 0.1335,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06554756313562393,
|
|
"step": 9590,
|
|
"valid_targets_mean": 1588.6,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 5.0261917234154,
|
|
"grad_norm": 0.7158738761162554,
|
|
"learning_rate": 8.935882772044798e-06,
|
|
"loss": 0.137,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07234790176153183,
|
|
"step": 9595,
|
|
"valid_targets_mean": 1708.5,
|
|
"valid_targets_min": 865
|
|
},
|
|
{
|
|
"epoch": 5.028810895756941,
|
|
"grad_norm": 0.8754088785499504,
|
|
"learning_rate": 8.91413031096963e-06,
|
|
"loss": 0.1317,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08297131210565567,
|
|
"step": 9600,
|
|
"valid_targets_mean": 1502.1,
|
|
"valid_targets_min": 593
|
|
},
|
|
{
|
|
"epoch": 5.031430068098481,
|
|
"grad_norm": 0.7442074558873492,
|
|
"learning_rate": 8.892396763183637e-06,
|
|
"loss": 0.1314,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06886909157037735,
|
|
"step": 9605,
|
|
"valid_targets_mean": 1667.9,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 5.034049240440021,
|
|
"grad_norm": 0.7627126439991847,
|
|
"learning_rate": 8.870682165765798e-06,
|
|
"loss": 0.1256,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05173000693321228,
|
|
"step": 9610,
|
|
"valid_targets_mean": 1233.4,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 5.036668412781561,
|
|
"grad_norm": 0.7142867945568193,
|
|
"learning_rate": 8.848986555762777e-06,
|
|
"loss": 0.1239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056171685457229614,
|
|
"step": 9615,
|
|
"valid_targets_mean": 1556.0,
|
|
"valid_targets_min": 968
|
|
},
|
|
{
|
|
"epoch": 5.039287585123101,
|
|
"grad_norm": 0.7988401443613027,
|
|
"learning_rate": 8.827309970188847e-06,
|
|
"loss": 0.1249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07681363821029663,
|
|
"step": 9620,
|
|
"valid_targets_mean": 1773.1,
|
|
"valid_targets_min": 936
|
|
},
|
|
{
|
|
"epoch": 5.041906757464641,
|
|
"grad_norm": 0.8914213786316774,
|
|
"learning_rate": 8.805652446025815e-06,
|
|
"loss": 0.1343,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06082507595419884,
|
|
"step": 9625,
|
|
"valid_targets_mean": 1383.6,
|
|
"valid_targets_min": 728
|
|
},
|
|
{
|
|
"epoch": 5.0445259298061815,
|
|
"grad_norm": 0.7635151965573758,
|
|
"learning_rate": 8.784014020222968e-06,
|
|
"loss": 0.1333,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08338280022144318,
|
|
"step": 9630,
|
|
"valid_targets_mean": 2023.5,
|
|
"valid_targets_min": 1143
|
|
},
|
|
{
|
|
"epoch": 5.047145102147721,
|
|
"grad_norm": 0.8405113623184243,
|
|
"learning_rate": 8.762394729697022e-06,
|
|
"loss": 0.1339,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07841154932975769,
|
|
"step": 9635,
|
|
"valid_targets_mean": 1411.9,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 5.049764274489261,
|
|
"grad_norm": 0.8164628534220753,
|
|
"learning_rate": 8.740794611332038e-06,
|
|
"loss": 0.127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06135069951415062,
|
|
"step": 9640,
|
|
"valid_targets_mean": 1393.2,
|
|
"valid_targets_min": 625
|
|
},
|
|
{
|
|
"epoch": 5.0523834468308015,
|
|
"grad_norm": 0.7403910338312527,
|
|
"learning_rate": 8.719213701979348e-06,
|
|
"loss": 0.1291,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07470349967479706,
|
|
"step": 9645,
|
|
"valid_targets_mean": 1949.8,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 5.055002619172342,
|
|
"grad_norm": 0.8074833036677275,
|
|
"learning_rate": 8.697652038457563e-06,
|
|
"loss": 0.1344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09135995805263519,
|
|
"step": 9650,
|
|
"valid_targets_mean": 2193.8,
|
|
"valid_targets_min": 891
|
|
},
|
|
{
|
|
"epoch": 5.057621791513881,
|
|
"grad_norm": 0.7255751179833262,
|
|
"learning_rate": 8.676109657552396e-06,
|
|
"loss": 0.128,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05861027538776398,
|
|
"step": 9655,
|
|
"valid_targets_mean": 1569.0,
|
|
"valid_targets_min": 850
|
|
},
|
|
{
|
|
"epoch": 5.0602409638554215,
|
|
"grad_norm": 0.7615798411288203,
|
|
"learning_rate": 8.654586596016726e-06,
|
|
"loss": 0.1291,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05548068881034851,
|
|
"step": 9660,
|
|
"valid_targets_mean": 1384.2,
|
|
"valid_targets_min": 974
|
|
},
|
|
{
|
|
"epoch": 5.062860136196962,
|
|
"grad_norm": 0.8130706992942511,
|
|
"learning_rate": 8.633082890570417e-06,
|
|
"loss": 0.1224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06968683749437332,
|
|
"step": 9665,
|
|
"valid_targets_mean": 1502.9,
|
|
"valid_targets_min": 998
|
|
},
|
|
{
|
|
"epoch": 5.065479308538502,
|
|
"grad_norm": 0.7913334909686525,
|
|
"learning_rate": 8.611598577900343e-06,
|
|
"loss": 0.1271,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07109968364238739,
|
|
"step": 9670,
|
|
"valid_targets_mean": 1675.6,
|
|
"valid_targets_min": 1047
|
|
},
|
|
{
|
|
"epoch": 5.068098480880042,
|
|
"grad_norm": 0.8234985901502248,
|
|
"learning_rate": 8.590133694660278e-06,
|
|
"loss": 0.1278,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0685684010386467,
|
|
"step": 9675,
|
|
"valid_targets_mean": 1577.6,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 5.070717653221582,
|
|
"grad_norm": 0.8494760608549315,
|
|
"learning_rate": 8.568688277470858e-06,
|
|
"loss": 0.1329,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07787296175956726,
|
|
"step": 9680,
|
|
"valid_targets_mean": 1967.9,
|
|
"valid_targets_min": 1075
|
|
},
|
|
{
|
|
"epoch": 5.073336825563122,
|
|
"grad_norm": 0.8041528005781308,
|
|
"learning_rate": 8.547262362919504e-06,
|
|
"loss": 0.1351,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07973786443471909,
|
|
"step": 9685,
|
|
"valid_targets_mean": 1872.6,
|
|
"valid_targets_min": 1317
|
|
},
|
|
{
|
|
"epoch": 5.0759559979046625,
|
|
"grad_norm": 0.7778820031817429,
|
|
"learning_rate": 8.525855987560368e-06,
|
|
"loss": 0.1271,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07093140482902527,
|
|
"step": 9690,
|
|
"valid_targets_mean": 1697.6,
|
|
"valid_targets_min": 957
|
|
},
|
|
{
|
|
"epoch": 5.078575170246202,
|
|
"grad_norm": 0.792525720698113,
|
|
"learning_rate": 8.504469187914242e-06,
|
|
"loss": 0.1302,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057369161397218704,
|
|
"step": 9695,
|
|
"valid_targets_mean": 1323.0,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 5.081194342587742,
|
|
"grad_norm": 0.8404788767956344,
|
|
"learning_rate": 8.483102000468569e-06,
|
|
"loss": 0.1316,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055390436202287674,
|
|
"step": 9700,
|
|
"valid_targets_mean": 1488.8,
|
|
"valid_targets_min": 1111
|
|
},
|
|
{
|
|
"epoch": 5.0838135149292825,
|
|
"grad_norm": 0.7641603228234167,
|
|
"learning_rate": 8.461754461677275e-06,
|
|
"loss": 0.1252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06761826574802399,
|
|
"step": 9705,
|
|
"valid_targets_mean": 1649.2,
|
|
"valid_targets_min": 1147
|
|
},
|
|
{
|
|
"epoch": 5.086432687270823,
|
|
"grad_norm": 0.8392729696383808,
|
|
"learning_rate": 8.440426607960821e-06,
|
|
"loss": 0.1263,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07057835161685944,
|
|
"step": 9710,
|
|
"valid_targets_mean": 1526.2,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 5.089051859612362,
|
|
"grad_norm": 0.7783269470106086,
|
|
"learning_rate": 8.419118475706032e-06,
|
|
"loss": 0.1156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05206485465168953,
|
|
"step": 9715,
|
|
"valid_targets_mean": 1219.6,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 5.0916710319539025,
|
|
"grad_norm": 0.7982999694650752,
|
|
"learning_rate": 8.397830101266133e-06,
|
|
"loss": 0.1285,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061915069818496704,
|
|
"step": 9720,
|
|
"valid_targets_mean": 1574.5,
|
|
"valid_targets_min": 1192
|
|
},
|
|
{
|
|
"epoch": 5.094290204295443,
|
|
"grad_norm": 0.7793045637084354,
|
|
"learning_rate": 8.376561520960603e-06,
|
|
"loss": 0.123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06052004545927048,
|
|
"step": 9725,
|
|
"valid_targets_mean": 1496.1,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 5.096909376636983,
|
|
"grad_norm": 0.8395426162046974,
|
|
"learning_rate": 8.355312771075165e-06,
|
|
"loss": 0.1262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0652417466044426,
|
|
"step": 9730,
|
|
"valid_targets_mean": 1402.0,
|
|
"valid_targets_min": 657
|
|
},
|
|
{
|
|
"epoch": 5.0995285489785225,
|
|
"grad_norm": 0.9063254762504421,
|
|
"learning_rate": 8.33408388786172e-06,
|
|
"loss": 0.1258,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06874285638332367,
|
|
"step": 9735,
|
|
"valid_targets_mean": 1471.2,
|
|
"valid_targets_min": 880
|
|
},
|
|
{
|
|
"epoch": 5.102147721320063,
|
|
"grad_norm": 0.8635233820789475,
|
|
"learning_rate": 8.312874907538259e-06,
|
|
"loss": 0.1261,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07698352634906769,
|
|
"step": 9740,
|
|
"valid_targets_mean": 1567.6,
|
|
"valid_targets_min": 1163
|
|
},
|
|
{
|
|
"epoch": 5.104766893661603,
|
|
"grad_norm": 0.8507730376845948,
|
|
"learning_rate": 8.29168586628883e-06,
|
|
"loss": 0.1216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06012817099690437,
|
|
"step": 9745,
|
|
"valid_targets_mean": 1493.1,
|
|
"valid_targets_min": 998
|
|
},
|
|
{
|
|
"epoch": 5.107386066003143,
|
|
"grad_norm": 0.7628871482302233,
|
|
"learning_rate": 8.270516800263461e-06,
|
|
"loss": 0.1222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050013184547424316,
|
|
"step": 9750,
|
|
"valid_targets_mean": 1214.1,
|
|
"valid_targets_min": 703
|
|
},
|
|
{
|
|
"epoch": 5.110005238344683,
|
|
"grad_norm": 0.7848283673414309,
|
|
"learning_rate": 8.249367745578086e-06,
|
|
"loss": 0.1337,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0638660341501236,
|
|
"step": 9755,
|
|
"valid_targets_mean": 1459.2,
|
|
"valid_targets_min": 934
|
|
},
|
|
{
|
|
"epoch": 5.112624410686223,
|
|
"grad_norm": 0.8613497551716048,
|
|
"learning_rate": 8.228238738314514e-06,
|
|
"loss": 0.137,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09289310872554779,
|
|
"step": 9760,
|
|
"valid_targets_mean": 1754.4,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 5.115243583027763,
|
|
"grad_norm": 0.7958306842436589,
|
|
"learning_rate": 8.207129814520349e-06,
|
|
"loss": 0.12,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07653972506523132,
|
|
"step": 9765,
|
|
"valid_targets_mean": 1864.2,
|
|
"valid_targets_min": 1270
|
|
},
|
|
{
|
|
"epoch": 5.117862755369304,
|
|
"grad_norm": 0.7443615561171484,
|
|
"learning_rate": 8.186041010208932e-06,
|
|
"loss": 0.121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06417173147201538,
|
|
"step": 9770,
|
|
"valid_targets_mean": 1765.5,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 5.120481927710843,
|
|
"grad_norm": 0.8704389863666305,
|
|
"learning_rate": 8.164972361359274e-06,
|
|
"loss": 0.1268,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08009348809719086,
|
|
"step": 9775,
|
|
"valid_targets_mean": 1876.4,
|
|
"valid_targets_min": 1172
|
|
},
|
|
{
|
|
"epoch": 5.123101100052383,
|
|
"grad_norm": 0.8254417455907435,
|
|
"learning_rate": 8.143923903916e-06,
|
|
"loss": 0.1242,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06702611595392227,
|
|
"step": 9780,
|
|
"valid_targets_mean": 1642.0,
|
|
"valid_targets_min": 1033
|
|
},
|
|
{
|
|
"epoch": 5.125720272393924,
|
|
"grad_norm": 0.7889086458607909,
|
|
"learning_rate": 8.1228956737893e-06,
|
|
"loss": 0.1181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07090294361114502,
|
|
"step": 9785,
|
|
"valid_targets_mean": 1689.0,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 5.128339444735464,
|
|
"grad_norm": 0.8058169834033115,
|
|
"learning_rate": 8.10188770685482e-06,
|
|
"loss": 0.1148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05185168236494064,
|
|
"step": 9790,
|
|
"valid_targets_mean": 1164.2,
|
|
"valid_targets_min": 737
|
|
},
|
|
{
|
|
"epoch": 5.130958617077003,
|
|
"grad_norm": 0.9519237857968278,
|
|
"learning_rate": 8.080900038953686e-06,
|
|
"loss": 0.1306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06645987927913666,
|
|
"step": 9795,
|
|
"valid_targets_mean": 1427.9,
|
|
"valid_targets_min": 791
|
|
},
|
|
{
|
|
"epoch": 5.133577789418544,
|
|
"grad_norm": 0.7608506328312206,
|
|
"learning_rate": 8.05993270589234e-06,
|
|
"loss": 0.1234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05785384774208069,
|
|
"step": 9800,
|
|
"valid_targets_mean": 1559.5,
|
|
"valid_targets_min": 786
|
|
},
|
|
{
|
|
"epoch": 5.136196961760084,
|
|
"grad_norm": 0.7612550220763856,
|
|
"learning_rate": 8.038985743442582e-06,
|
|
"loss": 0.1225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05324278026819229,
|
|
"step": 9805,
|
|
"valid_targets_mean": 1411.5,
|
|
"valid_targets_min": 944
|
|
},
|
|
{
|
|
"epoch": 5.138816134101624,
|
|
"grad_norm": 0.7614465632357175,
|
|
"learning_rate": 8.018059187341414e-06,
|
|
"loss": 0.1207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055258601903915405,
|
|
"step": 9810,
|
|
"valid_targets_mean": 1454.9,
|
|
"valid_targets_min": 1227
|
|
},
|
|
{
|
|
"epoch": 5.141435306443164,
|
|
"grad_norm": 0.7433802949513004,
|
|
"learning_rate": 7.997153073291049e-06,
|
|
"loss": 0.1264,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05270681530237198,
|
|
"step": 9815,
|
|
"valid_targets_mean": 1446.9,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 5.144054478784704,
|
|
"grad_norm": 0.9612758343128028,
|
|
"learning_rate": 7.976267436958817e-06,
|
|
"loss": 0.1286,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08071742206811905,
|
|
"step": 9820,
|
|
"valid_targets_mean": 2122.9,
|
|
"valid_targets_min": 1227
|
|
},
|
|
{
|
|
"epoch": 5.146673651126244,
|
|
"grad_norm": 0.7703023931906229,
|
|
"learning_rate": 7.955402313977115e-06,
|
|
"loss": 0.1247,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05489915609359741,
|
|
"step": 9825,
|
|
"valid_targets_mean": 1256.0,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 5.149292823467785,
|
|
"grad_norm": 0.7565720696589413,
|
|
"learning_rate": 7.934557739943333e-06,
|
|
"loss": 0.1224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045493997633457184,
|
|
"step": 9830,
|
|
"valid_targets_mean": 1168.8,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 5.151911995809324,
|
|
"grad_norm": 0.8230795234990084,
|
|
"learning_rate": 7.91373375041982e-06,
|
|
"loss": 0.1168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0581836998462677,
|
|
"step": 9835,
|
|
"valid_targets_mean": 1275.9,
|
|
"valid_targets_min": 911
|
|
},
|
|
{
|
|
"epoch": 5.154531168150864,
|
|
"grad_norm": 0.7663342792339701,
|
|
"learning_rate": 7.89293038093379e-06,
|
|
"loss": 0.1231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05573280155658722,
|
|
"step": 9840,
|
|
"valid_targets_mean": 1431.9,
|
|
"valid_targets_min": 661
|
|
},
|
|
{
|
|
"epoch": 5.157150340492405,
|
|
"grad_norm": 0.8663535122859962,
|
|
"learning_rate": 7.87214766697729e-06,
|
|
"loss": 0.1291,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0651249885559082,
|
|
"step": 9845,
|
|
"valid_targets_mean": 1572.1,
|
|
"valid_targets_min": 1164
|
|
},
|
|
{
|
|
"epoch": 5.159769512833945,
|
|
"grad_norm": 0.8382575650021854,
|
|
"learning_rate": 7.851385644007103e-06,
|
|
"loss": 0.1218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055575910955667496,
|
|
"step": 9850,
|
|
"valid_targets_mean": 1396.6,
|
|
"valid_targets_min": 692
|
|
},
|
|
{
|
|
"epoch": 5.162388685175484,
|
|
"grad_norm": 0.7875059756241013,
|
|
"learning_rate": 7.830644347444761e-06,
|
|
"loss": 0.1255,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05687501281499863,
|
|
"step": 9855,
|
|
"valid_targets_mean": 1405.8,
|
|
"valid_targets_min": 923
|
|
},
|
|
{
|
|
"epoch": 5.165007857517025,
|
|
"grad_norm": 0.8593071246128563,
|
|
"learning_rate": 7.80992381267637e-06,
|
|
"loss": 0.1274,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06243991106748581,
|
|
"step": 9860,
|
|
"valid_targets_mean": 1375.1,
|
|
"valid_targets_min": 922
|
|
},
|
|
{
|
|
"epoch": 5.167627029858565,
|
|
"grad_norm": 0.7415385381072802,
|
|
"learning_rate": 7.78922407505268e-06,
|
|
"loss": 0.1221,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06369824707508087,
|
|
"step": 9865,
|
|
"valid_targets_mean": 1806.0,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 5.170246202200104,
|
|
"grad_norm": 0.7676587590809413,
|
|
"learning_rate": 7.768545169888909e-06,
|
|
"loss": 0.1263,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0626119002699852,
|
|
"step": 9870,
|
|
"valid_targets_mean": 1342.9,
|
|
"valid_targets_min": 836
|
|
},
|
|
{
|
|
"epoch": 5.172865374541645,
|
|
"grad_norm": 0.8280038560088043,
|
|
"learning_rate": 7.747887132464757e-06,
|
|
"loss": 0.1229,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07222384959459305,
|
|
"step": 9875,
|
|
"valid_targets_mean": 1504.5,
|
|
"valid_targets_min": 1013
|
|
},
|
|
{
|
|
"epoch": 5.175484546883185,
|
|
"grad_norm": 0.8480684316574301,
|
|
"learning_rate": 7.72724999802432e-06,
|
|
"loss": 0.1245,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05442634969949722,
|
|
"step": 9880,
|
|
"valid_targets_mean": 1351.6,
|
|
"valid_targets_min": 978
|
|
},
|
|
{
|
|
"epoch": 5.178103719224725,
|
|
"grad_norm": 0.8943098137120158,
|
|
"learning_rate": 7.706633801776031e-06,
|
|
"loss": 0.1231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07496099174022675,
|
|
"step": 9885,
|
|
"valid_targets_mean": 1812.5,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 5.180722891566265,
|
|
"grad_norm": 0.7710715592164034,
|
|
"learning_rate": 7.686038578892599e-06,
|
|
"loss": 0.1327,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05917992815375328,
|
|
"step": 9890,
|
|
"valid_targets_mean": 1657.9,
|
|
"valid_targets_min": 1219
|
|
},
|
|
{
|
|
"epoch": 5.183342063907805,
|
|
"grad_norm": 0.768425230191172,
|
|
"learning_rate": 7.665464364510954e-06,
|
|
"loss": 0.1253,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07286841422319412,
|
|
"step": 9895,
|
|
"valid_targets_mean": 1814.4,
|
|
"valid_targets_min": 1002
|
|
},
|
|
{
|
|
"epoch": 5.185961236249345,
|
|
"grad_norm": 0.783120110067545,
|
|
"learning_rate": 7.644911193732185e-06,
|
|
"loss": 0.1225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05758904665708542,
|
|
"step": 9900,
|
|
"valid_targets_mean": 1390.4,
|
|
"valid_targets_min": 646
|
|
},
|
|
{
|
|
"epoch": 5.1885804085908855,
|
|
"grad_norm": 0.8756284618345157,
|
|
"learning_rate": 7.624379101621464e-06,
|
|
"loss": 0.1236,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06255842745304108,
|
|
"step": 9905,
|
|
"valid_targets_mean": 1307.4,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 5.191199580932425,
|
|
"grad_norm": 0.7865895847705132,
|
|
"learning_rate": 7.603868123208023e-06,
|
|
"loss": 0.1168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05866144597530365,
|
|
"step": 9910,
|
|
"valid_targets_mean": 1574.8,
|
|
"valid_targets_min": 808
|
|
},
|
|
{
|
|
"epoch": 5.193818753273965,
|
|
"grad_norm": 0.8621724591786725,
|
|
"learning_rate": 7.583378293485059e-06,
|
|
"loss": 0.1281,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05954202264547348,
|
|
"step": 9915,
|
|
"valid_targets_mean": 1173.9,
|
|
"valid_targets_min": 913
|
|
},
|
|
{
|
|
"epoch": 5.1964379256155055,
|
|
"grad_norm": 0.8240980648230771,
|
|
"learning_rate": 7.5629096474096954e-06,
|
|
"loss": 0.1168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05828775838017464,
|
|
"step": 9920,
|
|
"valid_targets_mean": 1354.2,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 5.199057097957046,
|
|
"grad_norm": 0.8571886369802686,
|
|
"learning_rate": 7.542462219902909e-06,
|
|
"loss": 0.115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055858269333839417,
|
|
"step": 9925,
|
|
"valid_targets_mean": 1508.4,
|
|
"valid_targets_min": 1013
|
|
},
|
|
{
|
|
"epoch": 5.201676270298585,
|
|
"grad_norm": 0.7750722608631837,
|
|
"learning_rate": 7.5220360458494855e-06,
|
|
"loss": 0.1188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06182383745908737,
|
|
"step": 9930,
|
|
"valid_targets_mean": 1517.8,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 5.204295442640126,
|
|
"grad_norm": 0.8070247648475251,
|
|
"learning_rate": 7.5016311600979245e-06,
|
|
"loss": 0.1274,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05417487397789955,
|
|
"step": 9935,
|
|
"valid_targets_mean": 1243.6,
|
|
"valid_targets_min": 548
|
|
},
|
|
{
|
|
"epoch": 5.206914614981666,
|
|
"grad_norm": 0.7338574303878064,
|
|
"learning_rate": 7.481247597460448e-06,
|
|
"loss": 0.1183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05301428586244583,
|
|
"step": 9940,
|
|
"valid_targets_mean": 1269.0,
|
|
"valid_targets_min": 694
|
|
},
|
|
{
|
|
"epoch": 5.209533787323206,
|
|
"grad_norm": 0.7525189322704464,
|
|
"learning_rate": 7.460885392712856e-06,
|
|
"loss": 0.1257,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07202500849962234,
|
|
"step": 9945,
|
|
"valid_targets_mean": 2141.0,
|
|
"valid_targets_min": 1030
|
|
},
|
|
{
|
|
"epoch": 5.212152959664746,
|
|
"grad_norm": 0.6971483181611564,
|
|
"learning_rate": 7.440544580594553e-06,
|
|
"loss": 0.1135,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05420509725809097,
|
|
"step": 9950,
|
|
"valid_targets_mean": 1738.6,
|
|
"valid_targets_min": 682
|
|
},
|
|
{
|
|
"epoch": 5.214772132006286,
|
|
"grad_norm": 0.9016247140440351,
|
|
"learning_rate": 7.420225195808397e-06,
|
|
"loss": 0.1153,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06308970600366592,
|
|
"step": 9955,
|
|
"valid_targets_mean": 1407.2,
|
|
"valid_targets_min": 990
|
|
},
|
|
{
|
|
"epoch": 5.217391304347826,
|
|
"grad_norm": 0.7498791828385352,
|
|
"learning_rate": 7.399927273020748e-06,
|
|
"loss": 0.1201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05194121599197388,
|
|
"step": 9960,
|
|
"valid_targets_mean": 1551.6,
|
|
"valid_targets_min": 639
|
|
},
|
|
{
|
|
"epoch": 5.2200104766893665,
|
|
"grad_norm": 0.8056681058630909,
|
|
"learning_rate": 7.379650846861295e-06,
|
|
"loss": 0.119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05170396342873573,
|
|
"step": 9965,
|
|
"valid_targets_mean": 1442.6,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 5.222629649030906,
|
|
"grad_norm": 0.8173097296836717,
|
|
"learning_rate": 7.359395951923081e-06,
|
|
"loss": 0.1211,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06028531491756439,
|
|
"step": 9970,
|
|
"valid_targets_mean": 1426.5,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 5.225248821372446,
|
|
"grad_norm": 0.7462986566841966,
|
|
"learning_rate": 7.3391626227624145e-06,
|
|
"loss": 0.124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056972358375787735,
|
|
"step": 9975,
|
|
"valid_targets_mean": 1298.2,
|
|
"valid_targets_min": 557
|
|
},
|
|
{
|
|
"epoch": 5.2278679937139865,
|
|
"grad_norm": 0.8632265642665938,
|
|
"learning_rate": 7.318950893898804e-06,
|
|
"loss": 0.1218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07349438965320587,
|
|
"step": 9980,
|
|
"valid_targets_mean": 1472.2,
|
|
"valid_targets_min": 713
|
|
},
|
|
{
|
|
"epoch": 5.230487166055527,
|
|
"grad_norm": 0.8247613670529226,
|
|
"learning_rate": 7.29876079981491e-06,
|
|
"loss": 0.124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0822034552693367,
|
|
"step": 9985,
|
|
"valid_targets_mean": 3139.9,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 5.233106338397066,
|
|
"grad_norm": 0.7016447370775042,
|
|
"learning_rate": 7.2785923749564905e-06,
|
|
"loss": 0.172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08335396647453308,
|
|
"step": 9990,
|
|
"valid_targets_mean": 7612.6,
|
|
"valid_targets_min": 5111
|
|
},
|
|
{
|
|
"epoch": 5.2357255107386065,
|
|
"grad_norm": 0.5823301815088959,
|
|
"learning_rate": 7.2584456537323025e-06,
|
|
"loss": 0.1609,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08028516173362732,
|
|
"step": 9995,
|
|
"valid_targets_mean": 6669.0,
|
|
"valid_targets_min": 4480
|
|
},
|
|
{
|
|
"epoch": 5.238344683080147,
|
|
"grad_norm": 0.42686852024385435,
|
|
"learning_rate": 7.238320670514125e-06,
|
|
"loss": 0.1499,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07357516139745712,
|
|
"step": 10000,
|
|
"valid_targets_mean": 6538.6,
|
|
"valid_targets_min": 4620
|
|
},
|
|
{
|
|
"epoch": 5.240963855421687,
|
|
"grad_norm": 0.3860267732299242,
|
|
"learning_rate": 7.218217459636605e-06,
|
|
"loss": 0.1494,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08040302991867065,
|
|
"step": 10005,
|
|
"valid_targets_mean": 6802.6,
|
|
"valid_targets_min": 4873
|
|
},
|
|
{
|
|
"epoch": 5.2435830277632265,
|
|
"grad_norm": 0.41018469570895844,
|
|
"learning_rate": 7.198136055397269e-06,
|
|
"loss": 0.1619,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09762245416641235,
|
|
"step": 10010,
|
|
"valid_targets_mean": 7563.2,
|
|
"valid_targets_min": 4229
|
|
},
|
|
{
|
|
"epoch": 5.246202200104767,
|
|
"grad_norm": 0.44350311791841235,
|
|
"learning_rate": 7.178076492056434e-06,
|
|
"loss": 0.1576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08566471934318542,
|
|
"step": 10015,
|
|
"valid_targets_mean": 6109.0,
|
|
"valid_targets_min": 4279
|
|
},
|
|
{
|
|
"epoch": 5.248821372446307,
|
|
"grad_norm": 0.3751231386489058,
|
|
"learning_rate": 7.1580388038371525e-06,
|
|
"loss": 0.1474,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05747370794415474,
|
|
"step": 10020,
|
|
"valid_targets_mean": 5549.6,
|
|
"valid_targets_min": 4542
|
|
},
|
|
{
|
|
"epoch": 5.251440544787847,
|
|
"grad_norm": 0.5169337392227329,
|
|
"learning_rate": 7.1380230249251605e-06,
|
|
"loss": 0.1668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07669960707426071,
|
|
"step": 10025,
|
|
"valid_targets_mean": 2525.8,
|
|
"valid_targets_min": 1497
|
|
},
|
|
{
|
|
"epoch": 5.254059717129387,
|
|
"grad_norm": 0.37923721697545987,
|
|
"learning_rate": 7.118029189468812e-06,
|
|
"loss": 0.152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07565468549728394,
|
|
"step": 10030,
|
|
"valid_targets_mean": 6166.2,
|
|
"valid_targets_min": 5332
|
|
},
|
|
{
|
|
"epoch": 5.256678889470927,
|
|
"grad_norm": 0.40425635783083314,
|
|
"learning_rate": 7.098057331579031e-06,
|
|
"loss": 0.1512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08241156488656998,
|
|
"step": 10035,
|
|
"valid_targets_mean": 5823.5,
|
|
"valid_targets_min": 4612
|
|
},
|
|
{
|
|
"epoch": 5.259298061812467,
|
|
"grad_norm": 0.3907609589866494,
|
|
"learning_rate": 7.078107485329222e-06,
|
|
"loss": 0.1592,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0800551027059555,
|
|
"step": 10040,
|
|
"valid_targets_mean": 6896.5,
|
|
"valid_targets_min": 4491
|
|
},
|
|
{
|
|
"epoch": 5.261917234154008,
|
|
"grad_norm": 0.3778040111634576,
|
|
"learning_rate": 7.05817968475528e-06,
|
|
"loss": 0.1486,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07752105593681335,
|
|
"step": 10045,
|
|
"valid_targets_mean": 6611.1,
|
|
"valid_targets_min": 4754
|
|
},
|
|
{
|
|
"epoch": 5.264536406495547,
|
|
"grad_norm": 0.3812696012808861,
|
|
"learning_rate": 7.038273963855442e-06,
|
|
"loss": 0.145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06828771531581879,
|
|
"step": 10050,
|
|
"valid_targets_mean": 5907.1,
|
|
"valid_targets_min": 4787
|
|
},
|
|
{
|
|
"epoch": 5.267155578837087,
|
|
"grad_norm": 0.38896075575746797,
|
|
"learning_rate": 7.018390356590303e-06,
|
|
"loss": 0.1371,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06619839370250702,
|
|
"step": 10055,
|
|
"valid_targets_mean": 6055.5,
|
|
"valid_targets_min": 5052
|
|
},
|
|
{
|
|
"epoch": 5.269774751178628,
|
|
"grad_norm": 0.39253571485970484,
|
|
"learning_rate": 6.998528896882726e-06,
|
|
"loss": 0.1397,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06778645515441895,
|
|
"step": 10060,
|
|
"valid_targets_mean": 6365.0,
|
|
"valid_targets_min": 4664
|
|
},
|
|
{
|
|
"epoch": 5.272393923520168,
|
|
"grad_norm": 0.3861560859423902,
|
|
"learning_rate": 6.978689618617784e-06,
|
|
"loss": 0.1504,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06602711230516434,
|
|
"step": 10065,
|
|
"valid_targets_mean": 5928.2,
|
|
"valid_targets_min": 4211
|
|
},
|
|
{
|
|
"epoch": 5.275013095861707,
|
|
"grad_norm": 0.3713955201784082,
|
|
"learning_rate": 6.958872555642711e-06,
|
|
"loss": 0.1486,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07369357347488403,
|
|
"step": 10070,
|
|
"valid_targets_mean": 6415.9,
|
|
"valid_targets_min": 4582
|
|
},
|
|
{
|
|
"epoch": 5.277632268203248,
|
|
"grad_norm": 0.38669814277909503,
|
|
"learning_rate": 6.939077741766842e-06,
|
|
"loss": 0.1348,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06115833669900894,
|
|
"step": 10075,
|
|
"valid_targets_mean": 6044.8,
|
|
"valid_targets_min": 4800
|
|
},
|
|
{
|
|
"epoch": 5.280251440544788,
|
|
"grad_norm": 0.3441598004906199,
|
|
"learning_rate": 6.9193052107615335e-06,
|
|
"loss": 0.1341,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06817680597305298,
|
|
"step": 10080,
|
|
"valid_targets_mean": 8009.0,
|
|
"valid_targets_min": 6651
|
|
},
|
|
{
|
|
"epoch": 5.282870612886328,
|
|
"grad_norm": 0.4086751254195686,
|
|
"learning_rate": 6.8995549963601675e-06,
|
|
"loss": 0.1525,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0852825790643692,
|
|
"step": 10085,
|
|
"valid_targets_mean": 5887.5,
|
|
"valid_targets_min": 3477
|
|
},
|
|
{
|
|
"epoch": 5.285489785227868,
|
|
"grad_norm": 0.4855072130935138,
|
|
"learning_rate": 6.8798271322580015e-06,
|
|
"loss": 0.1676,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09230928122997284,
|
|
"step": 10090,
|
|
"valid_targets_mean": 5471.1,
|
|
"valid_targets_min": 965
|
|
},
|
|
{
|
|
"epoch": 5.288108957569408,
|
|
"grad_norm": 0.3672577879646038,
|
|
"learning_rate": 6.860121652112211e-06,
|
|
"loss": 0.1377,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07655905187129974,
|
|
"step": 10095,
|
|
"valid_targets_mean": 6654.1,
|
|
"valid_targets_min": 5226
|
|
},
|
|
{
|
|
"epoch": 5.290728129910948,
|
|
"grad_norm": 0.4055655338645675,
|
|
"learning_rate": 6.840438589541733e-06,
|
|
"loss": 0.1474,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06398853659629822,
|
|
"step": 10100,
|
|
"valid_targets_mean": 6363.4,
|
|
"valid_targets_min": 4213
|
|
},
|
|
{
|
|
"epoch": 5.293347302252489,
|
|
"grad_norm": 0.42587886910638884,
|
|
"learning_rate": 6.820777978127315e-06,
|
|
"loss": 0.1606,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08191876113414764,
|
|
"step": 10105,
|
|
"valid_targets_mean": 5469.9,
|
|
"valid_targets_min": 3761
|
|
},
|
|
{
|
|
"epoch": 5.295966474594028,
|
|
"grad_norm": 0.4134904846413301,
|
|
"learning_rate": 6.801139851411347e-06,
|
|
"loss": 0.1778,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10157156735658646,
|
|
"step": 10110,
|
|
"valid_targets_mean": 7206.2,
|
|
"valid_targets_min": 4993
|
|
},
|
|
{
|
|
"epoch": 5.298585646935568,
|
|
"grad_norm": 0.3917239886283864,
|
|
"learning_rate": 6.78152424289789e-06,
|
|
"loss": 0.173,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07931825518608093,
|
|
"step": 10115,
|
|
"valid_targets_mean": 6402.2,
|
|
"valid_targets_min": 5199
|
|
},
|
|
{
|
|
"epoch": 5.301204819277109,
|
|
"grad_norm": 0.3825102469209971,
|
|
"learning_rate": 6.761931186052579e-06,
|
|
"loss": 0.1508,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08583804219961166,
|
|
"step": 10120,
|
|
"valid_targets_mean": 7578.1,
|
|
"valid_targets_min": 4534
|
|
},
|
|
{
|
|
"epoch": 5.303823991618648,
|
|
"grad_norm": 0.4109790035922538,
|
|
"learning_rate": 6.742360714302576e-06,
|
|
"loss": 0.1511,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06634113192558289,
|
|
"step": 10125,
|
|
"valid_targets_mean": 6264.4,
|
|
"valid_targets_min": 4751
|
|
},
|
|
{
|
|
"epoch": 5.306443163960188,
|
|
"grad_norm": 0.4112262337186078,
|
|
"learning_rate": 6.7228128610365116e-06,
|
|
"loss": 0.152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06814072281122208,
|
|
"step": 10130,
|
|
"valid_targets_mean": 5604.1,
|
|
"valid_targets_min": 3888
|
|
},
|
|
{
|
|
"epoch": 5.309062336301729,
|
|
"grad_norm": 0.4031731450213549,
|
|
"learning_rate": 6.703287659604429e-06,
|
|
"loss": 0.1576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08444470167160034,
|
|
"step": 10135,
|
|
"valid_targets_mean": 7001.6,
|
|
"valid_targets_min": 4455
|
|
},
|
|
{
|
|
"epoch": 5.311681508643269,
|
|
"grad_norm": 0.39609044928145004,
|
|
"learning_rate": 6.683785143317707e-06,
|
|
"loss": 0.15,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08760792016983032,
|
|
"step": 10140,
|
|
"valid_targets_mean": 6955.5,
|
|
"valid_targets_min": 5230
|
|
},
|
|
{
|
|
"epoch": 5.314300680984809,
|
|
"grad_norm": 0.3808870932482914,
|
|
"learning_rate": 6.6643053454490605e-06,
|
|
"loss": 0.1577,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06994412839412689,
|
|
"step": 10145,
|
|
"valid_targets_mean": 7295.8,
|
|
"valid_targets_min": 5668
|
|
},
|
|
{
|
|
"epoch": 5.316919853326349,
|
|
"grad_norm": 0.413429926706307,
|
|
"learning_rate": 6.644848299232407e-06,
|
|
"loss": 0.1551,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0801548883318901,
|
|
"step": 10150,
|
|
"valid_targets_mean": 6350.4,
|
|
"valid_targets_min": 4871
|
|
},
|
|
{
|
|
"epoch": 5.319539025667889,
|
|
"grad_norm": 0.37475407501719343,
|
|
"learning_rate": 6.625414037862865e-06,
|
|
"loss": 0.1449,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08898886293172836,
|
|
"step": 10155,
|
|
"valid_targets_mean": 8015.2,
|
|
"valid_targets_min": 6018
|
|
},
|
|
{
|
|
"epoch": 5.322158198009429,
|
|
"grad_norm": 0.35684775592882495,
|
|
"learning_rate": 6.606002594496683e-06,
|
|
"loss": 0.1313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06981298327445984,
|
|
"step": 10160,
|
|
"valid_targets_mean": 7110.1,
|
|
"valid_targets_min": 3533
|
|
},
|
|
{
|
|
"epoch": 5.324777370350969,
|
|
"grad_norm": 0.4425639362014185,
|
|
"learning_rate": 6.586614002251173e-06,
|
|
"loss": 0.1573,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08674512058496475,
|
|
"step": 10165,
|
|
"valid_targets_mean": 6875.0,
|
|
"valid_targets_min": 5199
|
|
},
|
|
{
|
|
"epoch": 5.327396542692509,
|
|
"grad_norm": 0.42824677791656335,
|
|
"learning_rate": 6.567248294204669e-06,
|
|
"loss": 0.1549,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07547745108604431,
|
|
"step": 10170,
|
|
"valid_targets_mean": 5568.4,
|
|
"valid_targets_min": 4923
|
|
},
|
|
{
|
|
"epoch": 5.330015715034049,
|
|
"grad_norm": 0.36854650462535177,
|
|
"learning_rate": 6.547905503396454e-06,
|
|
"loss": 0.151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059491705149412155,
|
|
"step": 10175,
|
|
"valid_targets_mean": 6421.9,
|
|
"valid_targets_min": 4527
|
|
},
|
|
{
|
|
"epoch": 5.3326348873755895,
|
|
"grad_norm": 0.39351311799382077,
|
|
"learning_rate": 6.528585662826725e-06,
|
|
"loss": 0.1485,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09477847069501877,
|
|
"step": 10180,
|
|
"valid_targets_mean": 7147.9,
|
|
"valid_targets_min": 3082
|
|
},
|
|
{
|
|
"epoch": 5.33525405971713,
|
|
"grad_norm": 0.38167891224919565,
|
|
"learning_rate": 6.5092888054564995e-06,
|
|
"loss": 0.1369,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06909792125225067,
|
|
"step": 10185,
|
|
"valid_targets_mean": 6880.6,
|
|
"valid_targets_min": 4685
|
|
},
|
|
{
|
|
"epoch": 5.337873232058669,
|
|
"grad_norm": 0.40048444743253364,
|
|
"learning_rate": 6.490014964207622e-06,
|
|
"loss": 0.1509,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06907521188259125,
|
|
"step": 10190,
|
|
"valid_targets_mean": 5915.5,
|
|
"valid_targets_min": 4875
|
|
},
|
|
{
|
|
"epoch": 5.34049240440021,
|
|
"grad_norm": 0.3729187881083819,
|
|
"learning_rate": 6.470764171962633e-06,
|
|
"loss": 0.1539,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06739333271980286,
|
|
"step": 10195,
|
|
"valid_targets_mean": 6234.5,
|
|
"valid_targets_min": 4433
|
|
},
|
|
{
|
|
"epoch": 5.34311157674175,
|
|
"grad_norm": 0.3943046453666481,
|
|
"learning_rate": 6.451536461564774e-06,
|
|
"loss": 0.144,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06472145020961761,
|
|
"step": 10200,
|
|
"valid_targets_mean": 5684.0,
|
|
"valid_targets_min": 4540
|
|
},
|
|
{
|
|
"epoch": 5.345730749083289,
|
|
"grad_norm": 0.43902753507712433,
|
|
"learning_rate": 6.432331865817891e-06,
|
|
"loss": 0.1557,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09635139256715775,
|
|
"step": 10205,
|
|
"valid_targets_mean": 7353.4,
|
|
"valid_targets_min": 5101
|
|
},
|
|
{
|
|
"epoch": 5.34834992142483,
|
|
"grad_norm": 0.4040924994306939,
|
|
"learning_rate": 6.413150417486409e-06,
|
|
"loss": 0.147,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06020814925432205,
|
|
"step": 10210,
|
|
"valid_targets_mean": 5300.5,
|
|
"valid_targets_min": 4442
|
|
},
|
|
{
|
|
"epoch": 5.35096909376637,
|
|
"grad_norm": 0.3703374972314483,
|
|
"learning_rate": 6.3939921492952514e-06,
|
|
"loss": 0.1273,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06047903001308441,
|
|
"step": 10215,
|
|
"valid_targets_mean": 6084.1,
|
|
"valid_targets_min": 5268
|
|
},
|
|
{
|
|
"epoch": 5.35358826610791,
|
|
"grad_norm": 0.3636680625578178,
|
|
"learning_rate": 6.374857093929802e-06,
|
|
"loss": 0.1303,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07073706388473511,
|
|
"step": 10220,
|
|
"valid_targets_mean": 6446.6,
|
|
"valid_targets_min": 5074
|
|
},
|
|
{
|
|
"epoch": 5.35620743844945,
|
|
"grad_norm": 0.3788299313306084,
|
|
"learning_rate": 6.355745284035835e-06,
|
|
"loss": 0.1361,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06560464203357697,
|
|
"step": 10225,
|
|
"valid_targets_mean": 6195.0,
|
|
"valid_targets_min": 5221
|
|
},
|
|
{
|
|
"epoch": 5.35882661079099,
|
|
"grad_norm": 0.4113542279653514,
|
|
"learning_rate": 6.336656752219479e-06,
|
|
"loss": 0.1458,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07294383645057678,
|
|
"step": 10230,
|
|
"valid_targets_mean": 5689.1,
|
|
"valid_targets_min": 4642
|
|
},
|
|
{
|
|
"epoch": 5.36144578313253,
|
|
"grad_norm": 0.47569693807252555,
|
|
"learning_rate": 6.3175915310471155e-06,
|
|
"loss": 0.1676,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08518552035093307,
|
|
"step": 10235,
|
|
"valid_targets_mean": 4461.4,
|
|
"valid_targets_min": 3222
|
|
},
|
|
{
|
|
"epoch": 5.3640649554740705,
|
|
"grad_norm": 0.4401813564141267,
|
|
"learning_rate": 6.2985496530454115e-06,
|
|
"loss": 0.1766,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07736337929964066,
|
|
"step": 10240,
|
|
"valid_targets_mean": 6822.2,
|
|
"valid_targets_min": 4371
|
|
},
|
|
{
|
|
"epoch": 5.36668412781561,
|
|
"grad_norm": 0.4407114697967697,
|
|
"learning_rate": 6.279531150701146e-06,
|
|
"loss": 0.1422,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07739515602588654,
|
|
"step": 10245,
|
|
"valid_targets_mean": 5632.6,
|
|
"valid_targets_min": 3689
|
|
},
|
|
{
|
|
"epoch": 5.36930330015715,
|
|
"grad_norm": 0.439335184296538,
|
|
"learning_rate": 6.260536056461275e-06,
|
|
"loss": 0.1502,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07690007984638214,
|
|
"step": 10250,
|
|
"valid_targets_mean": 6364.1,
|
|
"valid_targets_min": 4281
|
|
},
|
|
{
|
|
"epoch": 5.3719224724986905,
|
|
"grad_norm": 0.4600670189533514,
|
|
"learning_rate": 6.241564402732776e-06,
|
|
"loss": 0.1517,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09007485955953598,
|
|
"step": 10255,
|
|
"valid_targets_mean": 6337.8,
|
|
"valid_targets_min": 4844
|
|
},
|
|
{
|
|
"epoch": 5.374541644840231,
|
|
"grad_norm": 0.3955954710935925,
|
|
"learning_rate": 6.222616221882658e-06,
|
|
"loss": 0.163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06693617254495621,
|
|
"step": 10260,
|
|
"valid_targets_mean": 5507.5,
|
|
"valid_targets_min": 4549
|
|
},
|
|
{
|
|
"epoch": 5.37716081718177,
|
|
"grad_norm": 0.39154404681959404,
|
|
"learning_rate": 6.203691546237876e-06,
|
|
"loss": 0.1512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06701533496379852,
|
|
"step": 10265,
|
|
"valid_targets_mean": 5274.8,
|
|
"valid_targets_min": 2461
|
|
},
|
|
{
|
|
"epoch": 5.3797799895233105,
|
|
"grad_norm": 0.405735916954033,
|
|
"learning_rate": 6.184790408085281e-06,
|
|
"loss": 0.153,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0845988541841507,
|
|
"step": 10270,
|
|
"valid_targets_mean": 6465.2,
|
|
"valid_targets_min": 2815
|
|
},
|
|
{
|
|
"epoch": 5.382399161864851,
|
|
"grad_norm": 0.37252624761638364,
|
|
"learning_rate": 6.165912839671577e-06,
|
|
"loss": 0.1502,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06499049812555313,
|
|
"step": 10275,
|
|
"valid_targets_mean": 6715.5,
|
|
"valid_targets_min": 5192
|
|
},
|
|
{
|
|
"epoch": 5.385018334206391,
|
|
"grad_norm": 0.4138794283918792,
|
|
"learning_rate": 6.147058873203247e-06,
|
|
"loss": 0.1362,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06541548669338226,
|
|
"step": 10280,
|
|
"valid_targets_mean": 5737.2,
|
|
"valid_targets_min": 5186
|
|
},
|
|
{
|
|
"epoch": 5.3876375065479305,
|
|
"grad_norm": 0.4320162655413788,
|
|
"learning_rate": 6.12822854084651e-06,
|
|
"loss": 0.1482,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05883662402629852,
|
|
"step": 10285,
|
|
"valid_targets_mean": 5329.5,
|
|
"valid_targets_min": 3786
|
|
},
|
|
{
|
|
"epoch": 5.390256678889471,
|
|
"grad_norm": 0.3993761360156539,
|
|
"learning_rate": 6.1094218747272725e-06,
|
|
"loss": 0.193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08521421253681183,
|
|
"step": 10290,
|
|
"valid_targets_mean": 6233.8,
|
|
"valid_targets_min": 5147
|
|
},
|
|
{
|
|
"epoch": 5.392875851231011,
|
|
"grad_norm": 0.39569544811437657,
|
|
"learning_rate": 6.0906389069310434e-06,
|
|
"loss": 0.141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08180350810289383,
|
|
"step": 10295,
|
|
"valid_targets_mean": 5807.6,
|
|
"valid_targets_min": 3952
|
|
},
|
|
{
|
|
"epoch": 5.395495023572551,
|
|
"grad_norm": 0.4311559510427168,
|
|
"learning_rate": 6.07187966950292e-06,
|
|
"loss": 0.1522,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08369584381580353,
|
|
"step": 10300,
|
|
"valid_targets_mean": 6452.8,
|
|
"valid_targets_min": 5079
|
|
},
|
|
{
|
|
"epoch": 5.398114195914091,
|
|
"grad_norm": 0.4342446058772487,
|
|
"learning_rate": 6.053144194447507e-06,
|
|
"loss": 0.1546,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08193376660346985,
|
|
"step": 10305,
|
|
"valid_targets_mean": 6432.9,
|
|
"valid_targets_min": 3762
|
|
},
|
|
{
|
|
"epoch": 5.400733368255631,
|
|
"grad_norm": 0.4094968353215779,
|
|
"learning_rate": 6.0344325137288675e-06,
|
|
"loss": 0.1353,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06287078559398651,
|
|
"step": 10310,
|
|
"valid_targets_mean": 5954.1,
|
|
"valid_targets_min": 4726
|
|
},
|
|
{
|
|
"epoch": 5.403352540597171,
|
|
"grad_norm": 0.4694056913890443,
|
|
"learning_rate": 6.015744659270475e-06,
|
|
"loss": 0.1419,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08099617063999176,
|
|
"step": 10315,
|
|
"valid_targets_mean": 5874.6,
|
|
"valid_targets_min": 3724
|
|
},
|
|
{
|
|
"epoch": 5.405971712938712,
|
|
"grad_norm": 0.3680358302163034,
|
|
"learning_rate": 5.997080662955148e-06,
|
|
"loss": 0.1311,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06874550879001617,
|
|
"step": 10320,
|
|
"valid_targets_mean": 6436.9,
|
|
"valid_targets_min": 5578
|
|
},
|
|
{
|
|
"epoch": 5.408590885280251,
|
|
"grad_norm": 0.40563622398332005,
|
|
"learning_rate": 5.978440556625009e-06,
|
|
"loss": 0.1483,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0687650516629219,
|
|
"step": 10325,
|
|
"valid_targets_mean": 6727.8,
|
|
"valid_targets_min": 5795
|
|
},
|
|
{
|
|
"epoch": 5.411210057621791,
|
|
"grad_norm": 0.38579147373260586,
|
|
"learning_rate": 5.9598243720814e-06,
|
|
"loss": 0.1406,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06289946287870407,
|
|
"step": 10330,
|
|
"valid_targets_mean": 6116.2,
|
|
"valid_targets_min": 5268
|
|
},
|
|
{
|
|
"epoch": 5.413829229963332,
|
|
"grad_norm": 0.36715823726572966,
|
|
"learning_rate": 5.941232141084892e-06,
|
|
"loss": 0.1424,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06658988445997238,
|
|
"step": 10335,
|
|
"valid_targets_mean": 7379.5,
|
|
"valid_targets_min": 4920
|
|
},
|
|
{
|
|
"epoch": 5.416448402304872,
|
|
"grad_norm": 0.3570910803879039,
|
|
"learning_rate": 5.922663895355138e-06,
|
|
"loss": 0.1486,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055937521159648895,
|
|
"step": 10340,
|
|
"valid_targets_mean": 7048.4,
|
|
"valid_targets_min": 5243
|
|
},
|
|
{
|
|
"epoch": 5.419067574646411,
|
|
"grad_norm": 0.40882981172440686,
|
|
"learning_rate": 5.904119666570929e-06,
|
|
"loss": 0.1341,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07782215625047684,
|
|
"step": 10345,
|
|
"valid_targets_mean": 6385.8,
|
|
"valid_targets_min": 4794
|
|
},
|
|
{
|
|
"epoch": 5.421686746987952,
|
|
"grad_norm": 0.4339250956620702,
|
|
"learning_rate": 5.885599486370024e-06,
|
|
"loss": 0.1522,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06769482791423798,
|
|
"step": 10350,
|
|
"valid_targets_mean": 6215.4,
|
|
"valid_targets_min": 5168
|
|
},
|
|
{
|
|
"epoch": 5.424305919329492,
|
|
"grad_norm": 0.43044761296828704,
|
|
"learning_rate": 5.867103386349189e-06,
|
|
"loss": 0.1549,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08711810410022736,
|
|
"step": 10355,
|
|
"valid_targets_mean": 5333.2,
|
|
"valid_targets_min": 3130
|
|
},
|
|
{
|
|
"epoch": 5.426925091671032,
|
|
"grad_norm": 0.5643488195304203,
|
|
"learning_rate": 5.8486313980640975e-06,
|
|
"loss": 0.1862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09607409685850143,
|
|
"step": 10360,
|
|
"valid_targets_mean": 5546.2,
|
|
"valid_targets_min": 1971
|
|
},
|
|
{
|
|
"epoch": 5.429544264012572,
|
|
"grad_norm": 0.5850889872445502,
|
|
"learning_rate": 5.830183553029285e-06,
|
|
"loss": 0.2135,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13026058673858643,
|
|
"step": 10365,
|
|
"valid_targets_mean": 5648.9,
|
|
"valid_targets_min": 1631
|
|
},
|
|
{
|
|
"epoch": 5.432163436354112,
|
|
"grad_norm": 0.5666667792864185,
|
|
"learning_rate": 5.811759882718102e-06,
|
|
"loss": 0.2166,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12489330023527145,
|
|
"step": 10370,
|
|
"valid_targets_mean": 5411.4,
|
|
"valid_targets_min": 1509
|
|
},
|
|
{
|
|
"epoch": 5.434782608695652,
|
|
"grad_norm": 0.5258933268049755,
|
|
"learning_rate": 5.793360418562659e-06,
|
|
"loss": 0.2041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09626871347427368,
|
|
"step": 10375,
|
|
"valid_targets_mean": 4960.2,
|
|
"valid_targets_min": 2736
|
|
},
|
|
{
|
|
"epoch": 5.437401781037193,
|
|
"grad_norm": 0.6114570302288379,
|
|
"learning_rate": 5.7749851919537396e-06,
|
|
"loss": 0.2091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09290073066949844,
|
|
"step": 10380,
|
|
"valid_targets_mean": 3726.8,
|
|
"valid_targets_min": 1439
|
|
},
|
|
{
|
|
"epoch": 5.440020953378732,
|
|
"grad_norm": 0.5384858421780185,
|
|
"learning_rate": 5.756634234240825e-06,
|
|
"loss": 0.2165,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09385785460472107,
|
|
"step": 10385,
|
|
"valid_targets_mean": 4269.4,
|
|
"valid_targets_min": 1838
|
|
},
|
|
{
|
|
"epoch": 5.442640125720272,
|
|
"grad_norm": 0.5329360036480885,
|
|
"learning_rate": 5.738307576731943e-06,
|
|
"loss": 0.2023,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12725722789764404,
|
|
"step": 10390,
|
|
"valid_targets_mean": 5973.6,
|
|
"valid_targets_min": 2762
|
|
},
|
|
{
|
|
"epoch": 5.445259298061813,
|
|
"grad_norm": 0.5793722639828044,
|
|
"learning_rate": 5.7200052506937095e-06,
|
|
"loss": 0.211,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0956265777349472,
|
|
"step": 10395,
|
|
"valid_targets_mean": 3987.1,
|
|
"valid_targets_min": 1654
|
|
},
|
|
{
|
|
"epoch": 5.447878470403353,
|
|
"grad_norm": 0.496775138740992,
|
|
"learning_rate": 5.701727287351186e-06,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09867385029792786,
|
|
"step": 10400,
|
|
"valid_targets_mean": 5043.6,
|
|
"valid_targets_min": 2493
|
|
},
|
|
{
|
|
"epoch": 5.450497642744892,
|
|
"grad_norm": 0.5812377395766525,
|
|
"learning_rate": 5.6834737178879e-06,
|
|
"loss": 0.2188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09992499649524689,
|
|
"step": 10405,
|
|
"valid_targets_mean": 4468.9,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 5.453116815086433,
|
|
"grad_norm": 0.48686859517785813,
|
|
"learning_rate": 5.665244573445747e-06,
|
|
"loss": 0.208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09565421938896179,
|
|
"step": 10410,
|
|
"valid_targets_mean": 5241.0,
|
|
"valid_targets_min": 953
|
|
},
|
|
{
|
|
"epoch": 5.455735987427973,
|
|
"grad_norm": 0.5699010720655088,
|
|
"learning_rate": 5.647039885124956e-06,
|
|
"loss": 0.2108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1318211406469345,
|
|
"step": 10415,
|
|
"valid_targets_mean": 5277.0,
|
|
"valid_targets_min": 1265
|
|
},
|
|
{
|
|
"epoch": 5.458355159769513,
|
|
"grad_norm": 0.5097020234723649,
|
|
"learning_rate": 5.628859683984034e-06,
|
|
"loss": 0.2041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10591420531272888,
|
|
"step": 10420,
|
|
"valid_targets_mean": 6127.8,
|
|
"valid_targets_min": 2566
|
|
},
|
|
{
|
|
"epoch": 5.460974332111053,
|
|
"grad_norm": 0.5595183648685903,
|
|
"learning_rate": 5.610704001039706e-06,
|
|
"loss": 0.2101,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09012778103351593,
|
|
"step": 10425,
|
|
"valid_targets_mean": 3446.9,
|
|
"valid_targets_min": 1759
|
|
},
|
|
{
|
|
"epoch": 5.463593504452593,
|
|
"grad_norm": 0.5984489249129414,
|
|
"learning_rate": 5.592572867266877e-06,
|
|
"loss": 0.2007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09819424897432327,
|
|
"step": 10430,
|
|
"valid_targets_mean": 3659.4,
|
|
"valid_targets_min": 2179
|
|
},
|
|
{
|
|
"epoch": 5.466212676794133,
|
|
"grad_norm": 0.620023251286136,
|
|
"learning_rate": 5.57446631359855e-06,
|
|
"loss": 0.2045,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1056923121213913,
|
|
"step": 10435,
|
|
"valid_targets_mean": 4161.5,
|
|
"valid_targets_min": 1785
|
|
},
|
|
{
|
|
"epoch": 5.4688318491356736,
|
|
"grad_norm": 0.6052536403022243,
|
|
"learning_rate": 5.556384370925812e-06,
|
|
"loss": 0.1994,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11983170360326767,
|
|
"step": 10440,
|
|
"valid_targets_mean": 4467.9,
|
|
"valid_targets_min": 2345
|
|
},
|
|
{
|
|
"epoch": 5.471451021477213,
|
|
"grad_norm": 0.5988358860825104,
|
|
"learning_rate": 5.538327070097753e-06,
|
|
"loss": 0.2061,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11621883511543274,
|
|
"step": 10445,
|
|
"valid_targets_mean": 4057.2,
|
|
"valid_targets_min": 2767
|
|
},
|
|
{
|
|
"epoch": 5.474070193818753,
|
|
"grad_norm": 0.6636345304563924,
|
|
"learning_rate": 5.5202944419214236e-06,
|
|
"loss": 0.2099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10627111792564392,
|
|
"step": 10450,
|
|
"valid_targets_mean": 3450.2,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 5.476689366160294,
|
|
"grad_norm": 0.5966289856538027,
|
|
"learning_rate": 5.502286517161783e-06,
|
|
"loss": 0.2016,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10250097513198853,
|
|
"step": 10455,
|
|
"valid_targets_mean": 3777.1,
|
|
"valid_targets_min": 2004
|
|
},
|
|
{
|
|
"epoch": 5.479308538501833,
|
|
"grad_norm": 0.6305270129637696,
|
|
"learning_rate": 5.4843033265416426e-06,
|
|
"loss": 0.211,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11023075878620148,
|
|
"step": 10460,
|
|
"valid_targets_mean": 4273.0,
|
|
"valid_targets_min": 1469
|
|
},
|
|
{
|
|
"epoch": 5.481927710843373,
|
|
"grad_norm": 0.62686315049717,
|
|
"learning_rate": 5.466344900741616e-06,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11784721165895462,
|
|
"step": 10465,
|
|
"valid_targets_mean": 4196.8,
|
|
"valid_targets_min": 1889
|
|
},
|
|
{
|
|
"epoch": 5.484546883184914,
|
|
"grad_norm": 0.6191613884937408,
|
|
"learning_rate": 5.448411270400071e-06,
|
|
"loss": 0.1917,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10129757970571518,
|
|
"step": 10470,
|
|
"valid_targets_mean": 4319.2,
|
|
"valid_targets_min": 2046
|
|
},
|
|
{
|
|
"epoch": 5.487166055526454,
|
|
"grad_norm": 0.6381998419206792,
|
|
"learning_rate": 5.430502466113052e-06,
|
|
"loss": 0.1921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11723365634679794,
|
|
"step": 10475,
|
|
"valid_targets_mean": 4580.1,
|
|
"valid_targets_min": 3373
|
|
},
|
|
{
|
|
"epoch": 5.489785227867994,
|
|
"grad_norm": 0.6853399916665825,
|
|
"learning_rate": 5.4126185184342875e-06,
|
|
"loss": 0.2037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10599417984485626,
|
|
"step": 10480,
|
|
"valid_targets_mean": 3786.0,
|
|
"valid_targets_min": 1747
|
|
},
|
|
{
|
|
"epoch": 5.492404400209534,
|
|
"grad_norm": 0.6176675884859637,
|
|
"learning_rate": 5.39475945787505e-06,
|
|
"loss": 0.2049,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08931992948055267,
|
|
"step": 10485,
|
|
"valid_targets_mean": 3339.9,
|
|
"valid_targets_min": 1166
|
|
},
|
|
{
|
|
"epoch": 5.495023572551074,
|
|
"grad_norm": 0.5792463399349149,
|
|
"learning_rate": 5.376925314904202e-06,
|
|
"loss": 0.1959,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09918844699859619,
|
|
"step": 10490,
|
|
"valid_targets_mean": 4114.1,
|
|
"valid_targets_min": 1537
|
|
},
|
|
{
|
|
"epoch": 5.497642744892614,
|
|
"grad_norm": 0.5807901724764257,
|
|
"learning_rate": 5.359116119948056e-06,
|
|
"loss": 0.1973,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09208104759454727,
|
|
"step": 10495,
|
|
"valid_targets_mean": 4550.4,
|
|
"valid_targets_min": 3119
|
|
},
|
|
{
|
|
"epoch": 5.500261917234154,
|
|
"grad_norm": 0.6137099431127251,
|
|
"learning_rate": 5.341331903390381e-06,
|
|
"loss": 0.1979,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09576673805713654,
|
|
"step": 10500,
|
|
"valid_targets_mean": 3810.8,
|
|
"valid_targets_min": 2562
|
|
},
|
|
{
|
|
"epoch": 5.502881089575694,
|
|
"grad_norm": 0.6073790234492764,
|
|
"learning_rate": 5.323572695572326e-06,
|
|
"loss": 0.1962,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10099849849939346,
|
|
"step": 10505,
|
|
"valid_targets_mean": 4045.5,
|
|
"valid_targets_min": 2659
|
|
},
|
|
{
|
|
"epoch": 5.505500261917234,
|
|
"grad_norm": 0.553561835142201,
|
|
"learning_rate": 5.305838526792375e-06,
|
|
"loss": 0.1878,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08134942501783371,
|
|
"step": 10510,
|
|
"valid_targets_mean": 4064.8,
|
|
"valid_targets_min": 1349
|
|
},
|
|
{
|
|
"epoch": 5.5081194342587745,
|
|
"grad_norm": 0.6775278036738944,
|
|
"learning_rate": 5.2881294273062926e-06,
|
|
"loss": 0.2001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09701518714427948,
|
|
"step": 10515,
|
|
"valid_targets_mean": 3371.4,
|
|
"valid_targets_min": 2474
|
|
},
|
|
{
|
|
"epoch": 5.510738606600315,
|
|
"grad_norm": 0.6909881239559398,
|
|
"learning_rate": 5.270445427327071e-06,
|
|
"loss": 0.1979,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0812319666147232,
|
|
"step": 10520,
|
|
"valid_targets_mean": 2590.5,
|
|
"valid_targets_min": 738
|
|
},
|
|
{
|
|
"epoch": 5.513357778941854,
|
|
"grad_norm": 0.6275919728018728,
|
|
"learning_rate": 5.252786557024885e-06,
|
|
"loss": 0.1893,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09055991470813751,
|
|
"step": 10525,
|
|
"valid_targets_mean": 3570.5,
|
|
"valid_targets_min": 2033
|
|
},
|
|
{
|
|
"epoch": 5.5159769512833945,
|
|
"grad_norm": 0.721356563398191,
|
|
"learning_rate": 5.2351528465270405e-06,
|
|
"loss": 0.195,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10568110644817352,
|
|
"step": 10530,
|
|
"valid_targets_mean": 2933.1,
|
|
"valid_targets_min": 1346
|
|
},
|
|
{
|
|
"epoch": 5.518596123624935,
|
|
"grad_norm": 0.8071270869299125,
|
|
"learning_rate": 5.217544325917893e-06,
|
|
"loss": 0.1954,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10447847098112106,
|
|
"step": 10535,
|
|
"valid_targets_mean": 2991.1,
|
|
"valid_targets_min": 1821
|
|
},
|
|
{
|
|
"epoch": 5.521215295966474,
|
|
"grad_norm": 0.6504779203035157,
|
|
"learning_rate": 5.199961025238871e-06,
|
|
"loss": 0.1948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09409458935260773,
|
|
"step": 10540,
|
|
"valid_targets_mean": 3400.1,
|
|
"valid_targets_min": 2535
|
|
},
|
|
{
|
|
"epoch": 5.5238344683080145,
|
|
"grad_norm": 0.671693602313302,
|
|
"learning_rate": 5.182402974488323e-06,
|
|
"loss": 0.1952,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09331804513931274,
|
|
"step": 10545,
|
|
"valid_targets_mean": 3411.5,
|
|
"valid_targets_min": 2352
|
|
},
|
|
{
|
|
"epoch": 5.526453640649555,
|
|
"grad_norm": 0.6471336187560586,
|
|
"learning_rate": 5.164870203621553e-06,
|
|
"loss": 0.1928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08596295118331909,
|
|
"step": 10550,
|
|
"valid_targets_mean": 3319.8,
|
|
"valid_targets_min": 1153
|
|
},
|
|
{
|
|
"epoch": 5.529072812991095,
|
|
"grad_norm": 0.9139238811458623,
|
|
"learning_rate": 5.147362742550725e-06,
|
|
"loss": 0.1943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10043549537658691,
|
|
"step": 10555,
|
|
"valid_targets_mean": 3718.0,
|
|
"valid_targets_min": 2072
|
|
},
|
|
{
|
|
"epoch": 5.5316919853326345,
|
|
"grad_norm": 0.6312099733530194,
|
|
"learning_rate": 5.129880621144827e-06,
|
|
"loss": 0.1953,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0861826241016388,
|
|
"step": 10560,
|
|
"valid_targets_mean": 3451.0,
|
|
"valid_targets_min": 922
|
|
},
|
|
{
|
|
"epoch": 5.534311157674175,
|
|
"grad_norm": 0.6036615572114501,
|
|
"learning_rate": 5.112423869229608e-06,
|
|
"loss": 0.1927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07662707567214966,
|
|
"step": 10565,
|
|
"valid_targets_mean": 3381.6,
|
|
"valid_targets_min": 1417
|
|
},
|
|
{
|
|
"epoch": 5.536930330015715,
|
|
"grad_norm": 0.6239335015347284,
|
|
"learning_rate": 5.09499251658754e-06,
|
|
"loss": 0.2003,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08969300985336304,
|
|
"step": 10570,
|
|
"valid_targets_mean": 4260.6,
|
|
"valid_targets_min": 1844
|
|
},
|
|
{
|
|
"epoch": 5.539549502357255,
|
|
"grad_norm": 0.7308611942195394,
|
|
"learning_rate": 5.077586592957771e-06,
|
|
"loss": 0.1931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1015428975224495,
|
|
"step": 10575,
|
|
"valid_targets_mean": 3525.2,
|
|
"valid_targets_min": 1261
|
|
},
|
|
{
|
|
"epoch": 5.542168674698795,
|
|
"grad_norm": 0.6311203968047921,
|
|
"learning_rate": 5.060206128036034e-06,
|
|
"loss": 0.1944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11546990275382996,
|
|
"step": 10580,
|
|
"valid_targets_mean": 4548.6,
|
|
"valid_targets_min": 2077
|
|
},
|
|
{
|
|
"epoch": 5.544787847040335,
|
|
"grad_norm": 0.6104267190252021,
|
|
"learning_rate": 5.042851151474675e-06,
|
|
"loss": 0.1919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10856322199106216,
|
|
"step": 10585,
|
|
"valid_targets_mean": 4140.9,
|
|
"valid_targets_min": 2724
|
|
},
|
|
{
|
|
"epoch": 5.547407019381875,
|
|
"grad_norm": 0.667555633639888,
|
|
"learning_rate": 5.025521692882513e-06,
|
|
"loss": 0.1946,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14028789103031158,
|
|
"step": 10590,
|
|
"valid_targets_mean": 5291.4,
|
|
"valid_targets_min": 2292
|
|
},
|
|
{
|
|
"epoch": 5.550026191723416,
|
|
"grad_norm": 0.7172987509908402,
|
|
"learning_rate": 5.008217781824849e-06,
|
|
"loss": 0.1971,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0968562513589859,
|
|
"step": 10595,
|
|
"valid_targets_mean": 2965.1,
|
|
"valid_targets_min": 1377
|
|
},
|
|
{
|
|
"epoch": 5.552645364064955,
|
|
"grad_norm": 0.6177201842022538,
|
|
"learning_rate": 4.990939447823402e-06,
|
|
"loss": 0.198,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08947396278381348,
|
|
"step": 10600,
|
|
"valid_targets_mean": 3099.5,
|
|
"valid_targets_min": 1350
|
|
},
|
|
{
|
|
"epoch": 5.555264536406495,
|
|
"grad_norm": 0.6603323375361946,
|
|
"learning_rate": 4.973686720356248e-06,
|
|
"loss": 0.1945,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09418955445289612,
|
|
"step": 10605,
|
|
"valid_targets_mean": 3031.9,
|
|
"valid_targets_min": 654
|
|
},
|
|
{
|
|
"epoch": 5.557883708748036,
|
|
"grad_norm": 0.6490782382697062,
|
|
"learning_rate": 4.956459628857775e-06,
|
|
"loss": 0.1958,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07847204059362411,
|
|
"step": 10610,
|
|
"valid_targets_mean": 3463.8,
|
|
"valid_targets_min": 1706
|
|
},
|
|
{
|
|
"epoch": 5.560502881089576,
|
|
"grad_norm": 0.6090670662855951,
|
|
"learning_rate": 4.9392582027186466e-06,
|
|
"loss": 0.2016,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10062491148710251,
|
|
"step": 10615,
|
|
"valid_targets_mean": 4105.5,
|
|
"valid_targets_min": 1739
|
|
},
|
|
{
|
|
"epoch": 5.563122053431115,
|
|
"grad_norm": 0.6709318550615645,
|
|
"learning_rate": 4.922082471285712e-06,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08609208464622498,
|
|
"step": 10620,
|
|
"valid_targets_mean": 2944.4,
|
|
"valid_targets_min": 1302
|
|
},
|
|
{
|
|
"epoch": 5.565741225772656,
|
|
"grad_norm": 0.5884399803714986,
|
|
"learning_rate": 4.9049324638620245e-06,
|
|
"loss": 0.182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0818995013833046,
|
|
"step": 10625,
|
|
"valid_targets_mean": 3531.5,
|
|
"valid_targets_min": 2501
|
|
},
|
|
{
|
|
"epoch": 5.568360398114196,
|
|
"grad_norm": 0.5836595702624258,
|
|
"learning_rate": 4.8878082097067055e-06,
|
|
"loss": 0.1925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09440377354621887,
|
|
"step": 10630,
|
|
"valid_targets_mean": 4673.8,
|
|
"valid_targets_min": 3232
|
|
},
|
|
{
|
|
"epoch": 5.570979570455736,
|
|
"grad_norm": 0.6058345329660421,
|
|
"learning_rate": 4.870709738034982e-06,
|
|
"loss": 0.1831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1017429381608963,
|
|
"step": 10635,
|
|
"valid_targets_mean": 4465.1,
|
|
"valid_targets_min": 1101
|
|
},
|
|
{
|
|
"epoch": 5.573598742797276,
|
|
"grad_norm": 0.6198171565286366,
|
|
"learning_rate": 4.85363707801805e-06,
|
|
"loss": 0.2082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10447974503040314,
|
|
"step": 10640,
|
|
"valid_targets_mean": 4113.2,
|
|
"valid_targets_min": 2079
|
|
},
|
|
{
|
|
"epoch": 5.576217915138816,
|
|
"grad_norm": 1.4356392096515265,
|
|
"learning_rate": 4.836590258783118e-06,
|
|
"loss": 0.1954,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10219666361808777,
|
|
"step": 10645,
|
|
"valid_targets_mean": 3891.4,
|
|
"valid_targets_min": 2138
|
|
},
|
|
{
|
|
"epoch": 5.578837087480356,
|
|
"grad_norm": 0.5999752289086019,
|
|
"learning_rate": 4.819569309413266e-06,
|
|
"loss": 0.1962,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09590047597885132,
|
|
"step": 10650,
|
|
"valid_targets_mean": 3470.2,
|
|
"valid_targets_min": 1234
|
|
},
|
|
{
|
|
"epoch": 5.581456259821897,
|
|
"grad_norm": 0.6151013944912155,
|
|
"learning_rate": 4.802574258947459e-06,
|
|
"loss": 0.1968,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09681589901447296,
|
|
"step": 10655,
|
|
"valid_targets_mean": 4038.5,
|
|
"valid_targets_min": 1822
|
|
},
|
|
{
|
|
"epoch": 5.584075432163436,
|
|
"grad_norm": 0.5968394738555188,
|
|
"learning_rate": 4.7856051363804715e-06,
|
|
"loss": 0.2043,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09785608947277069,
|
|
"step": 10660,
|
|
"valid_targets_mean": 4214.6,
|
|
"valid_targets_min": 1651
|
|
},
|
|
{
|
|
"epoch": 5.586694604504976,
|
|
"grad_norm": 0.6446692840414369,
|
|
"learning_rate": 4.768661970662847e-06,
|
|
"loss": 0.1932,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11061255633831024,
|
|
"step": 10665,
|
|
"valid_targets_mean": 3447.9,
|
|
"valid_targets_min": 1434
|
|
},
|
|
{
|
|
"epoch": 5.589313776846517,
|
|
"grad_norm": 0.5956230197811474,
|
|
"learning_rate": 4.751744790700841e-06,
|
|
"loss": 0.1861,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10208860039710999,
|
|
"step": 10670,
|
|
"valid_targets_mean": 4753.4,
|
|
"valid_targets_min": 2725
|
|
},
|
|
{
|
|
"epoch": 5.591932949188057,
|
|
"grad_norm": 0.592416691781473,
|
|
"learning_rate": 4.7348536253563815e-06,
|
|
"loss": 0.1832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09418331831693649,
|
|
"step": 10675,
|
|
"valid_targets_mean": 4005.8,
|
|
"valid_targets_min": 1966
|
|
},
|
|
{
|
|
"epoch": 5.594552121529596,
|
|
"grad_norm": 0.6320872525497788,
|
|
"learning_rate": 4.717988503446993e-06,
|
|
"loss": 0.1802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08939506113529205,
|
|
"step": 10680,
|
|
"valid_targets_mean": 3669.0,
|
|
"valid_targets_min": 1407
|
|
},
|
|
{
|
|
"epoch": 5.597171293871137,
|
|
"grad_norm": 0.6581995553668584,
|
|
"learning_rate": 4.7011494537458035e-06,
|
|
"loss": 0.1822,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09837263822555542,
|
|
"step": 10685,
|
|
"valid_targets_mean": 4330.5,
|
|
"valid_targets_min": 1444
|
|
},
|
|
{
|
|
"epoch": 5.599790466212677,
|
|
"grad_norm": 0.6160637343893121,
|
|
"learning_rate": 4.684336504981426e-06,
|
|
"loss": 0.1746,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07830608636140823,
|
|
"step": 10690,
|
|
"valid_targets_mean": 3077.8,
|
|
"valid_targets_min": 1504
|
|
},
|
|
{
|
|
"epoch": 5.602409638554217,
|
|
"grad_norm": 0.6184021344561963,
|
|
"learning_rate": 4.667549685837962e-06,
|
|
"loss": 0.1721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0702492818236351,
|
|
"step": 10695,
|
|
"valid_targets_mean": 3051.6,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 5.605028810895757,
|
|
"grad_norm": 0.6898866591128014,
|
|
"learning_rate": 4.650789024954929e-06,
|
|
"loss": 0.1887,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07904627919197083,
|
|
"step": 10700,
|
|
"valid_targets_mean": 3231.8,
|
|
"valid_targets_min": 1288
|
|
},
|
|
{
|
|
"epoch": 5.607647983237297,
|
|
"grad_norm": 0.664381529980434,
|
|
"learning_rate": 4.634054550927218e-06,
|
|
"loss": 0.2058,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09320087730884552,
|
|
"step": 10705,
|
|
"valid_targets_mean": 4053.0,
|
|
"valid_targets_min": 1537
|
|
},
|
|
{
|
|
"epoch": 5.610267155578837,
|
|
"grad_norm": 0.6771694791746062,
|
|
"learning_rate": 4.6173462923050424e-06,
|
|
"loss": 0.1862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09745335578918457,
|
|
"step": 10710,
|
|
"valid_targets_mean": 3917.1,
|
|
"valid_targets_min": 2060
|
|
},
|
|
{
|
|
"epoch": 5.612886327920377,
|
|
"grad_norm": 0.6242309125978881,
|
|
"learning_rate": 4.600664277593892e-06,
|
|
"loss": 0.1913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1041133925318718,
|
|
"step": 10715,
|
|
"valid_targets_mean": 4482.6,
|
|
"valid_targets_min": 2515
|
|
},
|
|
{
|
|
"epoch": 5.615505500261917,
|
|
"grad_norm": 0.628486105645744,
|
|
"learning_rate": 4.584008535254482e-06,
|
|
"loss": 0.1862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09165383875370026,
|
|
"step": 10720,
|
|
"valid_targets_mean": 3847.2,
|
|
"valid_targets_min": 1744
|
|
},
|
|
{
|
|
"epoch": 5.618124672603457,
|
|
"grad_norm": 0.6165649846115652,
|
|
"learning_rate": 4.567379093702691e-06,
|
|
"loss": 0.1913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09430750459432602,
|
|
"step": 10725,
|
|
"valid_targets_mean": 5009.9,
|
|
"valid_targets_min": 2062
|
|
},
|
|
{
|
|
"epoch": 5.620743844944998,
|
|
"grad_norm": 0.6345298501930282,
|
|
"learning_rate": 4.550775981309561e-06,
|
|
"loss": 0.1865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08381767570972443,
|
|
"step": 10730,
|
|
"valid_targets_mean": 3550.4,
|
|
"valid_targets_min": 1181
|
|
},
|
|
{
|
|
"epoch": 5.623363017286538,
|
|
"grad_norm": 0.6446078817200693,
|
|
"learning_rate": 4.534199226401177e-06,
|
|
"loss": 0.1785,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10016798228025436,
|
|
"step": 10735,
|
|
"valid_targets_mean": 4010.6,
|
|
"valid_targets_min": 1937
|
|
},
|
|
{
|
|
"epoch": 5.625982189628077,
|
|
"grad_norm": 0.6983886083346015,
|
|
"learning_rate": 4.5176488572586765e-06,
|
|
"loss": 0.1948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10849693417549133,
|
|
"step": 10740,
|
|
"valid_targets_mean": 3624.9,
|
|
"valid_targets_min": 1649
|
|
},
|
|
{
|
|
"epoch": 5.628601361969618,
|
|
"grad_norm": 0.7505911544275667,
|
|
"learning_rate": 4.501124902118177e-06,
|
|
"loss": 0.1855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09305763244628906,
|
|
"step": 10745,
|
|
"valid_targets_mean": 2808.1,
|
|
"valid_targets_min": 1197
|
|
},
|
|
{
|
|
"epoch": 5.631220534311158,
|
|
"grad_norm": 0.7966629300365251,
|
|
"learning_rate": 4.48462738917073e-06,
|
|
"loss": 0.1888,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09724985063076019,
|
|
"step": 10750,
|
|
"valid_targets_mean": 3251.4,
|
|
"valid_targets_min": 1824
|
|
},
|
|
{
|
|
"epoch": 5.633839706652697,
|
|
"grad_norm": 0.6871211929063602,
|
|
"learning_rate": 4.468156346562282e-06,
|
|
"loss": 0.1978,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09823998808860779,
|
|
"step": 10755,
|
|
"valid_targets_mean": 3145.6,
|
|
"valid_targets_min": 538
|
|
},
|
|
{
|
|
"epoch": 5.636458878994238,
|
|
"grad_norm": 0.6329632570910867,
|
|
"learning_rate": 4.451711802393615e-06,
|
|
"loss": 0.1914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08536972105503082,
|
|
"step": 10760,
|
|
"valid_targets_mean": 3597.1,
|
|
"valid_targets_min": 537
|
|
},
|
|
{
|
|
"epoch": 5.639078051335778,
|
|
"grad_norm": 0.6906245376287569,
|
|
"learning_rate": 4.435293784720283e-06,
|
|
"loss": 0.2037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09746938198804855,
|
|
"step": 10765,
|
|
"valid_targets_mean": 3386.5,
|
|
"valid_targets_min": 2169
|
|
},
|
|
{
|
|
"epoch": 5.641697223677318,
|
|
"grad_norm": 0.6556767054751694,
|
|
"learning_rate": 4.418902321552629e-06,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0984741672873497,
|
|
"step": 10770,
|
|
"valid_targets_mean": 3687.6,
|
|
"valid_targets_min": 1625
|
|
},
|
|
{
|
|
"epoch": 5.6443163960188585,
|
|
"grad_norm": 0.6336878369135004,
|
|
"learning_rate": 4.4025374408556406e-06,
|
|
"loss": 0.2017,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08894580602645874,
|
|
"step": 10775,
|
|
"valid_targets_mean": 3807.0,
|
|
"valid_targets_min": 2642
|
|
},
|
|
{
|
|
"epoch": 5.646935568360398,
|
|
"grad_norm": 0.6360101774608651,
|
|
"learning_rate": 4.386199170549002e-06,
|
|
"loss": 0.1967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11144876480102539,
|
|
"step": 10780,
|
|
"valid_targets_mean": 4289.9,
|
|
"valid_targets_min": 2935
|
|
},
|
|
{
|
|
"epoch": 5.649554740701938,
|
|
"grad_norm": 0.5724879038589,
|
|
"learning_rate": 4.36988753850695e-06,
|
|
"loss": 0.1933,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10036087781190872,
|
|
"step": 10785,
|
|
"valid_targets_mean": 5189.9,
|
|
"valid_targets_min": 2652
|
|
},
|
|
{
|
|
"epoch": 5.6521739130434785,
|
|
"grad_norm": 0.6385118450508577,
|
|
"learning_rate": 4.353602572558322e-06,
|
|
"loss": 0.1943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1132982149720192,
|
|
"step": 10790,
|
|
"valid_targets_mean": 4117.2,
|
|
"valid_targets_min": 648
|
|
},
|
|
{
|
|
"epoch": 5.654793085385018,
|
|
"grad_norm": 0.6732253239872739,
|
|
"learning_rate": 4.337344300486426e-06,
|
|
"loss": 0.1921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08556917309761047,
|
|
"step": 10795,
|
|
"valid_targets_mean": 3492.8,
|
|
"valid_targets_min": 637
|
|
},
|
|
{
|
|
"epoch": 5.657412257726558,
|
|
"grad_norm": 0.7495367417437739,
|
|
"learning_rate": 4.3211127500290396e-06,
|
|
"loss": 0.189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09711206704378128,
|
|
"step": 10800,
|
|
"valid_targets_mean": 3412.5,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 5.6600314300680985,
|
|
"grad_norm": 0.6042811234142205,
|
|
"learning_rate": 4.3049079488783584e-06,
|
|
"loss": 0.1942,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08225598931312561,
|
|
"step": 10805,
|
|
"valid_targets_mean": 4227.4,
|
|
"valid_targets_min": 2182
|
|
},
|
|
{
|
|
"epoch": 5.662650602409639,
|
|
"grad_norm": 0.6598304132631697,
|
|
"learning_rate": 4.288729924680935e-06,
|
|
"loss": 0.1822,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08877980709075928,
|
|
"step": 10810,
|
|
"valid_targets_mean": 3211.1,
|
|
"valid_targets_min": 1372
|
|
},
|
|
{
|
|
"epoch": 5.665269774751179,
|
|
"grad_norm": 0.6929454788075643,
|
|
"learning_rate": 4.272578705037638e-06,
|
|
"loss": 0.1838,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11478810012340546,
|
|
"step": 10815,
|
|
"valid_targets_mean": 4490.9,
|
|
"valid_targets_min": 2594
|
|
},
|
|
{
|
|
"epoch": 5.6678889470927185,
|
|
"grad_norm": 0.6365936711525688,
|
|
"learning_rate": 4.256454317503611e-06,
|
|
"loss": 0.1881,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08295011520385742,
|
|
"step": 10820,
|
|
"valid_targets_mean": 3054.6,
|
|
"valid_targets_min": 530
|
|
},
|
|
{
|
|
"epoch": 5.670508119434259,
|
|
"grad_norm": 0.6655384406299206,
|
|
"learning_rate": 4.240356789588207e-06,
|
|
"loss": 0.1903,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1293245553970337,
|
|
"step": 10825,
|
|
"valid_targets_mean": 4110.2,
|
|
"valid_targets_min": 611
|
|
},
|
|
{
|
|
"epoch": 5.673127291775799,
|
|
"grad_norm": 0.7562332468745507,
|
|
"learning_rate": 4.22428614875497e-06,
|
|
"loss": 0.1888,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10113480687141418,
|
|
"step": 10830,
|
|
"valid_targets_mean": 3804.2,
|
|
"valid_targets_min": 1987
|
|
},
|
|
{
|
|
"epoch": 5.6757464641173385,
|
|
"grad_norm": 0.7150061744982886,
|
|
"learning_rate": 4.208242422421567e-06,
|
|
"loss": 0.1904,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09328234195709229,
|
|
"step": 10835,
|
|
"valid_targets_mean": 3203.2,
|
|
"valid_targets_min": 1362
|
|
},
|
|
{
|
|
"epoch": 5.678365636458879,
|
|
"grad_norm": 0.602777121567536,
|
|
"learning_rate": 4.192225637959743e-06,
|
|
"loss": 0.1924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.090776726603508,
|
|
"step": 10840,
|
|
"valid_targets_mean": 3245.1,
|
|
"valid_targets_min": 852
|
|
},
|
|
{
|
|
"epoch": 5.680984808800419,
|
|
"grad_norm": 0.6780464690155531,
|
|
"learning_rate": 4.176235822695285e-06,
|
|
"loss": 0.1886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07259844988584518,
|
|
"step": 10845,
|
|
"valid_targets_mean": 2537.4,
|
|
"valid_targets_min": 1195
|
|
},
|
|
{
|
|
"epoch": 5.683603981141959,
|
|
"grad_norm": 0.7248873902463573,
|
|
"learning_rate": 4.1602730039079645e-06,
|
|
"loss": 0.1928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10924700647592545,
|
|
"step": 10850,
|
|
"valid_targets_mean": 4383.0,
|
|
"valid_targets_min": 1397
|
|
},
|
|
{
|
|
"epoch": 5.686223153483499,
|
|
"grad_norm": 0.6970237986287124,
|
|
"learning_rate": 4.144337208831499e-06,
|
|
"loss": 0.1858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07984375953674316,
|
|
"step": 10855,
|
|
"valid_targets_mean": 3421.0,
|
|
"valid_targets_min": 1975
|
|
},
|
|
{
|
|
"epoch": 5.688842325825039,
|
|
"grad_norm": 0.6622949330317035,
|
|
"learning_rate": 4.128428464653484e-06,
|
|
"loss": 0.1755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09665311127901077,
|
|
"step": 10860,
|
|
"valid_targets_mean": 4100.6,
|
|
"valid_targets_min": 2564
|
|
},
|
|
{
|
|
"epoch": 5.691461498166579,
|
|
"grad_norm": 0.5573716439983707,
|
|
"learning_rate": 4.112546798515402e-06,
|
|
"loss": 0.1862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08861871808767319,
|
|
"step": 10865,
|
|
"valid_targets_mean": 4986.9,
|
|
"valid_targets_min": 2595
|
|
},
|
|
{
|
|
"epoch": 5.69408067050812,
|
|
"grad_norm": 0.6361048862555745,
|
|
"learning_rate": 4.0966922375124894e-06,
|
|
"loss": 0.1855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10304940491914749,
|
|
"step": 10870,
|
|
"valid_targets_mean": 3992.9,
|
|
"valid_targets_min": 2845
|
|
},
|
|
{
|
|
"epoch": 5.696699842849659,
|
|
"grad_norm": 0.6974644732997708,
|
|
"learning_rate": 4.080864808693787e-06,
|
|
"loss": 0.1796,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06854330003261566,
|
|
"step": 10875,
|
|
"valid_targets_mean": 2390.1,
|
|
"valid_targets_min": 993
|
|
},
|
|
{
|
|
"epoch": 5.699319015191199,
|
|
"grad_norm": 0.6106160850582569,
|
|
"learning_rate": 4.0650645390620135e-06,
|
|
"loss": 0.1731,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07218549400568008,
|
|
"step": 10880,
|
|
"valid_targets_mean": 3729.1,
|
|
"valid_targets_min": 1575
|
|
},
|
|
{
|
|
"epoch": 5.70193818753274,
|
|
"grad_norm": 0.6832198918686105,
|
|
"learning_rate": 4.049291455573561e-06,
|
|
"loss": 0.1807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09670407325029373,
|
|
"step": 10885,
|
|
"valid_targets_mean": 3469.0,
|
|
"valid_targets_min": 1360
|
|
},
|
|
{
|
|
"epoch": 5.70455735987428,
|
|
"grad_norm": 0.6375939367279172,
|
|
"learning_rate": 4.033545585138445e-06,
|
|
"loss": 0.1801,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08804913610219955,
|
|
"step": 10890,
|
|
"valid_targets_mean": 3555.5,
|
|
"valid_targets_min": 2278
|
|
},
|
|
{
|
|
"epoch": 5.707176532215819,
|
|
"grad_norm": 0.6613298545136447,
|
|
"learning_rate": 4.01782695462025e-06,
|
|
"loss": 0.197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10436461865901947,
|
|
"step": 10895,
|
|
"valid_targets_mean": 3812.2,
|
|
"valid_targets_min": 1454
|
|
},
|
|
{
|
|
"epoch": 5.70979570455736,
|
|
"grad_norm": 0.6612294096053106,
|
|
"learning_rate": 4.002135590836089e-06,
|
|
"loss": 0.1821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09107402712106705,
|
|
"step": 10900,
|
|
"valid_targets_mean": 3796.6,
|
|
"valid_targets_min": 2344
|
|
},
|
|
{
|
|
"epoch": 5.7124148768989,
|
|
"grad_norm": 0.7213039975427157,
|
|
"learning_rate": 3.986471520556552e-06,
|
|
"loss": 0.1841,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09840808808803558,
|
|
"step": 10905,
|
|
"valid_targets_mean": 3324.1,
|
|
"valid_targets_min": 1278
|
|
},
|
|
{
|
|
"epoch": 5.71503404924044,
|
|
"grad_norm": 0.7352845265339429,
|
|
"learning_rate": 3.9708347705056675e-06,
|
|
"loss": 0.1742,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09732574224472046,
|
|
"step": 10910,
|
|
"valid_targets_mean": 3381.4,
|
|
"valid_targets_min": 1858
|
|
},
|
|
{
|
|
"epoch": 5.71765322158198,
|
|
"grad_norm": 0.6978698405212208,
|
|
"learning_rate": 3.95522536736086e-06,
|
|
"loss": 0.1836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07258081436157227,
|
|
"step": 10915,
|
|
"valid_targets_mean": 2566.0,
|
|
"valid_targets_min": 420
|
|
},
|
|
{
|
|
"epoch": 5.72027239392352,
|
|
"grad_norm": 0.9922074600950967,
|
|
"learning_rate": 3.9396433377528746e-06,
|
|
"loss": 0.185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08637166023254395,
|
|
"step": 10920,
|
|
"valid_targets_mean": 2649.4,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 5.72289156626506,
|
|
"grad_norm": 0.6583115136938968,
|
|
"learning_rate": 3.924088708265792e-06,
|
|
"loss": 0.1898,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08856631070375443,
|
|
"step": 10925,
|
|
"valid_targets_mean": 4475.1,
|
|
"valid_targets_min": 1737
|
|
},
|
|
{
|
|
"epoch": 5.725510738606601,
|
|
"grad_norm": 0.6787761111080356,
|
|
"learning_rate": 3.908561505436905e-06,
|
|
"loss": 0.1847,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11043144762516022,
|
|
"step": 10930,
|
|
"valid_targets_mean": 4150.8,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 5.72812991094814,
|
|
"grad_norm": 0.6197335955768134,
|
|
"learning_rate": 3.89306175575676e-06,
|
|
"loss": 0.1829,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08951283246278763,
|
|
"step": 10935,
|
|
"valid_targets_mean": 3864.1,
|
|
"valid_targets_min": 1780
|
|
},
|
|
{
|
|
"epoch": 5.73074908328968,
|
|
"grad_norm": 0.5999510307672616,
|
|
"learning_rate": 3.8775894856690245e-06,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08658986538648605,
|
|
"step": 10940,
|
|
"valid_targets_mean": 4058.0,
|
|
"valid_targets_min": 2185
|
|
},
|
|
{
|
|
"epoch": 5.733368255631221,
|
|
"grad_norm": 0.4682407816392999,
|
|
"learning_rate": 3.862144721570504e-06,
|
|
"loss": 0.1553,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04221086576581001,
|
|
"step": 10945,
|
|
"valid_targets_mean": 2736.0,
|
|
"valid_targets_min": 876
|
|
},
|
|
{
|
|
"epoch": 5.735987427972761,
|
|
"grad_norm": 0.547156207816398,
|
|
"learning_rate": 3.84672748981108e-06,
|
|
"loss": 0.0959,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03678930923342705,
|
|
"step": 10950,
|
|
"valid_targets_mean": 1472.5,
|
|
"valid_targets_min": 324
|
|
},
|
|
{
|
|
"epoch": 5.7386066003143,
|
|
"grad_norm": 0.3865703266841375,
|
|
"learning_rate": 3.83133781669365e-06,
|
|
"loss": 0.0739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04523061215877533,
|
|
"step": 10955,
|
|
"valid_targets_mean": 2699.9,
|
|
"valid_targets_min": 1064
|
|
},
|
|
{
|
|
"epoch": 5.741225772655841,
|
|
"grad_norm": 0.40660624020334646,
|
|
"learning_rate": 3.815975728474106e-06,
|
|
"loss": 0.0732,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03883367404341698,
|
|
"step": 10960,
|
|
"valid_targets_mean": 3137.0,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 5.743844944997381,
|
|
"grad_norm": 0.32419841244438224,
|
|
"learning_rate": 3.800641251361268e-06,
|
|
"loss": 0.0863,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02795051783323288,
|
|
"step": 10965,
|
|
"valid_targets_mean": 3472.9,
|
|
"valid_targets_min": 1223
|
|
},
|
|
{
|
|
"epoch": 5.746464117338921,
|
|
"grad_norm": 0.39505777994688596,
|
|
"learning_rate": 3.7853344115168633e-06,
|
|
"loss": 0.0739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04061180725693703,
|
|
"step": 10970,
|
|
"valid_targets_mean": 3326.0,
|
|
"valid_targets_min": 1431
|
|
},
|
|
{
|
|
"epoch": 5.749083289680461,
|
|
"grad_norm": 0.7389516836791228,
|
|
"learning_rate": 3.7700552350554477e-06,
|
|
"loss": 0.1073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04988997429609299,
|
|
"step": 10975,
|
|
"valid_targets_mean": 1161.2,
|
|
"valid_targets_min": 559
|
|
},
|
|
{
|
|
"epoch": 5.751702462022001,
|
|
"grad_norm": 0.37442300106849613,
|
|
"learning_rate": 3.7548037480443977e-06,
|
|
"loss": 0.1001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037189219146966934,
|
|
"step": 10980,
|
|
"valid_targets_mean": 3598.2,
|
|
"valid_targets_min": 1623
|
|
},
|
|
{
|
|
"epoch": 5.754321634363541,
|
|
"grad_norm": 0.42154321075664153,
|
|
"learning_rate": 3.7395799765038422e-06,
|
|
"loss": 0.0834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043459370732307434,
|
|
"step": 10985,
|
|
"valid_targets_mean": 3178.2,
|
|
"valid_targets_min": 846
|
|
},
|
|
{
|
|
"epoch": 5.756940806705082,
|
|
"grad_norm": 0.4928263714664533,
|
|
"learning_rate": 3.724383946406631e-06,
|
|
"loss": 0.119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0836353674530983,
|
|
"step": 10990,
|
|
"valid_targets_mean": 3622.6,
|
|
"valid_targets_min": 1949
|
|
},
|
|
{
|
|
"epoch": 5.759559979046621,
|
|
"grad_norm": 0.3711070967607157,
|
|
"learning_rate": 3.709215683678282e-06,
|
|
"loss": 0.0608,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03476902097463608,
|
|
"step": 10995,
|
|
"valid_targets_mean": 3344.9,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 5.762179151388161,
|
|
"grad_norm": 0.4475362921935387,
|
|
"learning_rate": 3.6940752141969415e-06,
|
|
"loss": 0.0734,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041336361318826675,
|
|
"step": 11000,
|
|
"valid_targets_mean": 2391.6,
|
|
"valid_targets_min": 776
|
|
},
|
|
{
|
|
"epoch": 5.764798323729702,
|
|
"grad_norm": 0.38741715814397437,
|
|
"learning_rate": 3.678962563793325e-06,
|
|
"loss": 0.0753,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04356415197253227,
|
|
"step": 11005,
|
|
"valid_targets_mean": 2818.5,
|
|
"valid_targets_min": 754
|
|
},
|
|
{
|
|
"epoch": 5.767417496071241,
|
|
"grad_norm": 0.5117880596971838,
|
|
"learning_rate": 3.6638777582507177e-06,
|
|
"loss": 0.0927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047148026525974274,
|
|
"step": 11010,
|
|
"valid_targets_mean": 1646.6,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 5.770036668412781,
|
|
"grad_norm": 0.41930356078044206,
|
|
"learning_rate": 3.6488208233048594e-06,
|
|
"loss": 0.0875,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03821966052055359,
|
|
"step": 11015,
|
|
"valid_targets_mean": 3374.8,
|
|
"valid_targets_min": 1862
|
|
},
|
|
{
|
|
"epoch": 5.772655840754322,
|
|
"grad_norm": 0.43534620045705974,
|
|
"learning_rate": 3.6337917846439806e-06,
|
|
"loss": 0.0737,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04205740988254547,
|
|
"step": 11020,
|
|
"valid_targets_mean": 2837.4,
|
|
"valid_targets_min": 574
|
|
},
|
|
{
|
|
"epoch": 5.775275013095862,
|
|
"grad_norm": 0.5249942594507854,
|
|
"learning_rate": 3.6187906679086827e-06,
|
|
"loss": 0.0705,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03166569769382477,
|
|
"step": 11025,
|
|
"valid_targets_mean": 2618.4,
|
|
"valid_targets_min": 523
|
|
},
|
|
{
|
|
"epoch": 5.777894185437402,
|
|
"grad_norm": 0.4626425302288907,
|
|
"learning_rate": 3.603817498691966e-06,
|
|
"loss": 0.0738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04045230522751808,
|
|
"step": 11030,
|
|
"valid_targets_mean": 3548.9,
|
|
"valid_targets_min": 2189
|
|
},
|
|
{
|
|
"epoch": 5.780513357778942,
|
|
"grad_norm": 0.37052238295789225,
|
|
"learning_rate": 3.588872302539117e-06,
|
|
"loss": 0.0775,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02780608832836151,
|
|
"step": 11035,
|
|
"valid_targets_mean": 3564.6,
|
|
"valid_targets_min": 827
|
|
},
|
|
{
|
|
"epoch": 5.783132530120482,
|
|
"grad_norm": 0.4707983746182179,
|
|
"learning_rate": 3.5739551049477174e-06,
|
|
"loss": 0.088,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04321026802062988,
|
|
"step": 11040,
|
|
"valid_targets_mean": 1936.6,
|
|
"valid_targets_min": 597
|
|
},
|
|
{
|
|
"epoch": 5.785751702462022,
|
|
"grad_norm": 0.5187164571096663,
|
|
"learning_rate": 3.5590659313675735e-06,
|
|
"loss": 0.0821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03667140007019043,
|
|
"step": 11045,
|
|
"valid_targets_mean": 1747.5,
|
|
"valid_targets_min": 701
|
|
},
|
|
{
|
|
"epoch": 5.788370874803562,
|
|
"grad_norm": 0.5938730178938206,
|
|
"learning_rate": 3.5442048072006884e-06,
|
|
"loss": 0.1082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10945126414299011,
|
|
"step": 11050,
|
|
"valid_targets_mean": 3329.6,
|
|
"valid_targets_min": 1652
|
|
},
|
|
{
|
|
"epoch": 5.790990047145102,
|
|
"grad_norm": 0.4288560923289433,
|
|
"learning_rate": 3.5293717578012055e-06,
|
|
"loss": 0.087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034772470593452454,
|
|
"step": 11055,
|
|
"valid_targets_mean": 3117.1,
|
|
"valid_targets_min": 786
|
|
},
|
|
{
|
|
"epoch": 5.793609219486642,
|
|
"grad_norm": 0.5250971863597397,
|
|
"learning_rate": 3.514566808475375e-06,
|
|
"loss": 0.0725,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04586917161941528,
|
|
"step": 11060,
|
|
"valid_targets_mean": 2574.4,
|
|
"valid_targets_min": 971
|
|
},
|
|
{
|
|
"epoch": 5.7962283918281825,
|
|
"grad_norm": 0.4121152028658852,
|
|
"learning_rate": 3.4997899844814874e-06,
|
|
"loss": 0.1037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03905129432678223,
|
|
"step": 11065,
|
|
"valid_targets_mean": 3320.0,
|
|
"valid_targets_min": 2610
|
|
},
|
|
{
|
|
"epoch": 5.798847564169723,
|
|
"grad_norm": 0.4432979042499083,
|
|
"learning_rate": 3.4850413110298864e-06,
|
|
"loss": 0.0886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02789444476366043,
|
|
"step": 11070,
|
|
"valid_targets_mean": 2580.6,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 5.801466736511262,
|
|
"grad_norm": 0.5486777703978877,
|
|
"learning_rate": 3.470320813282848e-06,
|
|
"loss": 0.0802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03787270560860634,
|
|
"step": 11075,
|
|
"valid_targets_mean": 1537.1,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 5.8040859088528025,
|
|
"grad_norm": 0.593247965443131,
|
|
"learning_rate": 3.4556285163546143e-06,
|
|
"loss": 0.2632,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10560999810695648,
|
|
"step": 11080,
|
|
"valid_targets_mean": 2271.2,
|
|
"valid_targets_min": 987
|
|
},
|
|
{
|
|
"epoch": 5.806705081194343,
|
|
"grad_norm": 0.605181929270286,
|
|
"learning_rate": 3.4409644453112857e-06,
|
|
"loss": 0.0934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04537982866168022,
|
|
"step": 11085,
|
|
"valid_targets_mean": 1351.5,
|
|
"valid_targets_min": 329
|
|
},
|
|
{
|
|
"epoch": 5.809324253535882,
|
|
"grad_norm": 0.5073357781037506,
|
|
"learning_rate": 3.426328625170823e-06,
|
|
"loss": 0.0821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03772849217057228,
|
|
"step": 11090,
|
|
"valid_targets_mean": 1553.9,
|
|
"valid_targets_min": 608
|
|
},
|
|
{
|
|
"epoch": 5.8119434258774225,
|
|
"grad_norm": 0.49130373857608134,
|
|
"learning_rate": 3.4117210809029876e-06,
|
|
"loss": 0.1719,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04965986683964729,
|
|
"step": 11095,
|
|
"valid_targets_mean": 3645.6,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 5.814562598218963,
|
|
"grad_norm": 0.30933644359373974,
|
|
"learning_rate": 3.3971418374292963e-06,
|
|
"loss": 0.0707,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02171337977051735,
|
|
"step": 11100,
|
|
"valid_targets_mean": 4720.6,
|
|
"valid_targets_min": 3529
|
|
},
|
|
{
|
|
"epoch": 5.817181770560503,
|
|
"grad_norm": 0.45661785074403527,
|
|
"learning_rate": 3.3825909196229858e-06,
|
|
"loss": 0.0842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04527841880917549,
|
|
"step": 11105,
|
|
"valid_targets_mean": 4437.6,
|
|
"valid_targets_min": 2336
|
|
},
|
|
{
|
|
"epoch": 5.819800942902043,
|
|
"grad_norm": 0.42559521589327926,
|
|
"learning_rate": 3.3680683523089662e-06,
|
|
"loss": 0.0775,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03421613574028015,
|
|
"step": 11110,
|
|
"valid_targets_mean": 3533.1,
|
|
"valid_targets_min": 1226
|
|
},
|
|
{
|
|
"epoch": 5.822420115243583,
|
|
"grad_norm": 0.4042524258233285,
|
|
"learning_rate": 3.353574160263784e-06,
|
|
"loss": 0.0726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03365694358944893,
|
|
"step": 11115,
|
|
"valid_targets_mean": 3299.2,
|
|
"valid_targets_min": 798
|
|
},
|
|
{
|
|
"epoch": 5.825039287585123,
|
|
"grad_norm": 0.43669496834047056,
|
|
"learning_rate": 3.33910836821556e-06,
|
|
"loss": 0.0711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03029424138367176,
|
|
"step": 11120,
|
|
"valid_targets_mean": 1836.9,
|
|
"valid_targets_min": 723
|
|
},
|
|
{
|
|
"epoch": 5.827658459926663,
|
|
"grad_norm": 0.43820954403937185,
|
|
"learning_rate": 3.3246710008439777e-06,
|
|
"loss": 0.0705,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03753548488020897,
|
|
"step": 11125,
|
|
"valid_targets_mean": 3316.0,
|
|
"valid_targets_min": 898
|
|
},
|
|
{
|
|
"epoch": 5.830277632268203,
|
|
"grad_norm": 0.3244610135700791,
|
|
"learning_rate": 3.3102620827802247e-06,
|
|
"loss": 0.0687,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026321982964873314,
|
|
"step": 11130,
|
|
"valid_targets_mean": 4027.2,
|
|
"valid_targets_min": 3278
|
|
},
|
|
{
|
|
"epoch": 5.832896804609743,
|
|
"grad_norm": 0.49542063669645076,
|
|
"learning_rate": 3.295881638606948e-06,
|
|
"loss": 0.0711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03268700838088989,
|
|
"step": 11135,
|
|
"valid_targets_mean": 2052.8,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 5.835515976951283,
|
|
"grad_norm": 0.5033031868779394,
|
|
"learning_rate": 3.2815296928582163e-06,
|
|
"loss": 0.1023,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03714592009782791,
|
|
"step": 11140,
|
|
"valid_targets_mean": 1574.9,
|
|
"valid_targets_min": 538
|
|
},
|
|
{
|
|
"epoch": 5.838135149292824,
|
|
"grad_norm": 0.37289485687861634,
|
|
"learning_rate": 3.267206270019481e-06,
|
|
"loss": 0.0582,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0287267304956913,
|
|
"step": 11145,
|
|
"valid_targets_mean": 1956.9,
|
|
"valid_targets_min": 654
|
|
},
|
|
{
|
|
"epoch": 5.840754321634364,
|
|
"grad_norm": 0.4629461873322304,
|
|
"learning_rate": 3.252911394527527e-06,
|
|
"loss": 0.0707,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04151737689971924,
|
|
"step": 11150,
|
|
"valid_targets_mean": 2735.4,
|
|
"valid_targets_min": 930
|
|
},
|
|
{
|
|
"epoch": 5.843373493975903,
|
|
"grad_norm": 0.4710233917879916,
|
|
"learning_rate": 3.2386450907704426e-06,
|
|
"loss": 0.064,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03753400593996048,
|
|
"step": 11155,
|
|
"valid_targets_mean": 2965.0,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 5.845992666317444,
|
|
"grad_norm": 0.43800059549274245,
|
|
"learning_rate": 3.2244073830875512e-06,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03819188103079796,
|
|
"step": 11160,
|
|
"valid_targets_mean": 3344.6,
|
|
"valid_targets_min": 999
|
|
},
|
|
{
|
|
"epoch": 5.848611838658984,
|
|
"grad_norm": 0.5721017706752082,
|
|
"learning_rate": 3.2101982957694244e-06,
|
|
"loss": 0.0856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04740625619888306,
|
|
"step": 11165,
|
|
"valid_targets_mean": 2445.4,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 5.8512310110005235,
|
|
"grad_norm": 0.4113218661338274,
|
|
"learning_rate": 3.1960178530577645e-06,
|
|
"loss": 0.0885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037107281386852264,
|
|
"step": 11170,
|
|
"valid_targets_mean": 3514.9,
|
|
"valid_targets_min": 1068
|
|
},
|
|
{
|
|
"epoch": 5.853850183342064,
|
|
"grad_norm": 0.49946959668671437,
|
|
"learning_rate": 3.181866079145446e-06,
|
|
"loss": 0.0747,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0346212312579155,
|
|
"step": 11175,
|
|
"valid_targets_mean": 1602.8,
|
|
"valid_targets_min": 536
|
|
},
|
|
{
|
|
"epoch": 5.856469355683604,
|
|
"grad_norm": 0.5657946144647372,
|
|
"learning_rate": 3.167742998176391e-06,
|
|
"loss": 0.1159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09299509227275848,
|
|
"step": 11180,
|
|
"valid_targets_mean": 3568.8,
|
|
"valid_targets_min": 1507
|
|
},
|
|
{
|
|
"epoch": 5.859088528025144,
|
|
"grad_norm": 0.47555773468925827,
|
|
"learning_rate": 3.1536486342455987e-06,
|
|
"loss": 0.0839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041637204587459564,
|
|
"step": 11185,
|
|
"valid_targets_mean": 3292.2,
|
|
"valid_targets_min": 2342
|
|
},
|
|
{
|
|
"epoch": 5.861707700366684,
|
|
"grad_norm": 0.6132193893444337,
|
|
"learning_rate": 3.1395830113990633e-06,
|
|
"loss": 0.1023,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0404975600540638,
|
|
"step": 11190,
|
|
"valid_targets_mean": 1567.5,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 5.864326872708224,
|
|
"grad_norm": 0.6062733694551393,
|
|
"learning_rate": 3.125546153633747e-06,
|
|
"loss": 0.0697,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026830414310097694,
|
|
"step": 11195,
|
|
"valid_targets_mean": 1145.4,
|
|
"valid_targets_min": 449
|
|
},
|
|
{
|
|
"epoch": 5.866946045049764,
|
|
"grad_norm": 0.3731457384576194,
|
|
"learning_rate": 3.1115380848975363e-06,
|
|
"loss": 0.0772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030195914208889008,
|
|
"step": 11200,
|
|
"valid_targets_mean": 3191.0,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 5.869565217391305,
|
|
"grad_norm": 0.6184036320128709,
|
|
"learning_rate": 3.0975588290892e-06,
|
|
"loss": 0.0827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043835822492837906,
|
|
"step": 11205,
|
|
"valid_targets_mean": 1924.8,
|
|
"valid_targets_min": 577
|
|
},
|
|
{
|
|
"epoch": 5.872184389732844,
|
|
"grad_norm": 0.3683210417331091,
|
|
"learning_rate": 3.083608410058356e-06,
|
|
"loss": 0.0651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03257773816585541,
|
|
"step": 11210,
|
|
"valid_targets_mean": 3992.5,
|
|
"valid_targets_min": 726
|
|
},
|
|
{
|
|
"epoch": 5.874803562074384,
|
|
"grad_norm": 0.4266228733303925,
|
|
"learning_rate": 3.06968685160542e-06,
|
|
"loss": 0.0641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03843538835644722,
|
|
"step": 11215,
|
|
"valid_targets_mean": 3626.4,
|
|
"valid_targets_min": 2252
|
|
},
|
|
{
|
|
"epoch": 5.877422734415925,
|
|
"grad_norm": 0.459369175896622,
|
|
"learning_rate": 3.055794177481559e-06,
|
|
"loss": 0.0716,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02921932563185692,
|
|
"step": 11220,
|
|
"valid_targets_mean": 1556.2,
|
|
"valid_targets_min": 531
|
|
},
|
|
{
|
|
"epoch": 5.880041906757465,
|
|
"grad_norm": 0.5033875693245169,
|
|
"learning_rate": 3.0419304113886894e-06,
|
|
"loss": 0.0885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037124909460544586,
|
|
"step": 11225,
|
|
"valid_targets_mean": 2410.5,
|
|
"valid_targets_min": 744
|
|
},
|
|
{
|
|
"epoch": 5.882661079099004,
|
|
"grad_norm": 0.4929635596819927,
|
|
"learning_rate": 3.028095576979375e-06,
|
|
"loss": 0.0848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04303295165300369,
|
|
"step": 11230,
|
|
"valid_targets_mean": 3795.1,
|
|
"valid_targets_min": 3136
|
|
},
|
|
{
|
|
"epoch": 5.885280251440545,
|
|
"grad_norm": 0.34164401492577223,
|
|
"learning_rate": 3.014289697856845e-06,
|
|
"loss": 0.0721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0300001110881567,
|
|
"step": 11235,
|
|
"valid_targets_mean": 3324.4,
|
|
"valid_targets_min": 2366
|
|
},
|
|
{
|
|
"epoch": 5.887899423782085,
|
|
"grad_norm": 0.4084330205871245,
|
|
"learning_rate": 3.000512797574917e-06,
|
|
"loss": 0.0692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02712399885058403,
|
|
"step": 11240,
|
|
"valid_targets_mean": 1668.1,
|
|
"valid_targets_min": 669
|
|
},
|
|
{
|
|
"epoch": 5.890518596123625,
|
|
"grad_norm": 0.5342207638177974,
|
|
"learning_rate": 2.9867648996379706e-06,
|
|
"loss": 0.0855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0455302894115448,
|
|
"step": 11245,
|
|
"valid_targets_mean": 3529.8,
|
|
"valid_targets_min": 1785
|
|
},
|
|
{
|
|
"epoch": 5.893137768465165,
|
|
"grad_norm": 0.34233018289648487,
|
|
"learning_rate": 2.973046027500912e-06,
|
|
"loss": 0.06,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03136247768998146,
|
|
"step": 11250,
|
|
"valid_targets_mean": 4120.1,
|
|
"valid_targets_min": 3862
|
|
},
|
|
{
|
|
"epoch": 5.895756940806705,
|
|
"grad_norm": 0.42123279871969677,
|
|
"learning_rate": 2.9593562045691104e-06,
|
|
"loss": 0.0742,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03143005073070526,
|
|
"step": 11255,
|
|
"valid_targets_mean": 3062.2,
|
|
"valid_targets_min": 539
|
|
},
|
|
{
|
|
"epoch": 5.898376113148245,
|
|
"grad_norm": 0.43786853410663623,
|
|
"learning_rate": 2.9456954541983982e-06,
|
|
"loss": 0.0754,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029972806572914124,
|
|
"step": 11260,
|
|
"valid_targets_mean": 2781.5,
|
|
"valid_targets_min": 1016
|
|
},
|
|
{
|
|
"epoch": 5.900995285489786,
|
|
"grad_norm": 0.7461927569998023,
|
|
"learning_rate": 2.932063799694984e-06,
|
|
"loss": 0.086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04962751269340515,
|
|
"step": 11265,
|
|
"valid_targets_mean": 1102.9,
|
|
"valid_targets_min": 888
|
|
},
|
|
{
|
|
"epoch": 5.903614457831325,
|
|
"grad_norm": 0.5102448970842115,
|
|
"learning_rate": 2.918461264315462e-06,
|
|
"loss": 0.0777,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028521444648504257,
|
|
"step": 11270,
|
|
"valid_targets_mean": 2602.5,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 5.906233630172865,
|
|
"grad_norm": 0.6825464712654099,
|
|
"learning_rate": 2.904887871266724e-06,
|
|
"loss": 0.0742,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047726817429065704,
|
|
"step": 11275,
|
|
"valid_targets_mean": 1592.9,
|
|
"valid_targets_min": 605
|
|
},
|
|
{
|
|
"epoch": 5.908852802514406,
|
|
"grad_norm": 0.3946021142719126,
|
|
"learning_rate": 2.8913436437059526e-06,
|
|
"loss": 0.0759,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030810732394456863,
|
|
"step": 11280,
|
|
"valid_targets_mean": 2899.1,
|
|
"valid_targets_min": 1117
|
|
},
|
|
{
|
|
"epoch": 5.911471974855946,
|
|
"grad_norm": 0.43552867821377794,
|
|
"learning_rate": 2.8778286047405755e-06,
|
|
"loss": 0.096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034283559769392014,
|
|
"step": 11285,
|
|
"valid_targets_mean": 3376.6,
|
|
"valid_targets_min": 1388
|
|
},
|
|
{
|
|
"epoch": 5.914091147197485,
|
|
"grad_norm": 0.43670750851172746,
|
|
"learning_rate": 2.8643427774282174e-06,
|
|
"loss": 0.0748,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032360583543777466,
|
|
"step": 11290,
|
|
"valid_targets_mean": 3243.4,
|
|
"valid_targets_min": 524
|
|
},
|
|
{
|
|
"epoch": 5.916710319539026,
|
|
"grad_norm": 0.480477215537576,
|
|
"learning_rate": 2.8508861847766665e-06,
|
|
"loss": 0.0655,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03470785170793533,
|
|
"step": 11295,
|
|
"valid_targets_mean": 2468.0,
|
|
"valid_targets_min": 970
|
|
},
|
|
{
|
|
"epoch": 5.919329491880566,
|
|
"grad_norm": 0.5347638590624609,
|
|
"learning_rate": 2.8374588497438415e-06,
|
|
"loss": 0.0859,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03941023349761963,
|
|
"step": 11300,
|
|
"valid_targets_mean": 2959.6,
|
|
"valid_targets_min": 877
|
|
},
|
|
{
|
|
"epoch": 5.921948664222106,
|
|
"grad_norm": 0.39492224392015796,
|
|
"learning_rate": 2.8240607952377265e-06,
|
|
"loss": 0.0801,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029691524803638458,
|
|
"step": 11305,
|
|
"valid_targets_mean": 3886.5,
|
|
"valid_targets_min": 3251
|
|
},
|
|
{
|
|
"epoch": 5.924567836563646,
|
|
"grad_norm": 0.5711688477381672,
|
|
"learning_rate": 2.810692044116381e-06,
|
|
"loss": 0.063,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0402698814868927,
|
|
"step": 11310,
|
|
"valid_targets_mean": 2695.6,
|
|
"valid_targets_min": 564
|
|
},
|
|
{
|
|
"epoch": 5.927187008905186,
|
|
"grad_norm": 0.39760929628729696,
|
|
"learning_rate": 2.797352619187834e-06,
|
|
"loss": 0.0721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03923739492893219,
|
|
"step": 11315,
|
|
"valid_targets_mean": 4404.8,
|
|
"valid_targets_min": 2555
|
|
},
|
|
{
|
|
"epoch": 5.929806181246726,
|
|
"grad_norm": 0.3715878331665758,
|
|
"learning_rate": 2.7840425432101237e-06,
|
|
"loss": 0.0646,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03085777349770069,
|
|
"step": 11320,
|
|
"valid_targets_mean": 4410.5,
|
|
"valid_targets_min": 3261
|
|
},
|
|
{
|
|
"epoch": 5.9324253535882665,
|
|
"grad_norm": 0.4504797890028626,
|
|
"learning_rate": 2.7707618388911715e-06,
|
|
"loss": 0.0589,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.024951647967100143,
|
|
"step": 11325,
|
|
"valid_targets_mean": 1768.5,
|
|
"valid_targets_min": 618
|
|
},
|
|
{
|
|
"epoch": 5.935044525929806,
|
|
"grad_norm": 0.3407615967389443,
|
|
"learning_rate": 2.757510528888836e-06,
|
|
"loss": 0.0818,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033380474895238876,
|
|
"step": 11330,
|
|
"valid_targets_mean": 3985.8,
|
|
"valid_targets_min": 3213
|
|
},
|
|
{
|
|
"epoch": 5.937663698271346,
|
|
"grad_norm": 0.38157658006958145,
|
|
"learning_rate": 2.7442886358107834e-06,
|
|
"loss": 0.0765,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08153171092271805,
|
|
"step": 11335,
|
|
"valid_targets_mean": 3149.1,
|
|
"valid_targets_min": 1539
|
|
},
|
|
{
|
|
"epoch": 5.9402828706128865,
|
|
"grad_norm": 0.3566791419158281,
|
|
"learning_rate": 2.731096182214521e-06,
|
|
"loss": 0.0842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026738319545984268,
|
|
"step": 11340,
|
|
"valid_targets_mean": 2094.8,
|
|
"valid_targets_min": 509
|
|
},
|
|
{
|
|
"epoch": 5.942902042954426,
|
|
"grad_norm": 0.34299001263736395,
|
|
"learning_rate": 2.717933190607318e-06,
|
|
"loss": 0.0804,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.01958930492401123,
|
|
"step": 11345,
|
|
"valid_targets_mean": 2441.4,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 5.945521215295966,
|
|
"grad_norm": 0.45419186364997993,
|
|
"learning_rate": 2.704799683446184e-06,
|
|
"loss": 0.0931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035698674619197845,
|
|
"step": 11350,
|
|
"valid_targets_mean": 2740.9,
|
|
"valid_targets_min": 737
|
|
},
|
|
{
|
|
"epoch": 5.9481403876375065,
|
|
"grad_norm": 0.3987062350948198,
|
|
"learning_rate": 2.691695683137827e-06,
|
|
"loss": 0.0628,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03007950820028782,
|
|
"step": 11355,
|
|
"valid_targets_mean": 2726.9,
|
|
"valid_targets_min": 492
|
|
},
|
|
{
|
|
"epoch": 5.950759559979047,
|
|
"grad_norm": 0.6154943388101383,
|
|
"learning_rate": 2.6786212120386124e-06,
|
|
"loss": 0.0802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04654261842370033,
|
|
"step": 11360,
|
|
"valid_targets_mean": 1607.9,
|
|
"valid_targets_min": 833
|
|
},
|
|
{
|
|
"epoch": 5.953378732320587,
|
|
"grad_norm": 0.35676724141161104,
|
|
"learning_rate": 2.665576292454519e-06,
|
|
"loss": 0.0656,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02375321462750435,
|
|
"step": 11365,
|
|
"valid_targets_mean": 3502.6,
|
|
"valid_targets_min": 923
|
|
},
|
|
{
|
|
"epoch": 5.9559979046621265,
|
|
"grad_norm": 0.7459787971829334,
|
|
"learning_rate": 2.6525609466411207e-06,
|
|
"loss": 0.1035,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051826655864715576,
|
|
"step": 11370,
|
|
"valid_targets_mean": 1553.9,
|
|
"valid_targets_min": 531
|
|
},
|
|
{
|
|
"epoch": 5.958617077003667,
|
|
"grad_norm": 0.5310568093938065,
|
|
"learning_rate": 2.6395751968035322e-06,
|
|
"loss": 0.0979,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06401866674423218,
|
|
"step": 11375,
|
|
"valid_targets_mean": 1825.6,
|
|
"valid_targets_min": 575
|
|
},
|
|
{
|
|
"epoch": 5.961236249345207,
|
|
"grad_norm": 0.4782124189664576,
|
|
"learning_rate": 2.6266190650963695e-06,
|
|
"loss": 0.0716,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038485944271087646,
|
|
"step": 11380,
|
|
"valid_targets_mean": 3701.9,
|
|
"valid_targets_min": 2784
|
|
},
|
|
{
|
|
"epoch": 5.9638554216867465,
|
|
"grad_norm": 0.4087593790616926,
|
|
"learning_rate": 2.6136925736237294e-06,
|
|
"loss": 0.0699,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03156688064336777,
|
|
"step": 11385,
|
|
"valid_targets_mean": 3575.8,
|
|
"valid_targets_min": 2525
|
|
},
|
|
{
|
|
"epoch": 5.966474594028287,
|
|
"grad_norm": 0.39190323540721084,
|
|
"learning_rate": 2.6007957444391285e-06,
|
|
"loss": 0.0683,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0255313441157341,
|
|
"step": 11390,
|
|
"valid_targets_mean": 3081.0,
|
|
"valid_targets_min": 970
|
|
},
|
|
{
|
|
"epoch": 5.969093766369827,
|
|
"grad_norm": 0.51699522650557,
|
|
"learning_rate": 2.5879285995454904e-06,
|
|
"loss": 0.0641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031044133007526398,
|
|
"step": 11395,
|
|
"valid_targets_mean": 1963.5,
|
|
"valid_targets_min": 817
|
|
},
|
|
{
|
|
"epoch": 5.971712938711367,
|
|
"grad_norm": 0.8952837747705978,
|
|
"learning_rate": 2.5750911608950756e-06,
|
|
"loss": 0.0837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05980469286441803,
|
|
"step": 11400,
|
|
"valid_targets_mean": 1270.9,
|
|
"valid_targets_min": 637
|
|
},
|
|
{
|
|
"epoch": 5.974332111052908,
|
|
"grad_norm": 0.4099808057463521,
|
|
"learning_rate": 2.562283450389489e-06,
|
|
"loss": 0.0621,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03177203610539436,
|
|
"step": 11405,
|
|
"valid_targets_mean": 2594.1,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 5.976951283394447,
|
|
"grad_norm": 0.40631750093059726,
|
|
"learning_rate": 2.549505489879589e-06,
|
|
"loss": 0.0653,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028754254803061485,
|
|
"step": 11410,
|
|
"valid_targets_mean": 3222.2,
|
|
"valid_targets_min": 1606
|
|
},
|
|
{
|
|
"epoch": 5.9795704557359874,
|
|
"grad_norm": 0.5522428400424914,
|
|
"learning_rate": 2.536757301165511e-06,
|
|
"loss": 0.0671,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028628377243876457,
|
|
"step": 11415,
|
|
"valid_targets_mean": 3023.8,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 5.982189628077528,
|
|
"grad_norm": 0.3924448115671155,
|
|
"learning_rate": 2.5240389059965642e-06,
|
|
"loss": 0.0753,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03105129860341549,
|
|
"step": 11420,
|
|
"valid_targets_mean": 4263.8,
|
|
"valid_targets_min": 1078
|
|
},
|
|
{
|
|
"epoch": 5.984808800419067,
|
|
"grad_norm": 0.38690813940980917,
|
|
"learning_rate": 2.5113503260712514e-06,
|
|
"loss": 0.0648,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03710411116480827,
|
|
"step": 11425,
|
|
"valid_targets_mean": 4150.0,
|
|
"valid_targets_min": 761
|
|
},
|
|
{
|
|
"epoch": 5.9874279727606075,
|
|
"grad_norm": 0.4826422485215377,
|
|
"learning_rate": 2.498691583037196e-06,
|
|
"loss": 0.0695,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0368012934923172,
|
|
"step": 11430,
|
|
"valid_targets_mean": 1368.4,
|
|
"valid_targets_min": 538
|
|
},
|
|
{
|
|
"epoch": 5.990047145102148,
|
|
"grad_norm": 0.519782763564192,
|
|
"learning_rate": 2.486062698491125e-06,
|
|
"loss": 0.0867,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07773230969905853,
|
|
"step": 11435,
|
|
"valid_targets_mean": 2630.4,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 5.992666317443688,
|
|
"grad_norm": 0.37054664090641287,
|
|
"learning_rate": 2.4734636939788216e-06,
|
|
"loss": 0.1409,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030455132946372032,
|
|
"step": 11440,
|
|
"valid_targets_mean": 3353.9,
|
|
"valid_targets_min": 2347
|
|
},
|
|
{
|
|
"epoch": 5.995285489785228,
|
|
"grad_norm": 0.4400819327957686,
|
|
"learning_rate": 2.460894590995091e-06,
|
|
"loss": 0.0703,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04020780324935913,
|
|
"step": 11445,
|
|
"valid_targets_mean": 3880.2,
|
|
"valid_targets_min": 2473
|
|
},
|
|
{
|
|
"epoch": 5.997904662126768,
|
|
"grad_norm": 0.3963239425973607,
|
|
"learning_rate": 2.448355410983723e-06,
|
|
"loss": 0.0676,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03425648808479309,
|
|
"step": 11450,
|
|
"valid_targets_mean": 3556.4,
|
|
"valid_targets_min": 2245
|
|
},
|
|
{
|
|
"epoch": 6.000523834468308,
|
|
"grad_norm": 0.9180097151052058,
|
|
"learning_rate": 2.4358461753374663e-06,
|
|
"loss": 0.1123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07151223719120026,
|
|
"step": 11455,
|
|
"valid_targets_mean": 1492.2,
|
|
"valid_targets_min": 544
|
|
},
|
|
{
|
|
"epoch": 6.003143006809848,
|
|
"grad_norm": 0.9599880015008919,
|
|
"learning_rate": 2.423366905397957e-06,
|
|
"loss": 0.1487,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06720415502786636,
|
|
"step": 11460,
|
|
"valid_targets_mean": 1270.4,
|
|
"valid_targets_min": 632
|
|
},
|
|
{
|
|
"epoch": 6.005762179151388,
|
|
"grad_norm": 1.1052017481465797,
|
|
"learning_rate": 2.4109176224557463e-06,
|
|
"loss": 0.1315,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055264897644519806,
|
|
"step": 11465,
|
|
"valid_targets_mean": 1358.4,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 6.008381351492928,
|
|
"grad_norm": 0.9280327830610237,
|
|
"learning_rate": 2.3984983477501866e-06,
|
|
"loss": 0.1336,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07304681837558746,
|
|
"step": 11470,
|
|
"valid_targets_mean": 1413.0,
|
|
"valid_targets_min": 701
|
|
},
|
|
{
|
|
"epoch": 6.011000523834468,
|
|
"grad_norm": 0.8440440917837605,
|
|
"learning_rate": 2.3861091024694673e-06,
|
|
"loss": 0.1224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07937653362751007,
|
|
"step": 11475,
|
|
"valid_targets_mean": 1745.6,
|
|
"valid_targets_min": 754
|
|
},
|
|
{
|
|
"epoch": 6.013619696176009,
|
|
"grad_norm": 0.7769512330057519,
|
|
"learning_rate": 2.373749907750518e-06,
|
|
"loss": 0.1288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06396303325891495,
|
|
"step": 11480,
|
|
"valid_targets_mean": 1337.9,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 6.016238868517548,
|
|
"grad_norm": 0.764887720106359,
|
|
"learning_rate": 2.361420784679016e-06,
|
|
"loss": 0.1312,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06879110634326935,
|
|
"step": 11485,
|
|
"valid_targets_mean": 1728.8,
|
|
"valid_targets_min": 970
|
|
},
|
|
{
|
|
"epoch": 6.018858040859088,
|
|
"grad_norm": 0.8047908910618264,
|
|
"learning_rate": 2.3491217542893274e-06,
|
|
"loss": 0.116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06783033907413483,
|
|
"step": 11490,
|
|
"valid_targets_mean": 1569.5,
|
|
"valid_targets_min": 648
|
|
},
|
|
{
|
|
"epoch": 6.021477213200629,
|
|
"grad_norm": 0.8550912758160515,
|
|
"learning_rate": 2.336852837564485e-06,
|
|
"loss": 0.1379,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07583297789096832,
|
|
"step": 11495,
|
|
"valid_targets_mean": 1538.8,
|
|
"valid_targets_min": 727
|
|
},
|
|
{
|
|
"epoch": 6.024096385542169,
|
|
"grad_norm": 0.7113626689321407,
|
|
"learning_rate": 2.3246140554361385e-06,
|
|
"loss": 0.1209,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04935118556022644,
|
|
"step": 11500,
|
|
"valid_targets_mean": 1519.5,
|
|
"valid_targets_min": 788
|
|
},
|
|
{
|
|
"epoch": 6.026715557883708,
|
|
"grad_norm": 0.7922064047732775,
|
|
"learning_rate": 2.312405428784532e-06,
|
|
"loss": 0.129,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054738614708185196,
|
|
"step": 11505,
|
|
"valid_targets_mean": 1325.1,
|
|
"valid_targets_min": 627
|
|
},
|
|
{
|
|
"epoch": 6.029334730225249,
|
|
"grad_norm": 0.7956421797633352,
|
|
"learning_rate": 2.3002269784384533e-06,
|
|
"loss": 0.1231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0749821588397026,
|
|
"step": 11510,
|
|
"valid_targets_mean": 1732.4,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 6.031953902566789,
|
|
"grad_norm": 0.8077100279931957,
|
|
"learning_rate": 2.2880787251752134e-06,
|
|
"loss": 0.1202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06252764910459518,
|
|
"step": 11515,
|
|
"valid_targets_mean": 1525.6,
|
|
"valid_targets_min": 923
|
|
},
|
|
{
|
|
"epoch": 6.034573074908329,
|
|
"grad_norm": 0.7804853727327673,
|
|
"learning_rate": 2.2759606897206043e-06,
|
|
"loss": 0.1186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06563736498355865,
|
|
"step": 11520,
|
|
"valid_targets_mean": 1603.5,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 6.037192247249869,
|
|
"grad_norm": 0.8622259356502991,
|
|
"learning_rate": 2.263872892748866e-06,
|
|
"loss": 0.1163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06406301259994507,
|
|
"step": 11525,
|
|
"valid_targets_mean": 1757.0,
|
|
"valid_targets_min": 895
|
|
},
|
|
{
|
|
"epoch": 6.039811419591409,
|
|
"grad_norm": 0.8644492816609076,
|
|
"learning_rate": 2.251815354882647e-06,
|
|
"loss": 0.1112,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05467421934008598,
|
|
"step": 11530,
|
|
"valid_targets_mean": 1391.5,
|
|
"valid_targets_min": 1005
|
|
},
|
|
{
|
|
"epoch": 6.042430591932949,
|
|
"grad_norm": 0.8034350215030263,
|
|
"learning_rate": 2.2397880966929718e-06,
|
|
"loss": 0.1281,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04740656539797783,
|
|
"step": 11535,
|
|
"valid_targets_mean": 1217.1,
|
|
"valid_targets_min": 870
|
|
},
|
|
{
|
|
"epoch": 6.04504976427449,
|
|
"grad_norm": 0.7982295042700137,
|
|
"learning_rate": 2.227791138699211e-06,
|
|
"loss": 0.123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06267914175987244,
|
|
"step": 11540,
|
|
"valid_targets_mean": 1619.1,
|
|
"valid_targets_min": 879
|
|
},
|
|
{
|
|
"epoch": 6.047668936616029,
|
|
"grad_norm": 0.7768829121969157,
|
|
"learning_rate": 2.2158245013690216e-06,
|
|
"loss": 0.1241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0555448904633522,
|
|
"step": 11545,
|
|
"valid_targets_mean": 1507.5,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 6.050288108957569,
|
|
"grad_norm": 0.8186437461505175,
|
|
"learning_rate": 2.2038882051183606e-06,
|
|
"loss": 0.1213,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07260645925998688,
|
|
"step": 11550,
|
|
"valid_targets_mean": 1707.1,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 6.05290728129911,
|
|
"grad_norm": 0.8297988799685673,
|
|
"learning_rate": 2.1919822703113945e-06,
|
|
"loss": 0.1235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06317691504955292,
|
|
"step": 11555,
|
|
"valid_targets_mean": 1747.9,
|
|
"valid_targets_min": 832
|
|
},
|
|
{
|
|
"epoch": 6.05552645364065,
|
|
"grad_norm": 0.7807605710807147,
|
|
"learning_rate": 2.180106717260515e-06,
|
|
"loss": 0.1191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05148398131132126,
|
|
"step": 11560,
|
|
"valid_targets_mean": 1581.9,
|
|
"valid_targets_min": 891
|
|
},
|
|
{
|
|
"epoch": 6.058145625982189,
|
|
"grad_norm": 0.9167811492382313,
|
|
"learning_rate": 2.168261566226255e-06,
|
|
"loss": 0.1216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07749298214912415,
|
|
"step": 11565,
|
|
"valid_targets_mean": 1746.8,
|
|
"valid_targets_min": 1052
|
|
},
|
|
{
|
|
"epoch": 6.06076479832373,
|
|
"grad_norm": 0.7718625888889671,
|
|
"learning_rate": 2.1564468374172986e-06,
|
|
"loss": 0.1118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05473483353853226,
|
|
"step": 11570,
|
|
"valid_targets_mean": 1469.5,
|
|
"valid_targets_min": 737
|
|
},
|
|
{
|
|
"epoch": 6.06338397066527,
|
|
"grad_norm": 0.7798331491197191,
|
|
"learning_rate": 2.144662550990422e-06,
|
|
"loss": 0.1174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06085578724741936,
|
|
"step": 11575,
|
|
"valid_targets_mean": 1762.2,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 6.06600314300681,
|
|
"grad_norm": 0.8299720107263232,
|
|
"learning_rate": 2.1329087270504577e-06,
|
|
"loss": 0.1138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06023997813463211,
|
|
"step": 11580,
|
|
"valid_targets_mean": 1482.1,
|
|
"valid_targets_min": 805
|
|
},
|
|
{
|
|
"epoch": 6.06862231534835,
|
|
"grad_norm": 0.7736914438669119,
|
|
"learning_rate": 2.1211853856502794e-06,
|
|
"loss": 0.1217,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06807412952184677,
|
|
"step": 11585,
|
|
"valid_targets_mean": 1962.8,
|
|
"valid_targets_min": 1253
|
|
},
|
|
{
|
|
"epoch": 6.07124148768989,
|
|
"grad_norm": 0.854125867360808,
|
|
"learning_rate": 2.1094925467907456e-06,
|
|
"loss": 0.1271,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07101461291313171,
|
|
"step": 11590,
|
|
"valid_targets_mean": 1958.0,
|
|
"valid_targets_min": 1264
|
|
},
|
|
{
|
|
"epoch": 6.07386066003143,
|
|
"grad_norm": 0.7946035281916202,
|
|
"learning_rate": 2.0978302304206766e-06,
|
|
"loss": 0.1187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04817923903465271,
|
|
"step": 11595,
|
|
"valid_targets_mean": 1201.4,
|
|
"valid_targets_min": 613
|
|
},
|
|
{
|
|
"epoch": 6.0764798323729705,
|
|
"grad_norm": 0.8165893535643656,
|
|
"learning_rate": 2.0861984564368297e-06,
|
|
"loss": 0.1207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061780575662851334,
|
|
"step": 11600,
|
|
"valid_targets_mean": 1559.0,
|
|
"valid_targets_min": 768
|
|
},
|
|
{
|
|
"epoch": 6.07909900471451,
|
|
"grad_norm": 1.2887836682489096,
|
|
"learning_rate": 2.0745972446838294e-06,
|
|
"loss": 0.1157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05495624244213104,
|
|
"step": 11605,
|
|
"valid_targets_mean": 1210.0,
|
|
"valid_targets_min": 714
|
|
},
|
|
{
|
|
"epoch": 6.08171817705605,
|
|
"grad_norm": 0.8710831103154718,
|
|
"learning_rate": 2.0630266149541934e-06,
|
|
"loss": 0.1249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04973271116614342,
|
|
"step": 11610,
|
|
"valid_targets_mean": 1242.2,
|
|
"valid_targets_min": 841
|
|
},
|
|
{
|
|
"epoch": 6.0843373493975905,
|
|
"grad_norm": 0.7665366475838662,
|
|
"learning_rate": 2.0514865869882295e-06,
|
|
"loss": 0.1176,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06063533201813698,
|
|
"step": 11615,
|
|
"valid_targets_mean": 1708.1,
|
|
"valid_targets_min": 936
|
|
},
|
|
{
|
|
"epoch": 6.086956521739131,
|
|
"grad_norm": 0.894116689823647,
|
|
"learning_rate": 2.039977180474071e-06,
|
|
"loss": 0.1159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05748577415943146,
|
|
"step": 11620,
|
|
"valid_targets_mean": 1163.1,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 6.08957569408067,
|
|
"grad_norm": 0.790949054699234,
|
|
"learning_rate": 2.0284984150475795e-06,
|
|
"loss": 0.1042,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05484342202544212,
|
|
"step": 11625,
|
|
"valid_targets_mean": 1380.4,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 6.0921948664222105,
|
|
"grad_norm": 0.7846854748574492,
|
|
"learning_rate": 2.0170503102923544e-06,
|
|
"loss": 0.1199,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0660267323255539,
|
|
"step": 11630,
|
|
"valid_targets_mean": 1625.4,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 6.094814038763751,
|
|
"grad_norm": 0.8299153408407629,
|
|
"learning_rate": 2.0056328857396856e-06,
|
|
"loss": 0.1146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04472474753856659,
|
|
"step": 11635,
|
|
"valid_targets_mean": 1153.4,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 6.097433211105291,
|
|
"grad_norm": 0.8316666817616704,
|
|
"learning_rate": 1.994246160868516e-06,
|
|
"loss": 0.1155,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06456252932548523,
|
|
"step": 11640,
|
|
"valid_targets_mean": 1487.1,
|
|
"valid_targets_min": 831
|
|
},
|
|
{
|
|
"epoch": 6.1000523834468305,
|
|
"grad_norm": 0.7880769174254484,
|
|
"learning_rate": 1.9828901551054193e-06,
|
|
"loss": 0.1151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05208251625299454,
|
|
"step": 11645,
|
|
"valid_targets_mean": 1479.1,
|
|
"valid_targets_min": 1015
|
|
},
|
|
{
|
|
"epoch": 6.102671555788371,
|
|
"grad_norm": 0.7833005697941994,
|
|
"learning_rate": 1.9715648878245532e-06,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05049904063344002,
|
|
"step": 11650,
|
|
"valid_targets_mean": 1366.5,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 6.105290728129911,
|
|
"grad_norm": 0.7489628605619162,
|
|
"learning_rate": 1.960270378347642e-06,
|
|
"loss": 0.1135,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042261186987161636,
|
|
"step": 11655,
|
|
"valid_targets_mean": 1318.1,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 6.107909900471451,
|
|
"grad_norm": 0.9224735642475472,
|
|
"learning_rate": 1.9490066459439184e-06,
|
|
"loss": 0.1178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07317248731851578,
|
|
"step": 11660,
|
|
"valid_targets_mean": 1541.2,
|
|
"valid_targets_min": 931
|
|
},
|
|
{
|
|
"epoch": 6.110529072812991,
|
|
"grad_norm": 0.7834065543973477,
|
|
"learning_rate": 1.9377737098301263e-06,
|
|
"loss": 0.1222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06088053807616234,
|
|
"step": 11665,
|
|
"valid_targets_mean": 1790.2,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 6.113148245154531,
|
|
"grad_norm": 0.8178539351959183,
|
|
"learning_rate": 1.9265715891704583e-06,
|
|
"loss": 0.1241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06058657169342041,
|
|
"step": 11670,
|
|
"valid_targets_mean": 1511.6,
|
|
"valid_targets_min": 748
|
|
},
|
|
{
|
|
"epoch": 6.1157674174960714,
|
|
"grad_norm": 0.8386374996628507,
|
|
"learning_rate": 1.9154003030765377e-06,
|
|
"loss": 0.113,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058460816740989685,
|
|
"step": 11675,
|
|
"valid_targets_mean": 1568.1,
|
|
"valid_targets_min": 699
|
|
},
|
|
{
|
|
"epoch": 6.118386589837612,
|
|
"grad_norm": 0.8067359342212168,
|
|
"learning_rate": 1.9042598706073767e-06,
|
|
"loss": 0.1065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038479067385196686,
|
|
"step": 11680,
|
|
"valid_targets_mean": 1002.4,
|
|
"valid_targets_min": 534
|
|
},
|
|
{
|
|
"epoch": 6.121005762179151,
|
|
"grad_norm": 0.8310382170192431,
|
|
"learning_rate": 1.8931503107693605e-06,
|
|
"loss": 0.121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05678606778383255,
|
|
"step": 11685,
|
|
"valid_targets_mean": 1367.1,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 6.1236249345206915,
|
|
"grad_norm": 0.8446260002301718,
|
|
"learning_rate": 1.8820716425161812e-06,
|
|
"loss": 0.1132,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04868333414196968,
|
|
"step": 11690,
|
|
"valid_targets_mean": 1259.8,
|
|
"valid_targets_min": 835
|
|
},
|
|
{
|
|
"epoch": 6.126244106862232,
|
|
"grad_norm": 0.834487265817137,
|
|
"learning_rate": 1.8710238847488572e-06,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05200199782848358,
|
|
"step": 11695,
|
|
"valid_targets_mean": 1330.9,
|
|
"valid_targets_min": 829
|
|
},
|
|
{
|
|
"epoch": 6.128863279203772,
|
|
"grad_norm": 0.7950437491710232,
|
|
"learning_rate": 1.8600070563156425e-06,
|
|
"loss": 0.1074,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06211322546005249,
|
|
"step": 11700,
|
|
"valid_targets_mean": 1756.4,
|
|
"valid_targets_min": 1138
|
|
},
|
|
{
|
|
"epoch": 6.1314824515453115,
|
|
"grad_norm": 0.8023437380624452,
|
|
"learning_rate": 1.8490211760120513e-06,
|
|
"loss": 0.1181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0646124929189682,
|
|
"step": 11705,
|
|
"valid_targets_mean": 1799.8,
|
|
"valid_targets_min": 1174
|
|
},
|
|
{
|
|
"epoch": 6.134101623886852,
|
|
"grad_norm": 1.5271236396896737,
|
|
"learning_rate": 1.8380662625807665e-06,
|
|
"loss": 0.1155,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057662054896354675,
|
|
"step": 11710,
|
|
"valid_targets_mean": 1524.5,
|
|
"valid_targets_min": 554
|
|
},
|
|
{
|
|
"epoch": 6.136720796228392,
|
|
"grad_norm": 0.7862694098761159,
|
|
"learning_rate": 1.8271423347116757e-06,
|
|
"loss": 0.1119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05381157621741295,
|
|
"step": 11715,
|
|
"valid_targets_mean": 1367.2,
|
|
"valid_targets_min": 843
|
|
},
|
|
{
|
|
"epoch": 6.1393399685699315,
|
|
"grad_norm": 0.8073338604081094,
|
|
"learning_rate": 1.8162494110417683e-06,
|
|
"loss": 0.1106,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05907204747200012,
|
|
"step": 11720,
|
|
"valid_targets_mean": 1673.1,
|
|
"valid_targets_min": 849
|
|
},
|
|
{
|
|
"epoch": 6.141959140911472,
|
|
"grad_norm": 0.8167808150540459,
|
|
"learning_rate": 1.805387510155161e-06,
|
|
"loss": 0.1171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05595570057630539,
|
|
"step": 11725,
|
|
"valid_targets_mean": 1509.0,
|
|
"valid_targets_min": 612
|
|
},
|
|
{
|
|
"epoch": 6.144578313253012,
|
|
"grad_norm": 0.856293701424457,
|
|
"learning_rate": 1.7945566505830325e-06,
|
|
"loss": 0.1175,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05260946974158287,
|
|
"step": 11730,
|
|
"valid_targets_mean": 1401.0,
|
|
"valid_targets_min": 841
|
|
},
|
|
{
|
|
"epoch": 6.147197485594552,
|
|
"grad_norm": 0.8906531417243976,
|
|
"learning_rate": 1.7837568508036108e-06,
|
|
"loss": 0.1158,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06653806567192078,
|
|
"step": 11735,
|
|
"valid_targets_mean": 1564.8,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 6.149816657936092,
|
|
"grad_norm": 0.8176579226961042,
|
|
"learning_rate": 1.7729881292421236e-06,
|
|
"loss": 0.1118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053530290722846985,
|
|
"step": 11740,
|
|
"valid_targets_mean": 1293.1,
|
|
"valid_targets_min": 822
|
|
},
|
|
{
|
|
"epoch": 6.152435830277632,
|
|
"grad_norm": 0.7511511259946063,
|
|
"learning_rate": 1.7622505042707927e-06,
|
|
"loss": 0.1029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050591953098773956,
|
|
"step": 11745,
|
|
"valid_targets_mean": 1731.0,
|
|
"valid_targets_min": 670
|
|
},
|
|
{
|
|
"epoch": 6.155055002619172,
|
|
"grad_norm": 0.8166327313618861,
|
|
"learning_rate": 1.7515439942087597e-06,
|
|
"loss": 0.1177,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0659506767988205,
|
|
"step": 11750,
|
|
"valid_targets_mean": 1624.1,
|
|
"valid_targets_min": 647
|
|
},
|
|
{
|
|
"epoch": 6.157674174960713,
|
|
"grad_norm": 0.9273321820099081,
|
|
"learning_rate": 1.7408686173221179e-06,
|
|
"loss": 0.1177,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04893888160586357,
|
|
"step": 11755,
|
|
"valid_targets_mean": 1429.1,
|
|
"valid_targets_min": 952
|
|
},
|
|
{
|
|
"epoch": 6.160293347302252,
|
|
"grad_norm": 0.9048561020505893,
|
|
"learning_rate": 1.7302243918238137e-06,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04985303804278374,
|
|
"step": 11760,
|
|
"valid_targets_mean": 1222.5,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 6.162912519643792,
|
|
"grad_norm": 0.813391650304158,
|
|
"learning_rate": 1.7196113358736655e-06,
|
|
"loss": 0.1141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0514526404440403,
|
|
"step": 11765,
|
|
"valid_targets_mean": 1417.8,
|
|
"valid_targets_min": 892
|
|
},
|
|
{
|
|
"epoch": 6.165531691985333,
|
|
"grad_norm": 0.8100378908596324,
|
|
"learning_rate": 1.7090294675783026e-06,
|
|
"loss": 0.1189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06349000334739685,
|
|
"step": 11770,
|
|
"valid_targets_mean": 1636.0,
|
|
"valid_targets_min": 985
|
|
},
|
|
{
|
|
"epoch": 6.168150864326873,
|
|
"grad_norm": 0.8672363204516796,
|
|
"learning_rate": 1.6984788049911572e-06,
|
|
"loss": 0.1139,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054531119763851166,
|
|
"step": 11775,
|
|
"valid_targets_mean": 1505.0,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 6.170770036668412,
|
|
"grad_norm": 0.8618687495589307,
|
|
"learning_rate": 1.687959366112415e-06,
|
|
"loss": 0.112,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04755813628435135,
|
|
"step": 11780,
|
|
"valid_targets_mean": 1286.5,
|
|
"valid_targets_min": 641
|
|
},
|
|
{
|
|
"epoch": 6.173389209009953,
|
|
"grad_norm": 0.7778918855830956,
|
|
"learning_rate": 1.6774711688889933e-06,
|
|
"loss": 0.1137,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06086455285549164,
|
|
"step": 11785,
|
|
"valid_targets_mean": 1567.5,
|
|
"valid_targets_min": 1020
|
|
},
|
|
{
|
|
"epoch": 6.176008381351493,
|
|
"grad_norm": 0.7632153150912035,
|
|
"learning_rate": 1.667014231214512e-06,
|
|
"loss": 0.1095,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.048788003623485565,
|
|
"step": 11790,
|
|
"valid_targets_mean": 1252.4,
|
|
"valid_targets_min": 729
|
|
},
|
|
{
|
|
"epoch": 6.178627553693033,
|
|
"grad_norm": 0.8316843161980305,
|
|
"learning_rate": 1.6565885709292496e-06,
|
|
"loss": 0.1194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06742323935031891,
|
|
"step": 11795,
|
|
"valid_targets_mean": 1518.2,
|
|
"valid_targets_min": 682
|
|
},
|
|
{
|
|
"epoch": 6.181246726034573,
|
|
"grad_norm": 0.8169650561144044,
|
|
"learning_rate": 1.6461942058201397e-06,
|
|
"loss": 0.1197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0598052479326725,
|
|
"step": 11800,
|
|
"valid_targets_mean": 1619.2,
|
|
"valid_targets_min": 1260
|
|
},
|
|
{
|
|
"epoch": 6.183865898376113,
|
|
"grad_norm": 0.8353797928606886,
|
|
"learning_rate": 1.6358311536207084e-06,
|
|
"loss": 0.1151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06587246060371399,
|
|
"step": 11805,
|
|
"valid_targets_mean": 1575.4,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 6.186485070717653,
|
|
"grad_norm": 0.8979736643597953,
|
|
"learning_rate": 1.6254994320110707e-06,
|
|
"loss": 0.1149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06438019871711731,
|
|
"step": 11810,
|
|
"valid_targets_mean": 1608.1,
|
|
"valid_targets_min": 801
|
|
},
|
|
{
|
|
"epoch": 6.189104243059194,
|
|
"grad_norm": 0.7569834562634008,
|
|
"learning_rate": 1.61519905861788e-06,
|
|
"loss": 0.1072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055120810866355896,
|
|
"step": 11815,
|
|
"valid_targets_mean": 1414.8,
|
|
"valid_targets_min": 564
|
|
},
|
|
{
|
|
"epoch": 6.191723415400733,
|
|
"grad_norm": 0.8072629139008018,
|
|
"learning_rate": 1.6049300510143173e-06,
|
|
"loss": 0.1085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056000012904405594,
|
|
"step": 11820,
|
|
"valid_targets_mean": 1512.5,
|
|
"valid_targets_min": 651
|
|
},
|
|
{
|
|
"epoch": 6.194342587742273,
|
|
"grad_norm": 0.8137590485392445,
|
|
"learning_rate": 1.5946924267200437e-06,
|
|
"loss": 0.1154,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04661045968532562,
|
|
"step": 11825,
|
|
"valid_targets_mean": 1262.4,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 6.196961760083814,
|
|
"grad_norm": 0.7701490102572036,
|
|
"learning_rate": 1.584486203201181e-06,
|
|
"loss": 0.1075,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04992928355932236,
|
|
"step": 11830,
|
|
"valid_targets_mean": 1230.0,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 6.199580932425354,
|
|
"grad_norm": 0.7514466436217656,
|
|
"learning_rate": 1.5743113978702807e-06,
|
|
"loss": 0.104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05396423488855362,
|
|
"step": 11835,
|
|
"valid_targets_mean": 1497.5,
|
|
"valid_targets_min": 1251
|
|
},
|
|
{
|
|
"epoch": 6.202200104766893,
|
|
"grad_norm": 0.8670940282351024,
|
|
"learning_rate": 1.5641680280862947e-06,
|
|
"loss": 0.1112,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056431740522384644,
|
|
"step": 11840,
|
|
"valid_targets_mean": 1470.8,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 6.204819277108434,
|
|
"grad_norm": 0.9243653707102031,
|
|
"learning_rate": 1.5540561111545272e-06,
|
|
"loss": 0.1194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06075423210859299,
|
|
"step": 11845,
|
|
"valid_targets_mean": 1469.4,
|
|
"valid_targets_min": 846
|
|
},
|
|
{
|
|
"epoch": 6.207438449449974,
|
|
"grad_norm": 0.8745717321879108,
|
|
"learning_rate": 1.5439756643266468e-06,
|
|
"loss": 0.107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06556703895330429,
|
|
"step": 11850,
|
|
"valid_targets_mean": 1774.4,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 6.210057621791514,
|
|
"grad_norm": 0.7920889225831382,
|
|
"learning_rate": 1.5339267048006078e-06,
|
|
"loss": 0.1129,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05082762986421585,
|
|
"step": 11855,
|
|
"valid_targets_mean": 1558.9,
|
|
"valid_targets_min": 823
|
|
},
|
|
{
|
|
"epoch": 6.212676794133054,
|
|
"grad_norm": 0.7701394032471317,
|
|
"learning_rate": 1.5239092497206697e-06,
|
|
"loss": 0.0991,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05166442319750786,
|
|
"step": 11860,
|
|
"valid_targets_mean": 1687.6,
|
|
"valid_targets_min": 595
|
|
},
|
|
{
|
|
"epoch": 6.215295966474594,
|
|
"grad_norm": 0.9627179393614519,
|
|
"learning_rate": 1.5139233161773193e-06,
|
|
"loss": 0.1085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05256475508213043,
|
|
"step": 11865,
|
|
"valid_targets_mean": 1336.9,
|
|
"valid_targets_min": 868
|
|
},
|
|
{
|
|
"epoch": 6.217915138816134,
|
|
"grad_norm": 0.8767320830018093,
|
|
"learning_rate": 1.5039689212072772e-06,
|
|
"loss": 0.1062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05469423532485962,
|
|
"step": 11870,
|
|
"valid_targets_mean": 1266.6,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 6.2205343111576745,
|
|
"grad_norm": 0.8010639984469596,
|
|
"learning_rate": 1.4940460817934587e-06,
|
|
"loss": 0.1114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06903129816055298,
|
|
"step": 11875,
|
|
"valid_targets_mean": 1976.8,
|
|
"valid_targets_min": 1080
|
|
},
|
|
{
|
|
"epoch": 6.223153483499214,
|
|
"grad_norm": 0.8188843415246827,
|
|
"learning_rate": 1.4841548148649376e-06,
|
|
"loss": 0.1104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061441607773303986,
|
|
"step": 11880,
|
|
"valid_targets_mean": 1554.1,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 6.225772655840754,
|
|
"grad_norm": 0.7992562221434741,
|
|
"learning_rate": 1.4742951372969262e-06,
|
|
"loss": 0.1131,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05371711403131485,
|
|
"step": 11885,
|
|
"valid_targets_mean": 1499.6,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 6.2283918281822945,
|
|
"grad_norm": 0.8391318946920864,
|
|
"learning_rate": 1.4644670659107419e-06,
|
|
"loss": 0.1117,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0553358718752861,
|
|
"step": 11890,
|
|
"valid_targets_mean": 1475.5,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 6.231011000523835,
|
|
"grad_norm": 0.9655110948032307,
|
|
"learning_rate": 1.454670617473779e-06,
|
|
"loss": 0.1288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08838294446468353,
|
|
"step": 11895,
|
|
"valid_targets_mean": 7107.2,
|
|
"valid_targets_min": 5295
|
|
},
|
|
{
|
|
"epoch": 6.233630172865374,
|
|
"grad_norm": 0.8958472755472283,
|
|
"learning_rate": 1.4449058086994839e-06,
|
|
"loss": 0.1636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08870230615139008,
|
|
"step": 11900,
|
|
"valid_targets_mean": 7098.2,
|
|
"valid_targets_min": 5066
|
|
},
|
|
{
|
|
"epoch": 6.2362493452069145,
|
|
"grad_norm": 0.8201353608539739,
|
|
"learning_rate": 1.4351726562473166e-06,
|
|
"loss": 0.1602,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07592824101448059,
|
|
"step": 11905,
|
|
"valid_targets_mean": 6838.1,
|
|
"valid_targets_min": 4923
|
|
},
|
|
{
|
|
"epoch": 6.238868517548455,
|
|
"grad_norm": 0.7182641005412113,
|
|
"learning_rate": 1.4254711767227347e-06,
|
|
"loss": 0.1529,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07479575276374817,
|
|
"step": 11910,
|
|
"valid_targets_mean": 5701.2,
|
|
"valid_targets_min": 3922
|
|
},
|
|
{
|
|
"epoch": 6.241487689889995,
|
|
"grad_norm": 0.6340844273775583,
|
|
"learning_rate": 1.4158013866771624e-06,
|
|
"loss": 0.1498,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07615907490253448,
|
|
"step": 11915,
|
|
"valid_targets_mean": 7340.9,
|
|
"valid_targets_min": 4797
|
|
},
|
|
{
|
|
"epoch": 6.2441068622315345,
|
|
"grad_norm": 0.6199312834660536,
|
|
"learning_rate": 1.4061633026079501e-06,
|
|
"loss": 0.1651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08893653750419617,
|
|
"step": 11920,
|
|
"valid_targets_mean": 6343.9,
|
|
"valid_targets_min": 3853
|
|
},
|
|
{
|
|
"epoch": 6.246726034573075,
|
|
"grad_norm": 0.5247604131189251,
|
|
"learning_rate": 1.3965569409583646e-06,
|
|
"loss": 0.1533,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0741259753704071,
|
|
"step": 11925,
|
|
"valid_targets_mean": 6566.2,
|
|
"valid_targets_min": 4586
|
|
},
|
|
{
|
|
"epoch": 6.249345206914615,
|
|
"grad_norm": 0.507275471084232,
|
|
"learning_rate": 1.3869823181175468e-06,
|
|
"loss": 0.1461,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08363064378499985,
|
|
"step": 11930,
|
|
"valid_targets_mean": 7153.6,
|
|
"valid_targets_min": 4728
|
|
},
|
|
{
|
|
"epoch": 6.2519643792561554,
|
|
"grad_norm": 0.6009936005906976,
|
|
"learning_rate": 1.3774394504204925e-06,
|
|
"loss": 0.172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12445858865976334,
|
|
"step": 11935,
|
|
"valid_targets_mean": 6562.0,
|
|
"valid_targets_min": 5203
|
|
},
|
|
{
|
|
"epoch": 6.254583551597695,
|
|
"grad_norm": 0.45766553236629404,
|
|
"learning_rate": 1.3679283541480116e-06,
|
|
"loss": 0.1423,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07799802720546722,
|
|
"step": 11940,
|
|
"valid_targets_mean": 7098.9,
|
|
"valid_targets_min": 4561
|
|
},
|
|
{
|
|
"epoch": 6.257202723939235,
|
|
"grad_norm": 0.524128723453135,
|
|
"learning_rate": 1.3584490455267263e-06,
|
|
"loss": 0.1578,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1061946228146553,
|
|
"step": 11945,
|
|
"valid_targets_mean": 4960.2,
|
|
"valid_targets_min": 3439
|
|
},
|
|
{
|
|
"epoch": 6.2598218962807755,
|
|
"grad_norm": 0.4413781153928334,
|
|
"learning_rate": 1.349001540729009e-06,
|
|
"loss": 0.143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06203271076083183,
|
|
"step": 11950,
|
|
"valid_targets_mean": 5384.2,
|
|
"valid_targets_min": 3635
|
|
},
|
|
{
|
|
"epoch": 6.262441068622316,
|
|
"grad_norm": 0.44634902795119863,
|
|
"learning_rate": 1.339585855872989e-06,
|
|
"loss": 0.1499,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0751538872718811,
|
|
"step": 11955,
|
|
"valid_targets_mean": 6940.2,
|
|
"valid_targets_min": 4393
|
|
},
|
|
{
|
|
"epoch": 6.265060240963855,
|
|
"grad_norm": 0.43672764749518056,
|
|
"learning_rate": 1.3302020070224942e-06,
|
|
"loss": 0.1393,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07703030109405518,
|
|
"step": 11960,
|
|
"valid_targets_mean": 6812.4,
|
|
"valid_targets_min": 4749
|
|
},
|
|
{
|
|
"epoch": 6.2676794133053955,
|
|
"grad_norm": 0.48392158271644026,
|
|
"learning_rate": 1.3208500101870446e-06,
|
|
"loss": 0.1364,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029781676828861237,
|
|
"step": 11965,
|
|
"valid_targets_mean": 2363.0,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 6.270298585646936,
|
|
"grad_norm": 0.4393216070908655,
|
|
"learning_rate": 1.3115298813218201e-06,
|
|
"loss": 0.1372,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06820503622293472,
|
|
"step": 11970,
|
|
"valid_targets_mean": 5841.2,
|
|
"valid_targets_min": 3119
|
|
},
|
|
{
|
|
"epoch": 6.272917757988475,
|
|
"grad_norm": 0.4006964994183144,
|
|
"learning_rate": 1.302241636327628e-06,
|
|
"loss": 0.1475,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08046795427799225,
|
|
"step": 11975,
|
|
"valid_targets_mean": 5761.4,
|
|
"valid_targets_min": 3774
|
|
},
|
|
{
|
|
"epoch": 6.2755369303300155,
|
|
"grad_norm": 0.4525483967505272,
|
|
"learning_rate": 1.2929852910508811e-06,
|
|
"loss": 0.1439,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07478667795658112,
|
|
"step": 11980,
|
|
"valid_targets_mean": 6412.4,
|
|
"valid_targets_min": 5293
|
|
},
|
|
{
|
|
"epoch": 6.278156102671556,
|
|
"grad_norm": 0.40466175114045916,
|
|
"learning_rate": 1.2837608612835717e-06,
|
|
"loss": 0.1295,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07089799642562866,
|
|
"step": 11985,
|
|
"valid_targets_mean": 6709.1,
|
|
"valid_targets_min": 5665
|
|
},
|
|
{
|
|
"epoch": 6.280775275013096,
|
|
"grad_norm": 0.433406964654351,
|
|
"learning_rate": 1.274568362763231e-06,
|
|
"loss": 0.1363,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07706020772457123,
|
|
"step": 11990,
|
|
"valid_targets_mean": 7344.6,
|
|
"valid_targets_min": 4428
|
|
},
|
|
{
|
|
"epoch": 6.283394447354636,
|
|
"grad_norm": 0.424450344960714,
|
|
"learning_rate": 1.2654078111729318e-06,
|
|
"loss": 0.1549,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07545164227485657,
|
|
"step": 11995,
|
|
"valid_targets_mean": 5780.6,
|
|
"valid_targets_min": 4690
|
|
},
|
|
{
|
|
"epoch": 6.286013619696176,
|
|
"grad_norm": 0.391086310864684,
|
|
"learning_rate": 1.256279222141219e-06,
|
|
"loss": 0.1552,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06854374706745148,
|
|
"step": 12000,
|
|
"valid_targets_mean": 5854.5,
|
|
"valid_targets_min": 4335
|
|
},
|
|
{
|
|
"epoch": 6.288632792037716,
|
|
"grad_norm": 0.4395805926833709,
|
|
"learning_rate": 1.2471826112421348e-06,
|
|
"loss": 0.1385,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06968638300895691,
|
|
"step": 12005,
|
|
"valid_targets_mean": 6378.9,
|
|
"valid_targets_min": 5342
|
|
},
|
|
{
|
|
"epoch": 6.291251964379256,
|
|
"grad_norm": 0.44846829068853294,
|
|
"learning_rate": 1.238117993995136e-06,
|
|
"loss": 0.1459,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06752711534500122,
|
|
"step": 12010,
|
|
"valid_targets_mean": 5756.5,
|
|
"valid_targets_min": 3840
|
|
},
|
|
{
|
|
"epoch": 6.293871136720796,
|
|
"grad_norm": 0.44086127341386794,
|
|
"learning_rate": 1.2290853858651252e-06,
|
|
"loss": 0.1596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08626071363687515,
|
|
"step": 12015,
|
|
"valid_targets_mean": 6438.6,
|
|
"valid_targets_min": 5104
|
|
},
|
|
{
|
|
"epoch": 6.296490309062336,
|
|
"grad_norm": 0.4560999311130661,
|
|
"learning_rate": 1.220084802262369e-06,
|
|
"loss": 0.176,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06795895099639893,
|
|
"step": 12020,
|
|
"valid_targets_mean": 5190.4,
|
|
"valid_targets_min": 4224
|
|
},
|
|
{
|
|
"epoch": 6.299109481403876,
|
|
"grad_norm": 0.43551584888346606,
|
|
"learning_rate": 1.211116258542513e-06,
|
|
"loss": 0.1651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07206246256828308,
|
|
"step": 12025,
|
|
"valid_targets_mean": 5932.8,
|
|
"valid_targets_min": 4318
|
|
},
|
|
{
|
|
"epoch": 6.301728653745417,
|
|
"grad_norm": 0.41431914162984274,
|
|
"learning_rate": 1.202179770006533e-06,
|
|
"loss": 0.147,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07387147843837738,
|
|
"step": 12030,
|
|
"valid_targets_mean": 6348.5,
|
|
"valid_targets_min": 4783
|
|
},
|
|
{
|
|
"epoch": 6.304347826086957,
|
|
"grad_norm": 0.45201928195808644,
|
|
"learning_rate": 1.1932753519007246e-06,
|
|
"loss": 0.1485,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0748416930437088,
|
|
"step": 12035,
|
|
"valid_targets_mean": 7200.1,
|
|
"valid_targets_min": 4412
|
|
},
|
|
{
|
|
"epoch": 6.306966998428496,
|
|
"grad_norm": 0.45895603504767424,
|
|
"learning_rate": 1.18440301941666e-06,
|
|
"loss": 0.1551,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08404980599880219,
|
|
"step": 12040,
|
|
"valid_targets_mean": 6137.5,
|
|
"valid_targets_min": 5033
|
|
},
|
|
{
|
|
"epoch": 6.309586170770037,
|
|
"grad_norm": 0.40531811665772616,
|
|
"learning_rate": 1.1755627876911823e-06,
|
|
"loss": 0.1464,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06787222623825073,
|
|
"step": 12045,
|
|
"valid_targets_mean": 5695.5,
|
|
"valid_targets_min": 4822
|
|
},
|
|
{
|
|
"epoch": 6.312205343111577,
|
|
"grad_norm": 0.4505022302270149,
|
|
"learning_rate": 1.1667546718063493e-06,
|
|
"loss": 0.1497,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07829370349645615,
|
|
"step": 12050,
|
|
"valid_targets_mean": 5906.1,
|
|
"valid_targets_min": 2364
|
|
},
|
|
{
|
|
"epoch": 6.314824515453116,
|
|
"grad_norm": 0.4518121518021245,
|
|
"learning_rate": 1.1579786867894471e-06,
|
|
"loss": 0.1574,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08290734142065048,
|
|
"step": 12055,
|
|
"valid_targets_mean": 6359.0,
|
|
"valid_targets_min": 4793
|
|
},
|
|
{
|
|
"epoch": 6.317443687794657,
|
|
"grad_norm": 0.4299181018801878,
|
|
"learning_rate": 1.1492348476129323e-06,
|
|
"loss": 0.1495,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08029110729694366,
|
|
"step": 12060,
|
|
"valid_targets_mean": 6104.9,
|
|
"valid_targets_min": 4733
|
|
},
|
|
{
|
|
"epoch": 6.320062860136197,
|
|
"grad_norm": 0.3650039174519398,
|
|
"learning_rate": 1.1405231691944252e-06,
|
|
"loss": 0.1367,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06358305364847183,
|
|
"step": 12065,
|
|
"valid_targets_mean": 6780.8,
|
|
"valid_targets_min": 5388
|
|
},
|
|
{
|
|
"epoch": 6.322682032477737,
|
|
"grad_norm": 0.40302057029858435,
|
|
"learning_rate": 1.1318436663966703e-06,
|
|
"loss": 0.133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07508301734924316,
|
|
"step": 12070,
|
|
"valid_targets_mean": 5956.2,
|
|
"valid_targets_min": 3182
|
|
},
|
|
{
|
|
"epoch": 6.325301204819277,
|
|
"grad_norm": 0.4250736175157111,
|
|
"learning_rate": 1.123196354027527e-06,
|
|
"loss": 0.1552,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08265276253223419,
|
|
"step": 12075,
|
|
"valid_targets_mean": 6389.4,
|
|
"valid_targets_min": 5077
|
|
},
|
|
{
|
|
"epoch": 6.327920377160817,
|
|
"grad_norm": 0.45821405058257997,
|
|
"learning_rate": 1.1145812468399274e-06,
|
|
"loss": 0.1536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08668242394924164,
|
|
"step": 12080,
|
|
"valid_targets_mean": 5607.8,
|
|
"valid_targets_min": 3663
|
|
},
|
|
{
|
|
"epoch": 6.330539549502357,
|
|
"grad_norm": 0.4249903395916202,
|
|
"learning_rate": 1.1059983595318613e-06,
|
|
"loss": 0.1438,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06720085442066193,
|
|
"step": 12085,
|
|
"valid_targets_mean": 5616.4,
|
|
"valid_targets_min": 4105
|
|
},
|
|
{
|
|
"epoch": 6.333158721843898,
|
|
"grad_norm": 0.3896070204803804,
|
|
"learning_rate": 1.0974477067463552e-06,
|
|
"loss": 0.1446,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0791177898645401,
|
|
"step": 12090,
|
|
"valid_targets_mean": 7271.0,
|
|
"valid_targets_min": 5201
|
|
},
|
|
{
|
|
"epoch": 6.335777894185437,
|
|
"grad_norm": 0.421885075677326,
|
|
"learning_rate": 1.0889293030714288e-06,
|
|
"loss": 0.1361,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07875902950763702,
|
|
"step": 12095,
|
|
"valid_targets_mean": 6797.6,
|
|
"valid_targets_min": 5069
|
|
},
|
|
{
|
|
"epoch": 6.338397066526977,
|
|
"grad_norm": 0.4143888260982762,
|
|
"learning_rate": 1.0804431630401013e-06,
|
|
"loss": 0.145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07206882536411285,
|
|
"step": 12100,
|
|
"valid_targets_mean": 5787.4,
|
|
"valid_targets_min": 4729
|
|
},
|
|
{
|
|
"epoch": 6.341016238868518,
|
|
"grad_norm": 0.4447957567250799,
|
|
"learning_rate": 1.071989301130325e-06,
|
|
"loss": 0.1525,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0739193931221962,
|
|
"step": 12105,
|
|
"valid_targets_mean": 5989.0,
|
|
"valid_targets_min": 4567
|
|
},
|
|
{
|
|
"epoch": 6.343635411210058,
|
|
"grad_norm": 0.4116999865777004,
|
|
"learning_rate": 1.0635677317650006e-06,
|
|
"loss": 0.1387,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06551603227853775,
|
|
"step": 12110,
|
|
"valid_targets_mean": 5873.4,
|
|
"valid_targets_min": 4464
|
|
},
|
|
{
|
|
"epoch": 6.346254583551597,
|
|
"grad_norm": 0.4261393249437143,
|
|
"learning_rate": 1.0551784693119281e-06,
|
|
"loss": 0.1549,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07234413921833038,
|
|
"step": 12115,
|
|
"valid_targets_mean": 5373.8,
|
|
"valid_targets_min": 4908
|
|
},
|
|
{
|
|
"epoch": 6.348873755893138,
|
|
"grad_norm": 0.3769429712793793,
|
|
"learning_rate": 1.0468215280837945e-06,
|
|
"loss": 0.1367,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059860795736312866,
|
|
"step": 12120,
|
|
"valid_targets_mean": 5441.6,
|
|
"valid_targets_min": 4443
|
|
},
|
|
{
|
|
"epoch": 6.351492928234678,
|
|
"grad_norm": 0.3943897450855496,
|
|
"learning_rate": 1.038496922338137e-06,
|
|
"loss": 0.1258,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0633818507194519,
|
|
"step": 12125,
|
|
"valid_targets_mean": 6286.8,
|
|
"valid_targets_min": 3381
|
|
},
|
|
{
|
|
"epoch": 6.354112100576218,
|
|
"grad_norm": 0.41650678916559775,
|
|
"learning_rate": 1.0302046662773301e-06,
|
|
"loss": 0.1293,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06222138553857803,
|
|
"step": 12130,
|
|
"valid_targets_mean": 6162.8,
|
|
"valid_targets_min": 4790
|
|
},
|
|
{
|
|
"epoch": 6.356731272917758,
|
|
"grad_norm": 0.3995135580856778,
|
|
"learning_rate": 1.0219447740485622e-06,
|
|
"loss": 0.1321,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06467224657535553,
|
|
"step": 12135,
|
|
"valid_targets_mean": 6336.1,
|
|
"valid_targets_min": 5161
|
|
},
|
|
{
|
|
"epoch": 6.359350445259298,
|
|
"grad_norm": 0.4200816063818399,
|
|
"learning_rate": 1.0137172597438006e-06,
|
|
"loss": 0.1472,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06530135124921799,
|
|
"step": 12140,
|
|
"valid_targets_mean": 6038.0,
|
|
"valid_targets_min": 4491
|
|
},
|
|
{
|
|
"epoch": 6.361969617600838,
|
|
"grad_norm": 0.6366831964221076,
|
|
"learning_rate": 1.0055221373997704e-06,
|
|
"loss": 0.1752,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11287716031074524,
|
|
"step": 12145,
|
|
"valid_targets_mean": 1423.0,
|
|
"valid_targets_min": 817
|
|
},
|
|
{
|
|
"epoch": 6.3645887899423785,
|
|
"grad_norm": 0.44211213435952007,
|
|
"learning_rate": 9.973594209979475e-07,
|
|
"loss": 0.1572,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06968766450881958,
|
|
"step": 12150,
|
|
"valid_targets_mean": 5531.1,
|
|
"valid_targets_min": 3893
|
|
},
|
|
{
|
|
"epoch": 6.367207962283918,
|
|
"grad_norm": 0.4166404237161675,
|
|
"learning_rate": 9.892291244645014e-07,
|
|
"loss": 0.1399,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0780155211687088,
|
|
"step": 12155,
|
|
"valid_targets_mean": 6276.2,
|
|
"valid_targets_min": 4622
|
|
},
|
|
{
|
|
"epoch": 6.369827134625458,
|
|
"grad_norm": 0.40278586878859884,
|
|
"learning_rate": 9.811312616703072e-07,
|
|
"loss": 0.1455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07156006246805191,
|
|
"step": 12160,
|
|
"valid_targets_mean": 8019.8,
|
|
"valid_targets_min": 6108
|
|
},
|
|
{
|
|
"epoch": 6.3724463069669985,
|
|
"grad_norm": 0.43210480480170727,
|
|
"learning_rate": 9.730658464309007e-07,
|
|
"loss": 0.1511,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07802318036556244,
|
|
"step": 12165,
|
|
"valid_targets_mean": 6822.1,
|
|
"valid_targets_min": 4253
|
|
},
|
|
{
|
|
"epoch": 6.375065479308539,
|
|
"grad_norm": 0.425610981171059,
|
|
"learning_rate": 9.650328925064545e-07,
|
|
"loss": 0.1588,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07639053463935852,
|
|
"step": 12170,
|
|
"valid_targets_mean": 6800.6,
|
|
"valid_targets_min": 4990
|
|
},
|
|
{
|
|
"epoch": 6.377684651650078,
|
|
"grad_norm": 0.43050386355730597,
|
|
"learning_rate": 9.570324136017683e-07,
|
|
"loss": 0.1494,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09192109107971191,
|
|
"step": 12175,
|
|
"valid_targets_mean": 6860.5,
|
|
"valid_targets_min": 4957
|
|
},
|
|
{
|
|
"epoch": 6.3803038239916186,
|
|
"grad_norm": 0.4091475548747802,
|
|
"learning_rate": 9.490644233662328e-07,
|
|
"loss": 0.1488,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06866734474897385,
|
|
"step": 12180,
|
|
"valid_targets_mean": 5820.2,
|
|
"valid_targets_min": 4775
|
|
},
|
|
{
|
|
"epoch": 6.382922996333159,
|
|
"grad_norm": 0.37973839807835397,
|
|
"learning_rate": 9.411289353938136e-07,
|
|
"loss": 0.1404,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05396395921707153,
|
|
"step": 12185,
|
|
"valid_targets_mean": 5628.8,
|
|
"valid_targets_min": 4098
|
|
},
|
|
{
|
|
"epoch": 6.385542168674699,
|
|
"grad_norm": 0.3950700056570422,
|
|
"learning_rate": 9.332259632230123e-07,
|
|
"loss": 0.1366,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07747262716293335,
|
|
"step": 12190,
|
|
"valid_targets_mean": 6170.6,
|
|
"valid_targets_min": 4801
|
|
},
|
|
{
|
|
"epoch": 6.388161341016239,
|
|
"grad_norm": 0.4635368797467224,
|
|
"learning_rate": 9.253555203368792e-07,
|
|
"loss": 0.1872,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18500551581382751,
|
|
"step": 12195,
|
|
"valid_targets_mean": 6196.0,
|
|
"valid_targets_min": 4734
|
|
},
|
|
{
|
|
"epoch": 6.390780513357779,
|
|
"grad_norm": 0.44116838575681694,
|
|
"learning_rate": 9.175176201629466e-07,
|
|
"loss": 0.1507,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08110781013965607,
|
|
"step": 12200,
|
|
"valid_targets_mean": 6086.4,
|
|
"valid_targets_min": 4400
|
|
},
|
|
{
|
|
"epoch": 6.393399685699319,
|
|
"grad_norm": 0.4394803002233574,
|
|
"learning_rate": 9.09712276073238e-07,
|
|
"loss": 0.1383,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07908235490322113,
|
|
"step": 12205,
|
|
"valid_targets_mean": 6544.9,
|
|
"valid_targets_min": 4919
|
|
},
|
|
{
|
|
"epoch": 6.3960188580408595,
|
|
"grad_norm": 0.4318114046998599,
|
|
"learning_rate": 9.019395013842303e-07,
|
|
"loss": 0.1465,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0808994323015213,
|
|
"step": 12210,
|
|
"valid_targets_mean": 6052.2,
|
|
"valid_targets_min": 4611
|
|
},
|
|
{
|
|
"epoch": 6.398638030382399,
|
|
"grad_norm": 0.43101304490405234,
|
|
"learning_rate": 8.941993093568314e-07,
|
|
"loss": 0.149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06643542647361755,
|
|
"step": 12215,
|
|
"valid_targets_mean": 6934.9,
|
|
"valid_targets_min": 2275
|
|
},
|
|
{
|
|
"epoch": 6.401257202723939,
|
|
"grad_norm": 0.4130400097681656,
|
|
"learning_rate": 8.86491713196369e-07,
|
|
"loss": 0.1284,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06309416890144348,
|
|
"step": 12220,
|
|
"valid_targets_mean": 5878.5,
|
|
"valid_targets_min": 4826
|
|
},
|
|
{
|
|
"epoch": 6.4038763750654795,
|
|
"grad_norm": 0.39866053805511975,
|
|
"learning_rate": 8.788167260525537e-07,
|
|
"loss": 0.1424,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06360381841659546,
|
|
"step": 12225,
|
|
"valid_targets_mean": 6784.9,
|
|
"valid_targets_min": 4707
|
|
},
|
|
{
|
|
"epoch": 6.40649554740702,
|
|
"grad_norm": 0.3952513558012143,
|
|
"learning_rate": 8.7117436101946e-07,
|
|
"loss": 0.1313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10245849192142487,
|
|
"step": 12230,
|
|
"valid_targets_mean": 7245.8,
|
|
"valid_targets_min": 5421
|
|
},
|
|
{
|
|
"epoch": 6.409114719748559,
|
|
"grad_norm": 0.3855066970646897,
|
|
"learning_rate": 8.635646311355206e-07,
|
|
"loss": 0.1371,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0684409886598587,
|
|
"step": 12235,
|
|
"valid_targets_mean": 7122.0,
|
|
"valid_targets_min": 4949
|
|
},
|
|
{
|
|
"epoch": 6.4117338920900995,
|
|
"grad_norm": 0.3819539134322198,
|
|
"learning_rate": 8.559875493834702e-07,
|
|
"loss": 0.1375,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0731293186545372,
|
|
"step": 12240,
|
|
"valid_targets_mean": 6444.5,
|
|
"valid_targets_min": 4831
|
|
},
|
|
{
|
|
"epoch": 6.41435306443164,
|
|
"grad_norm": 0.4052284152288746,
|
|
"learning_rate": 8.484431286903616e-07,
|
|
"loss": 0.1437,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08262995630502701,
|
|
"step": 12245,
|
|
"valid_targets_mean": 6454.9,
|
|
"valid_targets_min": 4018
|
|
},
|
|
{
|
|
"epoch": 6.41697223677318,
|
|
"grad_norm": 0.3474902598250319,
|
|
"learning_rate": 8.409313819275144e-07,
|
|
"loss": 0.1355,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0552181750535965,
|
|
"step": 12250,
|
|
"valid_targets_mean": 6851.4,
|
|
"valid_targets_min": 5034
|
|
},
|
|
{
|
|
"epoch": 6.4195914091147195,
|
|
"grad_norm": 0.5507816017274946,
|
|
"learning_rate": 8.334523219105084e-07,
|
|
"loss": 0.1383,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0801452100276947,
|
|
"step": 12255,
|
|
"valid_targets_mean": 6674.2,
|
|
"valid_targets_min": 4682
|
|
},
|
|
{
|
|
"epoch": 6.42221058145626,
|
|
"grad_norm": 0.447494323940838,
|
|
"learning_rate": 8.260059613991545e-07,
|
|
"loss": 0.1496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07584092020988464,
|
|
"step": 12260,
|
|
"valid_targets_mean": 5518.1,
|
|
"valid_targets_min": 4764
|
|
},
|
|
{
|
|
"epoch": 6.4248297537978,
|
|
"grad_norm": 0.4725890617126325,
|
|
"learning_rate": 8.185923130974816e-07,
|
|
"loss": 0.1496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08310903608798981,
|
|
"step": 12265,
|
|
"valid_targets_mean": 5329.2,
|
|
"valid_targets_min": 2481
|
|
},
|
|
{
|
|
"epoch": 6.42744892613934,
|
|
"grad_norm": 0.6737547623808363,
|
|
"learning_rate": 8.112113896537033e-07,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11236198246479034,
|
|
"step": 12270,
|
|
"valid_targets_mean": 4792.6,
|
|
"valid_targets_min": 702
|
|
},
|
|
{
|
|
"epoch": 6.43006809848088,
|
|
"grad_norm": 0.6449972081489183,
|
|
"learning_rate": 8.038632036602068e-07,
|
|
"loss": 0.2083,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11374415457248688,
|
|
"step": 12275,
|
|
"valid_targets_mean": 5322.8,
|
|
"valid_targets_min": 1005
|
|
},
|
|
{
|
|
"epoch": 6.43268727082242,
|
|
"grad_norm": 0.5654573992879217,
|
|
"learning_rate": 7.965477676535216e-07,
|
|
"loss": 0.206,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07705260813236237,
|
|
"step": 12280,
|
|
"valid_targets_mean": 3584.6,
|
|
"valid_targets_min": 2054
|
|
},
|
|
{
|
|
"epoch": 6.43530644316396,
|
|
"grad_norm": 0.6863852664683533,
|
|
"learning_rate": 7.892650941143109e-07,
|
|
"loss": 0.2056,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09170589596033096,
|
|
"step": 12285,
|
|
"valid_targets_mean": 3685.6,
|
|
"valid_targets_min": 1751
|
|
},
|
|
{
|
|
"epoch": 6.437925615505501,
|
|
"grad_norm": 0.602298364243919,
|
|
"learning_rate": 7.820151954673316e-07,
|
|
"loss": 0.2036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12648309767246246,
|
|
"step": 12290,
|
|
"valid_targets_mean": 5944.2,
|
|
"valid_targets_min": 1970
|
|
},
|
|
{
|
|
"epoch": 6.44054478784704,
|
|
"grad_norm": 0.5507065526632529,
|
|
"learning_rate": 7.747980840814384e-07,
|
|
"loss": 0.2099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09939628094434738,
|
|
"step": 12295,
|
|
"valid_targets_mean": 5916.0,
|
|
"valid_targets_min": 920
|
|
},
|
|
{
|
|
"epoch": 6.44316396018858,
|
|
"grad_norm": 0.6006661949709525,
|
|
"learning_rate": 7.676137722695376e-07,
|
|
"loss": 0.2027,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09916333854198456,
|
|
"step": 12300,
|
|
"valid_targets_mean": 6475.4,
|
|
"valid_targets_min": 2231
|
|
},
|
|
{
|
|
"epoch": 6.445783132530121,
|
|
"grad_norm": 0.5738573068037527,
|
|
"learning_rate": 7.60462272288578e-07,
|
|
"loss": 0.201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07840252667665482,
|
|
"step": 12305,
|
|
"valid_targets_mean": 4043.4,
|
|
"valid_targets_min": 467
|
|
},
|
|
{
|
|
"epoch": 6.44840230487166,
|
|
"grad_norm": 0.5439295935975441,
|
|
"learning_rate": 7.533435963395375e-07,
|
|
"loss": 0.212,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11592923104763031,
|
|
"step": 12310,
|
|
"valid_targets_mean": 5549.8,
|
|
"valid_targets_min": 3440
|
|
},
|
|
{
|
|
"epoch": 6.4510214772132,
|
|
"grad_norm": 0.8590229109445299,
|
|
"learning_rate": 7.462577565673835e-07,
|
|
"loss": 0.21,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10425989329814911,
|
|
"step": 12315,
|
|
"valid_targets_mean": 4736.9,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 6.453640649554741,
|
|
"grad_norm": 0.6337785584790162,
|
|
"learning_rate": 7.392047650610679e-07,
|
|
"loss": 0.2034,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11334839463233948,
|
|
"step": 12320,
|
|
"valid_targets_mean": 4838.2,
|
|
"valid_targets_min": 1223
|
|
},
|
|
{
|
|
"epoch": 6.456259821896281,
|
|
"grad_norm": 0.5547802198399892,
|
|
"learning_rate": 7.321846338535032e-07,
|
|
"loss": 0.204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10477716475725174,
|
|
"step": 12325,
|
|
"valid_targets_mean": 5466.9,
|
|
"valid_targets_min": 1552
|
|
},
|
|
{
|
|
"epoch": 6.458878994237821,
|
|
"grad_norm": 0.5876783595854257,
|
|
"learning_rate": 7.251973749215357e-07,
|
|
"loss": 0.2041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11823265999555588,
|
|
"step": 12330,
|
|
"valid_targets_mean": 5166.0,
|
|
"valid_targets_min": 1546
|
|
},
|
|
{
|
|
"epoch": 6.461498166579361,
|
|
"grad_norm": 0.5857999618397483,
|
|
"learning_rate": 7.182430001859253e-07,
|
|
"loss": 0.1984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09447768330574036,
|
|
"step": 12335,
|
|
"valid_targets_mean": 3848.2,
|
|
"valid_targets_min": 1957
|
|
},
|
|
{
|
|
"epoch": 6.464117338920901,
|
|
"grad_norm": 0.6194569579809597,
|
|
"learning_rate": 7.113215215113456e-07,
|
|
"loss": 0.2007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09726107120513916,
|
|
"step": 12340,
|
|
"valid_targets_mean": 3487.5,
|
|
"valid_targets_min": 1538
|
|
},
|
|
{
|
|
"epoch": 6.466736511262441,
|
|
"grad_norm": 0.6193271598241644,
|
|
"learning_rate": 7.044329507063241e-07,
|
|
"loss": 0.1939,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09087219834327698,
|
|
"step": 12345,
|
|
"valid_targets_mean": 3397.2,
|
|
"valid_targets_min": 1645
|
|
},
|
|
{
|
|
"epoch": 6.469355683603981,
|
|
"grad_norm": 0.6559474003243466,
|
|
"learning_rate": 6.975772995232621e-07,
|
|
"loss": 0.1986,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08520987629890442,
|
|
"step": 12350,
|
|
"valid_targets_mean": 2859.4,
|
|
"valid_targets_min": 935
|
|
},
|
|
{
|
|
"epoch": 6.471974855945521,
|
|
"grad_norm": 0.6428355270463898,
|
|
"learning_rate": 6.907545796583926e-07,
|
|
"loss": 0.2026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10408250987529755,
|
|
"step": 12355,
|
|
"valid_targets_mean": 3490.6,
|
|
"valid_targets_min": 1269
|
|
},
|
|
{
|
|
"epoch": 6.474594028287061,
|
|
"grad_norm": 0.6780616244891007,
|
|
"learning_rate": 6.839648027517665e-07,
|
|
"loss": 0.2007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09541037678718567,
|
|
"step": 12360,
|
|
"valid_targets_mean": 2860.0,
|
|
"valid_targets_min": 1173
|
|
},
|
|
{
|
|
"epoch": 6.477213200628602,
|
|
"grad_norm": 0.6127512057602541,
|
|
"learning_rate": 6.772079803872267e-07,
|
|
"loss": 0.1981,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09076771885156631,
|
|
"step": 12365,
|
|
"valid_targets_mean": 3480.9,
|
|
"valid_targets_min": 1392
|
|
},
|
|
{
|
|
"epoch": 6.479832372970141,
|
|
"grad_norm": 0.6625302661809119,
|
|
"learning_rate": 6.704841240924031e-07,
|
|
"loss": 0.2021,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1085045337677002,
|
|
"step": 12370,
|
|
"valid_targets_mean": 3878.6,
|
|
"valid_targets_min": 1883
|
|
},
|
|
{
|
|
"epoch": 6.482451545311681,
|
|
"grad_norm": 0.6526193395022808,
|
|
"learning_rate": 6.637932453386664e-07,
|
|
"loss": 0.1928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11360698938369751,
|
|
"step": 12375,
|
|
"valid_targets_mean": 3721.4,
|
|
"valid_targets_min": 2106
|
|
},
|
|
{
|
|
"epoch": 6.485070717653222,
|
|
"grad_norm": 0.5927849705531022,
|
|
"learning_rate": 6.571353555411475e-07,
|
|
"loss": 0.1782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07101786136627197,
|
|
"step": 12380,
|
|
"valid_targets_mean": 3868.5,
|
|
"valid_targets_min": 2222
|
|
},
|
|
{
|
|
"epoch": 6.487689889994762,
|
|
"grad_norm": 0.6476439861205255,
|
|
"learning_rate": 6.505104660586759e-07,
|
|
"loss": 0.1928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09150652587413788,
|
|
"step": 12385,
|
|
"valid_targets_mean": 3010.1,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 6.490309062336301,
|
|
"grad_norm": 0.6184049122484422,
|
|
"learning_rate": 6.439185881937949e-07,
|
|
"loss": 0.1963,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09595520049333572,
|
|
"step": 12390,
|
|
"valid_targets_mean": 3789.1,
|
|
"valid_targets_min": 771
|
|
},
|
|
{
|
|
"epoch": 6.492928234677842,
|
|
"grad_norm": 0.6302199004088047,
|
|
"learning_rate": 6.373597331927128e-07,
|
|
"loss": 0.1986,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06606753170490265,
|
|
"step": 12395,
|
|
"valid_targets_mean": 3578.9,
|
|
"valid_targets_min": 1774
|
|
},
|
|
{
|
|
"epoch": 6.495547407019382,
|
|
"grad_norm": 0.693365212113418,
|
|
"learning_rate": 6.308339122453189e-07,
|
|
"loss": 0.1908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08515545725822449,
|
|
"step": 12400,
|
|
"valid_targets_mean": 2545.8,
|
|
"valid_targets_min": 997
|
|
},
|
|
{
|
|
"epoch": 6.498166579360922,
|
|
"grad_norm": 0.6048352005477383,
|
|
"learning_rate": 6.243411364851204e-07,
|
|
"loss": 0.1921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09665270149707794,
|
|
"step": 12405,
|
|
"valid_targets_mean": 3667.1,
|
|
"valid_targets_min": 1741
|
|
},
|
|
{
|
|
"epoch": 6.500785751702462,
|
|
"grad_norm": 0.6792020303860473,
|
|
"learning_rate": 6.178814169892633e-07,
|
|
"loss": 0.1938,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09353077411651611,
|
|
"step": 12410,
|
|
"valid_targets_mean": 3606.1,
|
|
"valid_targets_min": 863
|
|
},
|
|
{
|
|
"epoch": 6.503404924044002,
|
|
"grad_norm": 0.6470828919156506,
|
|
"learning_rate": 6.114547647784941e-07,
|
|
"loss": 0.1894,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0979934111237526,
|
|
"step": 12415,
|
|
"valid_targets_mean": 3876.9,
|
|
"valid_targets_min": 1854
|
|
},
|
|
{
|
|
"epoch": 6.506024096385542,
|
|
"grad_norm": 0.5779518222287894,
|
|
"learning_rate": 6.0506119081714e-07,
|
|
"loss": 0.1819,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09203438460826874,
|
|
"step": 12420,
|
|
"valid_targets_mean": 4763.2,
|
|
"valid_targets_min": 2040
|
|
},
|
|
{
|
|
"epoch": 6.5086432687270825,
|
|
"grad_norm": 0.6346688915452731,
|
|
"learning_rate": 5.987007060130978e-07,
|
|
"loss": 0.1945,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08190780133008957,
|
|
"step": 12425,
|
|
"valid_targets_mean": 3521.9,
|
|
"valid_targets_min": 1036
|
|
},
|
|
{
|
|
"epoch": 6.511262441068622,
|
|
"grad_norm": 0.6318536696051502,
|
|
"learning_rate": 5.92373321217814e-07,
|
|
"loss": 0.1938,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10408752411603928,
|
|
"step": 12430,
|
|
"valid_targets_mean": 4304.8,
|
|
"valid_targets_min": 1279
|
|
},
|
|
{
|
|
"epoch": 6.513881613410162,
|
|
"grad_norm": 0.6421579092608279,
|
|
"learning_rate": 5.860790472262534e-07,
|
|
"loss": 0.186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09606657922267914,
|
|
"step": 12435,
|
|
"valid_targets_mean": 3600.0,
|
|
"valid_targets_min": 1798
|
|
},
|
|
{
|
|
"epoch": 6.5165007857517026,
|
|
"grad_norm": 0.6457038058905296,
|
|
"learning_rate": 5.798178947769106e-07,
|
|
"loss": 0.188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08636249601840973,
|
|
"step": 12440,
|
|
"valid_targets_mean": 3067.2,
|
|
"valid_targets_min": 1446
|
|
},
|
|
{
|
|
"epoch": 6.519119958093243,
|
|
"grad_norm": 0.6154324966048933,
|
|
"learning_rate": 5.735898745517499e-07,
|
|
"loss": 0.188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0979316383600235,
|
|
"step": 12445,
|
|
"valid_targets_mean": 4183.2,
|
|
"valid_targets_min": 1270
|
|
},
|
|
{
|
|
"epoch": 6.521739130434782,
|
|
"grad_norm": 0.5862027446389466,
|
|
"learning_rate": 5.673949971762249e-07,
|
|
"loss": 0.1931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10502250492572784,
|
|
"step": 12450,
|
|
"valid_targets_mean": 5485.8,
|
|
"valid_targets_min": 1216
|
|
},
|
|
{
|
|
"epoch": 6.524358302776323,
|
|
"grad_norm": 0.6376219612800321,
|
|
"learning_rate": 5.612332732192416e-07,
|
|
"loss": 0.1843,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10316480696201324,
|
|
"step": 12455,
|
|
"valid_targets_mean": 3994.0,
|
|
"valid_targets_min": 2028
|
|
},
|
|
{
|
|
"epoch": 6.526977475117863,
|
|
"grad_norm": 0.6787194184067051,
|
|
"learning_rate": 5.551047131931442e-07,
|
|
"loss": 0.195,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11441092193126678,
|
|
"step": 12460,
|
|
"valid_targets_mean": 3700.6,
|
|
"valid_targets_min": 2329
|
|
},
|
|
{
|
|
"epoch": 6.529596647459403,
|
|
"grad_norm": 0.6767800339802285,
|
|
"learning_rate": 5.490093275536934e-07,
|
|
"loss": 0.1916,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11958886682987213,
|
|
"step": 12465,
|
|
"valid_targets_mean": 4075.2,
|
|
"valid_targets_min": 1363
|
|
},
|
|
{
|
|
"epoch": 6.532215819800943,
|
|
"grad_norm": 0.6225915057487827,
|
|
"learning_rate": 5.429471267000552e-07,
|
|
"loss": 0.1814,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08626016974449158,
|
|
"step": 12470,
|
|
"valid_targets_mean": 3560.8,
|
|
"valid_targets_min": 1087
|
|
},
|
|
{
|
|
"epoch": 6.534834992142483,
|
|
"grad_norm": 0.6555202364215589,
|
|
"learning_rate": 5.36918120974781e-07,
|
|
"loss": 0.1948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09971264004707336,
|
|
"step": 12475,
|
|
"valid_targets_mean": 3860.1,
|
|
"valid_targets_min": 2483
|
|
},
|
|
{
|
|
"epoch": 6.537454164484023,
|
|
"grad_norm": 0.6733343811998177,
|
|
"learning_rate": 5.309223206637826e-07,
|
|
"loss": 0.1956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11346982419490814,
|
|
"step": 12480,
|
|
"valid_targets_mean": 4583.2,
|
|
"valid_targets_min": 2316
|
|
},
|
|
{
|
|
"epoch": 6.5400733368255635,
|
|
"grad_norm": 0.6473883245274624,
|
|
"learning_rate": 5.249597359963332e-07,
|
|
"loss": 0.1866,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09985611587762833,
|
|
"step": 12485,
|
|
"valid_targets_mean": 4198.9,
|
|
"valid_targets_min": 2022
|
|
},
|
|
{
|
|
"epoch": 6.542692509167103,
|
|
"grad_norm": 0.6039774253852044,
|
|
"learning_rate": 5.190303771450222e-07,
|
|
"loss": 0.1793,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07155510783195496,
|
|
"step": 12490,
|
|
"valid_targets_mean": 3228.9,
|
|
"valid_targets_min": 1294
|
|
},
|
|
{
|
|
"epoch": 6.545311681508643,
|
|
"grad_norm": 0.6989217643366192,
|
|
"learning_rate": 5.131342542257666e-07,
|
|
"loss": 0.1977,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11551254987716675,
|
|
"step": 12495,
|
|
"valid_targets_mean": 3846.8,
|
|
"valid_targets_min": 2716
|
|
},
|
|
{
|
|
"epoch": 6.5479308538501835,
|
|
"grad_norm": 0.6953253783559055,
|
|
"learning_rate": 5.072713772977711e-07,
|
|
"loss": 0.1892,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09119541943073273,
|
|
"step": 12500,
|
|
"valid_targets_mean": 3211.6,
|
|
"valid_targets_min": 2060
|
|
},
|
|
{
|
|
"epoch": 6.550550026191724,
|
|
"grad_norm": 0.6984431761510316,
|
|
"learning_rate": 5.014417563635276e-07,
|
|
"loss": 0.1887,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09894177317619324,
|
|
"step": 12505,
|
|
"valid_targets_mean": 3320.4,
|
|
"valid_targets_min": 2407
|
|
},
|
|
{
|
|
"epoch": 6.553169198533263,
|
|
"grad_norm": 0.611165944926459,
|
|
"learning_rate": 4.956454013687873e-07,
|
|
"loss": 0.1927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10009843111038208,
|
|
"step": 12510,
|
|
"valid_targets_mean": 4165.1,
|
|
"valid_targets_min": 1533
|
|
},
|
|
{
|
|
"epoch": 6.5557883708748035,
|
|
"grad_norm": 0.6677128161368928,
|
|
"learning_rate": 4.898823222025462e-07,
|
|
"loss": 0.1964,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12110663950443268,
|
|
"step": 12515,
|
|
"valid_targets_mean": 5201.2,
|
|
"valid_targets_min": 2189
|
|
},
|
|
{
|
|
"epoch": 6.558407543216344,
|
|
"grad_norm": 0.6393601054937041,
|
|
"learning_rate": 4.841525286970327e-07,
|
|
"loss": 0.1871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11260367929935455,
|
|
"step": 12520,
|
|
"valid_targets_mean": 4193.1,
|
|
"valid_targets_min": 1328
|
|
},
|
|
{
|
|
"epoch": 6.561026715557884,
|
|
"grad_norm": 0.6701988763152277,
|
|
"learning_rate": 4.784560306276875e-07,
|
|
"loss": 0.2032,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13839399814605713,
|
|
"step": 12525,
|
|
"valid_targets_mean": 4500.4,
|
|
"valid_targets_min": 1708
|
|
},
|
|
{
|
|
"epoch": 6.5636458878994235,
|
|
"grad_norm": 0.6291661582869631,
|
|
"learning_rate": 4.727928377131363e-07,
|
|
"loss": 0.1826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07673248648643494,
|
|
"step": 12530,
|
|
"valid_targets_mean": 3494.0,
|
|
"valid_targets_min": 1912
|
|
},
|
|
{
|
|
"epoch": 6.566265060240964,
|
|
"grad_norm": 0.676320891906221,
|
|
"learning_rate": 4.6716295961520384e-07,
|
|
"loss": 0.1807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11168824881315231,
|
|
"step": 12535,
|
|
"valid_targets_mean": 4145.2,
|
|
"valid_targets_min": 2606
|
|
},
|
|
{
|
|
"epoch": 6.568884232582504,
|
|
"grad_norm": 0.6125559428322571,
|
|
"learning_rate": 4.6156640593885805e-07,
|
|
"loss": 0.1812,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08832336217164993,
|
|
"step": 12540,
|
|
"valid_targets_mean": 4189.0,
|
|
"valid_targets_min": 1378
|
|
},
|
|
{
|
|
"epoch": 6.571503404924044,
|
|
"grad_norm": 0.6638100292937631,
|
|
"learning_rate": 4.560031862322278e-07,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11195237934589386,
|
|
"step": 12545,
|
|
"valid_targets_mean": 3983.4,
|
|
"valid_targets_min": 1219
|
|
},
|
|
{
|
|
"epoch": 6.574122577265584,
|
|
"grad_norm": 0.6212782016254619,
|
|
"learning_rate": 4.504733099865566e-07,
|
|
"loss": 0.1963,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09998372197151184,
|
|
"step": 12550,
|
|
"valid_targets_mean": 3997.0,
|
|
"valid_targets_min": 2631
|
|
},
|
|
{
|
|
"epoch": 6.576741749607124,
|
|
"grad_norm": 0.6954308718876611,
|
|
"learning_rate": 4.4497678663621756e-07,
|
|
"loss": 0.1938,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11565820127725601,
|
|
"step": 12555,
|
|
"valid_targets_mean": 4167.4,
|
|
"valid_targets_min": 1423
|
|
},
|
|
{
|
|
"epoch": 6.579360921948664,
|
|
"grad_norm": 0.6549184271938823,
|
|
"learning_rate": 4.395136255586696e-07,
|
|
"loss": 0.1874,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07491223514080048,
|
|
"step": 12560,
|
|
"valid_targets_mean": 3381.9,
|
|
"valid_targets_min": 2320
|
|
},
|
|
{
|
|
"epoch": 6.581980094290204,
|
|
"grad_norm": 0.7132956672725176,
|
|
"learning_rate": 4.340838360744615e-07,
|
|
"loss": 0.1946,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09730696678161621,
|
|
"step": 12565,
|
|
"valid_targets_mean": 3174.9,
|
|
"valid_targets_min": 1188
|
|
},
|
|
{
|
|
"epoch": 6.584599266631744,
|
|
"grad_norm": 0.613721618741684,
|
|
"learning_rate": 4.286874274471986e-07,
|
|
"loss": 0.1968,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07196123898029327,
|
|
"step": 12570,
|
|
"valid_targets_mean": 3378.6,
|
|
"valid_targets_min": 618
|
|
},
|
|
{
|
|
"epoch": 6.587218438973284,
|
|
"grad_norm": 0.7250752630540884,
|
|
"learning_rate": 4.233244088835453e-07,
|
|
"loss": 0.1914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10160218179225922,
|
|
"step": 12575,
|
|
"valid_targets_mean": 2988.1,
|
|
"valid_targets_min": 1364
|
|
},
|
|
{
|
|
"epoch": 6.589837611314825,
|
|
"grad_norm": 0.6503760009869628,
|
|
"learning_rate": 4.1799478953319814e-07,
|
|
"loss": 0.1826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08129014819860458,
|
|
"step": 12580,
|
|
"valid_targets_mean": 3199.1,
|
|
"valid_targets_min": 1397
|
|
},
|
|
{
|
|
"epoch": 6.592456783656365,
|
|
"grad_norm": 0.6238635228912598,
|
|
"learning_rate": 4.1269857848886153e-07,
|
|
"loss": 0.1776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10387745499610901,
|
|
"step": 12585,
|
|
"valid_targets_mean": 5348.0,
|
|
"valid_targets_min": 1778
|
|
},
|
|
{
|
|
"epoch": 6.595075955997904,
|
|
"grad_norm": 0.6460935080740302,
|
|
"learning_rate": 4.074357847862609e-07,
|
|
"loss": 0.1772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09542147815227509,
|
|
"step": 12590,
|
|
"valid_targets_mean": 3444.9,
|
|
"valid_targets_min": 621
|
|
},
|
|
{
|
|
"epoch": 6.597695128339445,
|
|
"grad_norm": 0.6121163929342296,
|
|
"learning_rate": 4.022064174040918e-07,
|
|
"loss": 0.1722,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07590524852275848,
|
|
"step": 12595,
|
|
"valid_targets_mean": 3872.9,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 6.600314300680985,
|
|
"grad_norm": 0.7923605265608449,
|
|
"learning_rate": 3.9701048526403774e-07,
|
|
"loss": 0.1738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08326010406017303,
|
|
"step": 12600,
|
|
"valid_targets_mean": 2546.6,
|
|
"valid_targets_min": 642
|
|
},
|
|
{
|
|
"epoch": 6.602933473022524,
|
|
"grad_norm": 0.6844825101895506,
|
|
"learning_rate": 3.9184799723072764e-07,
|
|
"loss": 0.1703,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09406258165836334,
|
|
"step": 12605,
|
|
"valid_targets_mean": 3428.2,
|
|
"valid_targets_min": 1978
|
|
},
|
|
{
|
|
"epoch": 6.605552645364065,
|
|
"grad_norm": 0.6745675038663468,
|
|
"learning_rate": 3.867189621117362e-07,
|
|
"loss": 0.1886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12714041769504547,
|
|
"step": 12610,
|
|
"valid_targets_mean": 4160.0,
|
|
"valid_targets_min": 2642
|
|
},
|
|
{
|
|
"epoch": 6.608171817705605,
|
|
"grad_norm": 0.6502643748669914,
|
|
"learning_rate": 3.81623388657566e-07,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07471387088298798,
|
|
"step": 12615,
|
|
"valid_targets_mean": 2937.1,
|
|
"valid_targets_min": 1510
|
|
},
|
|
{
|
|
"epoch": 6.610790990047145,
|
|
"grad_norm": 0.7399225689301697,
|
|
"learning_rate": 3.7656128556163407e-07,
|
|
"loss": 0.1815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09238269180059433,
|
|
"step": 12620,
|
|
"valid_targets_mean": 2744.2,
|
|
"valid_targets_min": 987
|
|
},
|
|
{
|
|
"epoch": 6.613410162388686,
|
|
"grad_norm": 0.6197270407292439,
|
|
"learning_rate": 3.7153266146024547e-07,
|
|
"loss": 0.1857,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08794254064559937,
|
|
"step": 12625,
|
|
"valid_targets_mean": 3638.4,
|
|
"valid_targets_min": 1428
|
|
},
|
|
{
|
|
"epoch": 6.616029334730225,
|
|
"grad_norm": 0.6334397777503624,
|
|
"learning_rate": 3.6653752493259976e-07,
|
|
"loss": 0.1843,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10795360803604126,
|
|
"step": 12630,
|
|
"valid_targets_mean": 4731.5,
|
|
"valid_targets_min": 1700
|
|
},
|
|
{
|
|
"epoch": 6.618648507071765,
|
|
"grad_norm": 0.5566634348297339,
|
|
"learning_rate": 3.615758845007533e-07,
|
|
"loss": 0.1791,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08727572858333588,
|
|
"step": 12635,
|
|
"valid_targets_mean": 5699.6,
|
|
"valid_targets_min": 1767
|
|
},
|
|
{
|
|
"epoch": 6.621267679413306,
|
|
"grad_norm": 0.5570405375278416,
|
|
"learning_rate": 3.5664774862962605e-07,
|
|
"loss": 0.1838,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08419948816299438,
|
|
"step": 12640,
|
|
"valid_targets_mean": 4101.5,
|
|
"valid_targets_min": 2684
|
|
},
|
|
{
|
|
"epoch": 6.623886851754845,
|
|
"grad_norm": 0.6816645231727136,
|
|
"learning_rate": 3.5175312572696795e-07,
|
|
"loss": 0.1787,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08527840673923492,
|
|
"step": 12645,
|
|
"valid_targets_mean": 2761.2,
|
|
"valid_targets_min": 1542
|
|
},
|
|
{
|
|
"epoch": 6.626506024096385,
|
|
"grad_norm": 0.6537470489254144,
|
|
"learning_rate": 3.468920241433571e-07,
|
|
"loss": 0.1835,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07553485035896301,
|
|
"step": 12650,
|
|
"valid_targets_mean": 2765.8,
|
|
"valid_targets_min": 1476
|
|
},
|
|
{
|
|
"epoch": 6.629125196437926,
|
|
"grad_norm": 0.6249077159518814,
|
|
"learning_rate": 3.4206445217217944e-07,
|
|
"loss": 0.1839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08840804547071457,
|
|
"step": 12655,
|
|
"valid_targets_mean": 3884.5,
|
|
"valid_targets_min": 1711
|
|
},
|
|
{
|
|
"epoch": 6.631744368779466,
|
|
"grad_norm": 0.6499343409410591,
|
|
"learning_rate": 3.3727041804962447e-07,
|
|
"loss": 0.1858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09530726075172424,
|
|
"step": 12660,
|
|
"valid_targets_mean": 3235.6,
|
|
"valid_targets_min": 1669
|
|
},
|
|
{
|
|
"epoch": 6.634363541121006,
|
|
"grad_norm": 0.6919759268955735,
|
|
"learning_rate": 3.3250992995465416e-07,
|
|
"loss": 0.1928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.092196524143219,
|
|
"step": 12665,
|
|
"valid_targets_mean": 3388.5,
|
|
"valid_targets_min": 1624
|
|
},
|
|
{
|
|
"epoch": 6.636982713462546,
|
|
"grad_norm": 0.6748818891045982,
|
|
"learning_rate": 3.277829960090073e-07,
|
|
"loss": 0.1874,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09625837206840515,
|
|
"step": 12670,
|
|
"valid_targets_mean": 3380.9,
|
|
"valid_targets_min": 1676
|
|
},
|
|
{
|
|
"epoch": 6.639601885804086,
|
|
"grad_norm": 0.6481965100625632,
|
|
"learning_rate": 3.230896242771642e-07,
|
|
"loss": 0.197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07984872162342072,
|
|
"step": 12675,
|
|
"valid_targets_mean": 3148.4,
|
|
"valid_targets_min": 1500
|
|
},
|
|
{
|
|
"epoch": 6.642221058145626,
|
|
"grad_norm": 0.5984040546530887,
|
|
"learning_rate": 3.184298227663618e-07,
|
|
"loss": 0.1795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08513247966766357,
|
|
"step": 12680,
|
|
"valid_targets_mean": 4343.5,
|
|
"valid_targets_min": 2556
|
|
},
|
|
{
|
|
"epoch": 6.644840230487166,
|
|
"grad_norm": 0.677679937873208,
|
|
"learning_rate": 3.13803599426552e-07,
|
|
"loss": 0.2025,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07720249146223068,
|
|
"step": 12685,
|
|
"valid_targets_mean": 3299.5,
|
|
"valid_targets_min": 1241
|
|
},
|
|
{
|
|
"epoch": 6.647459402828706,
|
|
"grad_norm": 0.7324573724604573,
|
|
"learning_rate": 3.092109621504058e-07,
|
|
"loss": 0.197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1175050139427185,
|
|
"step": 12690,
|
|
"valid_targets_mean": 3547.0,
|
|
"valid_targets_min": 992
|
|
},
|
|
{
|
|
"epoch": 6.650078575170246,
|
|
"grad_norm": 0.6946700460208859,
|
|
"learning_rate": 3.0465191877329104e-07,
|
|
"loss": 0.1854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10434059053659439,
|
|
"step": 12695,
|
|
"valid_targets_mean": 4328.6,
|
|
"valid_targets_min": 1133
|
|
},
|
|
{
|
|
"epoch": 6.6526977475117866,
|
|
"grad_norm": 0.6687980880353598,
|
|
"learning_rate": 3.001264770732615e-07,
|
|
"loss": 0.1854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09853892028331757,
|
|
"step": 12700,
|
|
"valid_targets_mean": 4202.6,
|
|
"valid_targets_min": 2369
|
|
},
|
|
{
|
|
"epoch": 6.655316919853326,
|
|
"grad_norm": 0.6663236224684165,
|
|
"learning_rate": 2.9563464477104786e-07,
|
|
"loss": 0.1876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07648585736751556,
|
|
"step": 12705,
|
|
"valid_targets_mean": 3563.0,
|
|
"valid_targets_min": 1919
|
|
},
|
|
{
|
|
"epoch": 6.657936092194866,
|
|
"grad_norm": 0.6637108325798059,
|
|
"learning_rate": 2.9117642953003564e-07,
|
|
"loss": 0.1919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11010406911373138,
|
|
"step": 12710,
|
|
"valid_targets_mean": 3975.5,
|
|
"valid_targets_min": 2902
|
|
},
|
|
{
|
|
"epoch": 6.660555264536407,
|
|
"grad_norm": 0.7498564576579266,
|
|
"learning_rate": 2.867518389562629e-07,
|
|
"loss": 0.1909,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11214965581893921,
|
|
"step": 12715,
|
|
"valid_targets_mean": 2975.5,
|
|
"valid_targets_min": 617
|
|
},
|
|
{
|
|
"epoch": 6.663174436877947,
|
|
"grad_norm": 0.6239320454745892,
|
|
"learning_rate": 2.823608805983957e-07,
|
|
"loss": 0.1719,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0791563019156456,
|
|
"step": 12720,
|
|
"valid_targets_mean": 3290.4,
|
|
"valid_targets_min": 1349
|
|
},
|
|
{
|
|
"epoch": 6.665793609219486,
|
|
"grad_norm": 0.6505688926922241,
|
|
"learning_rate": 2.7800356194772836e-07,
|
|
"loss": 0.1837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0979127585887909,
|
|
"step": 12725,
|
|
"valid_targets_mean": 3459.0,
|
|
"valid_targets_min": 1769
|
|
},
|
|
{
|
|
"epoch": 6.668412781561027,
|
|
"grad_norm": 0.6130763206181329,
|
|
"learning_rate": 2.7367989043815434e-07,
|
|
"loss": 0.18,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0859518051147461,
|
|
"step": 12730,
|
|
"valid_targets_mean": 4186.1,
|
|
"valid_targets_min": 2097
|
|
},
|
|
{
|
|
"epoch": 6.671031953902567,
|
|
"grad_norm": 0.682777529916058,
|
|
"learning_rate": 2.6938987344616864e-07,
|
|
"loss": 0.1875,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09927409142255783,
|
|
"step": 12735,
|
|
"valid_targets_mean": 3304.0,
|
|
"valid_targets_min": 1066
|
|
},
|
|
{
|
|
"epoch": 6.673651126244107,
|
|
"grad_norm": 0.6139750346228053,
|
|
"learning_rate": 2.651335182908499e-07,
|
|
"loss": 0.181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09072718769311905,
|
|
"step": 12740,
|
|
"valid_targets_mean": 4457.9,
|
|
"valid_targets_min": 2635
|
|
},
|
|
{
|
|
"epoch": 6.676270298585647,
|
|
"grad_norm": 0.6688779633764557,
|
|
"learning_rate": 2.6091083223384496e-07,
|
|
"loss": 0.1931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10317809879779816,
|
|
"step": 12745,
|
|
"valid_targets_mean": 4296.2,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 6.678889470927187,
|
|
"grad_norm": 0.6153838140431476,
|
|
"learning_rate": 2.5672182247935773e-07,
|
|
"loss": 0.1818,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08931320160627365,
|
|
"step": 12750,
|
|
"valid_targets_mean": 4020.1,
|
|
"valid_targets_min": 2131
|
|
},
|
|
{
|
|
"epoch": 6.681508643268727,
|
|
"grad_norm": 0.6841531773090127,
|
|
"learning_rate": 2.525664961741403e-07,
|
|
"loss": 0.1821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09330379217863083,
|
|
"step": 12755,
|
|
"valid_targets_mean": 3541.0,
|
|
"valid_targets_min": 1353
|
|
},
|
|
{
|
|
"epoch": 6.6841278156102675,
|
|
"grad_norm": 0.5892914974913471,
|
|
"learning_rate": 2.484448604074796e-07,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07329745590686798,
|
|
"step": 12760,
|
|
"valid_targets_mean": 3362.0,
|
|
"valid_targets_min": 1509
|
|
},
|
|
{
|
|
"epoch": 6.686746987951807,
|
|
"grad_norm": 0.6057942078920835,
|
|
"learning_rate": 2.4435692221117966e-07,
|
|
"loss": 0.179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08245901763439178,
|
|
"step": 12765,
|
|
"valid_targets_mean": 3918.9,
|
|
"valid_targets_min": 2128
|
|
},
|
|
{
|
|
"epoch": 6.689366160293347,
|
|
"grad_norm": 0.7912254274223812,
|
|
"learning_rate": 2.4030268855955495e-07,
|
|
"loss": 0.172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07037973403930664,
|
|
"step": 12770,
|
|
"valid_targets_mean": 2919.0,
|
|
"valid_targets_min": 1687
|
|
},
|
|
{
|
|
"epoch": 6.6919853326348875,
|
|
"grad_norm": 0.6624467951019597,
|
|
"learning_rate": 2.362821663694237e-07,
|
|
"loss": 0.1815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10190001875162125,
|
|
"step": 12775,
|
|
"valid_targets_mean": 3696.9,
|
|
"valid_targets_min": 1803
|
|
},
|
|
{
|
|
"epoch": 6.694604504976428,
|
|
"grad_norm": 0.6255604822386763,
|
|
"learning_rate": 2.32295362500079e-07,
|
|
"loss": 0.1771,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08623208105564117,
|
|
"step": 12780,
|
|
"valid_targets_mean": 4102.0,
|
|
"valid_targets_min": 2263
|
|
},
|
|
{
|
|
"epoch": 6.697223677317967,
|
|
"grad_norm": 0.6706980490532756,
|
|
"learning_rate": 2.2834228375330003e-07,
|
|
"loss": 0.1785,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10698755085468292,
|
|
"step": 12785,
|
|
"valid_targets_mean": 4621.4,
|
|
"valid_targets_min": 2232
|
|
},
|
|
{
|
|
"epoch": 6.6998428496595075,
|
|
"grad_norm": 0.6372593378139074,
|
|
"learning_rate": 2.2442293687331863e-07,
|
|
"loss": 0.1687,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10318998992443085,
|
|
"step": 12790,
|
|
"valid_targets_mean": 4051.6,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 6.702462022001048,
|
|
"grad_norm": 0.6193928640374642,
|
|
"learning_rate": 2.2053732854682598e-07,
|
|
"loss": 0.1729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07823237776756287,
|
|
"step": 12795,
|
|
"valid_targets_mean": 3340.1,
|
|
"valid_targets_min": 1942
|
|
},
|
|
{
|
|
"epoch": 6.705081194342588,
|
|
"grad_norm": 0.6780240423049211,
|
|
"learning_rate": 2.16685465402946e-07,
|
|
"loss": 0.179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08720569312572479,
|
|
"step": 12800,
|
|
"valid_targets_mean": 3179.4,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 6.7077003666841275,
|
|
"grad_norm": 0.659535201767842,
|
|
"learning_rate": 2.128673540132331e-07,
|
|
"loss": 0.193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09473605453968048,
|
|
"step": 12805,
|
|
"valid_targets_mean": 3492.2,
|
|
"valid_targets_min": 1695
|
|
},
|
|
{
|
|
"epoch": 6.710319539025668,
|
|
"grad_norm": 0.6495919509093001,
|
|
"learning_rate": 2.090830008916589e-07,
|
|
"loss": 0.1735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07860636711120605,
|
|
"step": 12810,
|
|
"valid_targets_mean": 4043.9,
|
|
"valid_targets_min": 1712
|
|
},
|
|
{
|
|
"epoch": 6.712938711367208,
|
|
"grad_norm": 0.611522543603785,
|
|
"learning_rate": 2.0533241249460545e-07,
|
|
"loss": 0.176,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07307950407266617,
|
|
"step": 12815,
|
|
"valid_targets_mean": 2987.9,
|
|
"valid_targets_min": 993
|
|
},
|
|
{
|
|
"epoch": 6.715557883708748,
|
|
"grad_norm": 0.5993522606495606,
|
|
"learning_rate": 2.0161559522084095e-07,
|
|
"loss": 0.1709,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08598138391971588,
|
|
"step": 12820,
|
|
"valid_targets_mean": 3854.8,
|
|
"valid_targets_min": 1679
|
|
},
|
|
{
|
|
"epoch": 6.718177056050288,
|
|
"grad_norm": 0.5714454997499965,
|
|
"learning_rate": 1.9793255541152633e-07,
|
|
"loss": 0.1803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08323566615581512,
|
|
"step": 12825,
|
|
"valid_targets_mean": 3938.8,
|
|
"valid_targets_min": 2345
|
|
},
|
|
{
|
|
"epoch": 6.720796228391828,
|
|
"grad_norm": 0.6519650606498112,
|
|
"learning_rate": 1.9428329935018642e-07,
|
|
"loss": 0.1845,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09488783776760101,
|
|
"step": 12830,
|
|
"valid_targets_mean": 4507.9,
|
|
"valid_targets_min": 3068
|
|
},
|
|
{
|
|
"epoch": 6.723415400733368,
|
|
"grad_norm": 0.6573069225818161,
|
|
"learning_rate": 1.9066783326272098e-07,
|
|
"loss": 0.1821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08799446374177933,
|
|
"step": 12835,
|
|
"valid_targets_mean": 4080.9,
|
|
"valid_targets_min": 2965
|
|
},
|
|
{
|
|
"epoch": 6.726034573074909,
|
|
"grad_norm": 0.6211160166989392,
|
|
"learning_rate": 1.870861633173693e-07,
|
|
"loss": 0.1795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07835432887077332,
|
|
"step": 12840,
|
|
"valid_targets_mean": 2993.5,
|
|
"valid_targets_min": 1817
|
|
},
|
|
{
|
|
"epoch": 6.728653745416448,
|
|
"grad_norm": 0.6320423602834774,
|
|
"learning_rate": 1.8353829562471893e-07,
|
|
"loss": 0.1778,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0895933210849762,
|
|
"step": 12845,
|
|
"valid_targets_mean": 3795.2,
|
|
"valid_targets_min": 1727
|
|
},
|
|
{
|
|
"epoch": 6.731272917757988,
|
|
"grad_norm": 0.5841924116488029,
|
|
"learning_rate": 1.8002423623768584e-07,
|
|
"loss": 0.1771,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08804034441709518,
|
|
"step": 12850,
|
|
"valid_targets_mean": 4095.9,
|
|
"valid_targets_min": 2491
|
|
},
|
|
{
|
|
"epoch": 6.733892090099529,
|
|
"grad_norm": 0.5169148104446686,
|
|
"learning_rate": 1.7654399115150988e-07,
|
|
"loss": 0.1459,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09014087915420532,
|
|
"step": 12855,
|
|
"valid_targets_mean": 2046.1,
|
|
"valid_targets_min": 333
|
|
},
|
|
{
|
|
"epoch": 6.736511262441068,
|
|
"grad_norm": 0.38211913498028294,
|
|
"learning_rate": 1.7309756630373486e-07,
|
|
"loss": 0.0801,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04085535183548927,
|
|
"step": 12860,
|
|
"valid_targets_mean": 3455.9,
|
|
"valid_targets_min": 712
|
|
},
|
|
{
|
|
"epoch": 6.739130434782608,
|
|
"grad_norm": 0.4488289713041321,
|
|
"learning_rate": 1.6968496757421294e-07,
|
|
"loss": 0.0747,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0355193093419075,
|
|
"step": 12865,
|
|
"valid_targets_mean": 2452.6,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 6.741749607124149,
|
|
"grad_norm": 0.5332516349959687,
|
|
"learning_rate": 1.663062007850824e-07,
|
|
"loss": 0.082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03894609957933426,
|
|
"step": 12870,
|
|
"valid_targets_mean": 2228.9,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 6.744368779465689,
|
|
"grad_norm": 0.3891080450363208,
|
|
"learning_rate": 1.6296127170075894e-07,
|
|
"loss": 0.0724,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035704925656318665,
|
|
"step": 12875,
|
|
"valid_targets_mean": 3038.6,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 6.746987951807229,
|
|
"grad_norm": 0.4282429329715162,
|
|
"learning_rate": 1.5965018602793315e-07,
|
|
"loss": 0.0767,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04017750546336174,
|
|
"step": 12880,
|
|
"valid_targets_mean": 3049.4,
|
|
"valid_targets_min": 1524
|
|
},
|
|
{
|
|
"epoch": 6.749607124148769,
|
|
"grad_norm": 0.802341758882023,
|
|
"learning_rate": 1.5637294941555526e-07,
|
|
"loss": 0.1187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05328008532524109,
|
|
"step": 12885,
|
|
"valid_targets_mean": 1256.1,
|
|
"valid_targets_min": 539
|
|
},
|
|
{
|
|
"epoch": 6.752226296490309,
|
|
"grad_norm": 0.390005962090855,
|
|
"learning_rate": 1.5312956745482388e-07,
|
|
"loss": 0.0878,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03711675852537155,
|
|
"step": 12890,
|
|
"valid_targets_mean": 3597.0,
|
|
"valid_targets_min": 2920
|
|
},
|
|
{
|
|
"epoch": 6.754845468831849,
|
|
"grad_norm": 0.46639672819519007,
|
|
"learning_rate": 1.4992004567918383e-07,
|
|
"loss": 0.0844,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04318893328309059,
|
|
"step": 12895,
|
|
"valid_targets_mean": 2737.9,
|
|
"valid_targets_min": 1048
|
|
},
|
|
{
|
|
"epoch": 6.757464641173389,
|
|
"grad_norm": 0.30713773620726875,
|
|
"learning_rate": 1.4674438956430837e-07,
|
|
"loss": 0.1103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02616284415125847,
|
|
"step": 12900,
|
|
"valid_targets_mean": 4340.1,
|
|
"valid_targets_min": 962
|
|
},
|
|
{
|
|
"epoch": 6.760083813514929,
|
|
"grad_norm": 0.3858031244610261,
|
|
"learning_rate": 1.4360260452809472e-07,
|
|
"loss": 0.0646,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041911303997039795,
|
|
"step": 12905,
|
|
"valid_targets_mean": 4283.1,
|
|
"valid_targets_min": 2091
|
|
},
|
|
{
|
|
"epoch": 6.762702985856469,
|
|
"grad_norm": 0.3659335945049293,
|
|
"learning_rate": 1.4049469593065524e-07,
|
|
"loss": 0.0727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03540034592151642,
|
|
"step": 12910,
|
|
"valid_targets_mean": 4320.9,
|
|
"valid_targets_min": 3349
|
|
},
|
|
{
|
|
"epoch": 6.76532215819801,
|
|
"grad_norm": 0.4146819659753387,
|
|
"learning_rate": 1.3742066907430185e-07,
|
|
"loss": 0.0741,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03829266503453255,
|
|
"step": 12915,
|
|
"valid_targets_mean": 3895.9,
|
|
"valid_targets_min": 518
|
|
},
|
|
{
|
|
"epoch": 6.76794133053955,
|
|
"grad_norm": 0.5409036972770743,
|
|
"learning_rate": 1.3438052920354827e-07,
|
|
"loss": 0.1116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04457290470600128,
|
|
"step": 12920,
|
|
"valid_targets_mean": 2219.6,
|
|
"valid_targets_min": 549
|
|
},
|
|
{
|
|
"epoch": 6.770560502881089,
|
|
"grad_norm": 0.35095890145217795,
|
|
"learning_rate": 1.313742815050878e-07,
|
|
"loss": 0.0665,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029218202456831932,
|
|
"step": 12925,
|
|
"valid_targets_mean": 3700.2,
|
|
"valid_targets_min": 2724
|
|
},
|
|
{
|
|
"epoch": 6.77317967522263,
|
|
"grad_norm": 0.49776452924136344,
|
|
"learning_rate": 1.2840193110779776e-07,
|
|
"loss": 0.0782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039242640137672424,
|
|
"step": 12930,
|
|
"valid_targets_mean": 1672.5,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 6.77579884756417,
|
|
"grad_norm": 0.3355855933765676,
|
|
"learning_rate": 1.2546348308271949e-07,
|
|
"loss": 0.0632,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02106427773833275,
|
|
"step": 12935,
|
|
"valid_targets_mean": 3546.2,
|
|
"valid_targets_min": 3183
|
|
},
|
|
{
|
|
"epoch": 6.778418019905709,
|
|
"grad_norm": 0.4377025120082717,
|
|
"learning_rate": 1.2255894244305843e-07,
|
|
"loss": 0.0773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03368924558162689,
|
|
"step": 12940,
|
|
"valid_targets_mean": 2473.2,
|
|
"valid_targets_min": 577
|
|
},
|
|
{
|
|
"epoch": 6.78103719224725,
|
|
"grad_norm": 0.5287519420200428,
|
|
"learning_rate": 1.1968831414416626e-07,
|
|
"loss": 0.0849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055708516389131546,
|
|
"step": 12945,
|
|
"valid_targets_mean": 1597.0,
|
|
"valid_targets_min": 557
|
|
},
|
|
{
|
|
"epoch": 6.78365636458879,
|
|
"grad_norm": 0.4376992157043782,
|
|
"learning_rate": 1.1685160308354315e-07,
|
|
"loss": 0.0806,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04162124544382095,
|
|
"step": 12950,
|
|
"valid_targets_mean": 3019.1,
|
|
"valid_targets_min": 774
|
|
},
|
|
{
|
|
"epoch": 6.78627553693033,
|
|
"grad_norm": 0.41988601074037835,
|
|
"learning_rate": 1.1404881410082002e-07,
|
|
"loss": 0.0823,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03479935973882675,
|
|
"step": 12955,
|
|
"valid_targets_mean": 1825.8,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 6.7888947092718706,
|
|
"grad_norm": 0.49979181398974065,
|
|
"learning_rate": 1.1127995197775632e-07,
|
|
"loss": 0.1213,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0622258298099041,
|
|
"step": 12960,
|
|
"valid_targets_mean": 3686.9,
|
|
"valid_targets_min": 2819
|
|
},
|
|
{
|
|
"epoch": 6.79151388161341,
|
|
"grad_norm": 0.31947197879194844,
|
|
"learning_rate": 1.085450214382311e-07,
|
|
"loss": 0.0689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02941582351922989,
|
|
"step": 12965,
|
|
"valid_targets_mean": 4137.8,
|
|
"valid_targets_min": 3927
|
|
},
|
|
{
|
|
"epoch": 6.79413305395495,
|
|
"grad_norm": 0.552564569567585,
|
|
"learning_rate": 1.0584402714822972e-07,
|
|
"loss": 0.0794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0524916835129261,
|
|
"step": 12970,
|
|
"valid_targets_mean": 1376.1,
|
|
"valid_targets_min": 714
|
|
},
|
|
{
|
|
"epoch": 6.796752226296491,
|
|
"grad_norm": 0.4017964533375128,
|
|
"learning_rate": 1.031769737158439e-07,
|
|
"loss": 0.1005,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03385963290929794,
|
|
"step": 12975,
|
|
"valid_targets_mean": 3432.0,
|
|
"valid_targets_min": 754
|
|
},
|
|
{
|
|
"epoch": 6.79937139863803,
|
|
"grad_norm": 0.39071375705825395,
|
|
"learning_rate": 1.0054386569126051e-07,
|
|
"loss": 0.0885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037942830473184586,
|
|
"step": 12980,
|
|
"valid_targets_mean": 3855.2,
|
|
"valid_targets_min": 3178
|
|
},
|
|
{
|
|
"epoch": 6.80199057097957,
|
|
"grad_norm": 0.7343800331543665,
|
|
"learning_rate": 9.79447075667439e-08,
|
|
"loss": 0.1086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11189083755016327,
|
|
"step": 12985,
|
|
"valid_targets_mean": 1566.6,
|
|
"valid_targets_min": 920
|
|
},
|
|
{
|
|
"epoch": 6.804609743321111,
|
|
"grad_norm": 0.460517421977809,
|
|
"learning_rate": 9.53795037766514e-08,
|
|
"loss": 0.2418,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03707824647426605,
|
|
"step": 12990,
|
|
"valid_targets_mean": 3478.0,
|
|
"valid_targets_min": 2314
|
|
},
|
|
{
|
|
"epoch": 6.807228915662651,
|
|
"grad_norm": 0.40921888799466477,
|
|
"learning_rate": 9.284825869739999e-08,
|
|
"loss": 0.0848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031269654631614685,
|
|
"step": 12995,
|
|
"valid_targets_mean": 2561.1,
|
|
"valid_targets_min": 833
|
|
},
|
|
{
|
|
"epoch": 6.809848088004191,
|
|
"grad_norm": 0.4426077347230882,
|
|
"learning_rate": 9.035097664747972e-08,
|
|
"loss": 0.0825,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049915894865989685,
|
|
"step": 13000,
|
|
"valid_targets_mean": 3509.2,
|
|
"valid_targets_min": 1803
|
|
},
|
|
{
|
|
"epoch": 6.812467260345731,
|
|
"grad_norm": 0.4285004055276,
|
|
"learning_rate": 8.788766188742914e-08,
|
|
"loss": 0.1731,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05492042750120163,
|
|
"step": 13005,
|
|
"valid_targets_mean": 4291.4,
|
|
"valid_targets_min": 3089
|
|
},
|
|
{
|
|
"epoch": 6.815086432687271,
|
|
"grad_norm": 0.2697662860913481,
|
|
"learning_rate": 8.545831861983989e-08,
|
|
"loss": 0.0619,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03071390651166439,
|
|
"step": 13010,
|
|
"valid_targets_mean": 4843.2,
|
|
"valid_targets_min": 2664
|
|
},
|
|
{
|
|
"epoch": 6.817705605028811,
|
|
"grad_norm": 0.3521097778524677,
|
|
"learning_rate": 8.306295098934991e-08,
|
|
"loss": 0.0851,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.023876160383224487,
|
|
"step": 13015,
|
|
"valid_targets_mean": 3042.1,
|
|
"valid_targets_min": 560
|
|
},
|
|
{
|
|
"epoch": 6.820324777370351,
|
|
"grad_norm": 0.4436727602143218,
|
|
"learning_rate": 8.070156308262356e-08,
|
|
"loss": 0.0817,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04059450328350067,
|
|
"step": 13020,
|
|
"valid_targets_mean": 3863.6,
|
|
"valid_targets_min": 1615
|
|
},
|
|
{
|
|
"epoch": 6.822943949711891,
|
|
"grad_norm": 0.37161958824132174,
|
|
"learning_rate": 7.837415892836264e-08,
|
|
"loss": 0.069,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02635052613914013,
|
|
"step": 13025,
|
|
"valid_targets_mean": 3312.5,
|
|
"valid_targets_min": 2172
|
|
},
|
|
{
|
|
"epoch": 6.825563122053431,
|
|
"grad_norm": 0.3756347701876398,
|
|
"learning_rate": 7.608074249728204e-08,
|
|
"loss": 0.0676,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0313107855618,
|
|
"step": 13030,
|
|
"valid_targets_mean": 3059.9,
|
|
"valid_targets_min": 536
|
|
},
|
|
{
|
|
"epoch": 6.8281822943949715,
|
|
"grad_norm": 0.38514610915482933,
|
|
"learning_rate": 7.382131770211409e-08,
|
|
"loss": 0.0712,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03413877636194229,
|
|
"step": 13035,
|
|
"valid_targets_mean": 3881.0,
|
|
"valid_targets_min": 3295
|
|
},
|
|
{
|
|
"epoch": 6.830801466736511,
|
|
"grad_norm": 0.3701505847408979,
|
|
"learning_rate": 7.159588839760423e-08,
|
|
"loss": 0.0668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027802327647805214,
|
|
"step": 13040,
|
|
"valid_targets_mean": 3287.2,
|
|
"valid_targets_min": 820
|
|
},
|
|
{
|
|
"epoch": 6.833420639078051,
|
|
"grad_norm": 0.6436076477730783,
|
|
"learning_rate": 6.940445838049093e-08,
|
|
"loss": 0.1063,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10774976760149002,
|
|
"step": 13045,
|
|
"valid_targets_mean": 2566.6,
|
|
"valid_targets_min": 1046
|
|
},
|
|
{
|
|
"epoch": 6.8360398114195915,
|
|
"grad_norm": 0.3211581572540816,
|
|
"learning_rate": 6.724703138951017e-08,
|
|
"loss": 0.064,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031185396015644073,
|
|
"step": 13050,
|
|
"valid_targets_mean": 3590.1,
|
|
"valid_targets_min": 2719
|
|
},
|
|
{
|
|
"epoch": 6.838658983761132,
|
|
"grad_norm": 0.6395463224449414,
|
|
"learning_rate": 6.5123611105391e-08,
|
|
"loss": 0.0663,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05162978917360306,
|
|
"step": 13055,
|
|
"valid_targets_mean": 1638.1,
|
|
"valid_targets_min": 699
|
|
},
|
|
{
|
|
"epoch": 6.841278156102671,
|
|
"grad_norm": 0.3482974711492975,
|
|
"learning_rate": 6.303420115083781e-08,
|
|
"loss": 0.063,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02813403680920601,
|
|
"step": 13060,
|
|
"valid_targets_mean": 3621.6,
|
|
"valid_targets_min": 2353
|
|
},
|
|
{
|
|
"epoch": 6.8438973284442115,
|
|
"grad_norm": 0.3862083760665773,
|
|
"learning_rate": 6.097880509053467e-08,
|
|
"loss": 0.0641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02983633428812027,
|
|
"step": 13065,
|
|
"valid_targets_mean": 2328.1,
|
|
"valid_targets_min": 549
|
|
},
|
|
{
|
|
"epoch": 6.846516500785752,
|
|
"grad_norm": 0.4292286502108055,
|
|
"learning_rate": 5.8957426431136554e-08,
|
|
"loss": 0.1103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028507661074399948,
|
|
"step": 13070,
|
|
"valid_targets_mean": 3330.4,
|
|
"valid_targets_min": 908
|
|
},
|
|
{
|
|
"epoch": 6.849135673127292,
|
|
"grad_norm": 0.4010713221943791,
|
|
"learning_rate": 5.69700686212582e-08,
|
|
"loss": 0.0929,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06634754687547684,
|
|
"step": 13075,
|
|
"valid_targets_mean": 3458.4,
|
|
"valid_targets_min": 2196
|
|
},
|
|
{
|
|
"epoch": 6.8517548454688315,
|
|
"grad_norm": 0.3699757785067211,
|
|
"learning_rate": 5.5016735051474093e-08,
|
|
"loss": 0.0769,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026269644498825073,
|
|
"step": 13080,
|
|
"valid_targets_mean": 1252.1,
|
|
"valid_targets_min": 598
|
|
},
|
|
{
|
|
"epoch": 6.854374017810372,
|
|
"grad_norm": 0.9234512573724675,
|
|
"learning_rate": 5.309742905431403e-08,
|
|
"loss": 0.0904,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08189341425895691,
|
|
"step": 13085,
|
|
"valid_targets_mean": 1353.1,
|
|
"valid_targets_min": 578
|
|
},
|
|
{
|
|
"epoch": 6.856993190151912,
|
|
"grad_norm": 0.6584269157012614,
|
|
"learning_rate": 5.121215390424761e-08,
|
|
"loss": 0.1143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06208275631070137,
|
|
"step": 13090,
|
|
"valid_targets_mean": 1974.2,
|
|
"valid_targets_min": 695
|
|
},
|
|
{
|
|
"epoch": 6.859612362493452,
|
|
"grad_norm": 0.4854169557630643,
|
|
"learning_rate": 4.9360912817695286e-08,
|
|
"loss": 0.0774,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03507557511329651,
|
|
"step": 13095,
|
|
"valid_targets_mean": 1724.5,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 6.862231534834992,
|
|
"grad_norm": 0.3355601007497446,
|
|
"learning_rate": 4.7543708953003974e-08,
|
|
"loss": 0.0944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029666226357221603,
|
|
"step": 13100,
|
|
"valid_targets_mean": 3842.6,
|
|
"valid_targets_min": 2024
|
|
},
|
|
{
|
|
"epoch": 6.864850707176532,
|
|
"grad_norm": 0.5702376430529507,
|
|
"learning_rate": 4.576054541045149e-08,
|
|
"loss": 0.0772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05020301789045334,
|
|
"step": 13105,
|
|
"valid_targets_mean": 1651.0,
|
|
"valid_targets_min": 523
|
|
},
|
|
{
|
|
"epoch": 6.867469879518072,
|
|
"grad_norm": 0.3188062480989316,
|
|
"learning_rate": 4.4011425232248774e-08,
|
|
"loss": 0.0713,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02868589386343956,
|
|
"step": 13110,
|
|
"valid_targets_mean": 3532.5,
|
|
"valid_targets_min": 468
|
|
},
|
|
{
|
|
"epoch": 6.870089051859613,
|
|
"grad_norm": 0.288519923378264,
|
|
"learning_rate": 4.229635140251542e-08,
|
|
"loss": 0.0776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.019699353724718094,
|
|
"step": 13115,
|
|
"valid_targets_mean": 5298.0,
|
|
"valid_targets_min": 1195
|
|
},
|
|
{
|
|
"epoch": 6.872708224201152,
|
|
"grad_norm": 0.3565904224985087,
|
|
"learning_rate": 4.061532684729308e-08,
|
|
"loss": 0.0679,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.022875187918543816,
|
|
"step": 13120,
|
|
"valid_targets_mean": 3562.6,
|
|
"valid_targets_min": 459
|
|
},
|
|
{
|
|
"epoch": 6.875327396542692,
|
|
"grad_norm": 0.36958882575456525,
|
|
"learning_rate": 3.896835443452762e-08,
|
|
"loss": 0.0673,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03134230524301529,
|
|
"step": 13125,
|
|
"valid_targets_mean": 3548.0,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 6.877946568884233,
|
|
"grad_norm": 0.5216557128177458,
|
|
"learning_rate": 3.735543697407362e-08,
|
|
"loss": 0.077,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047654591500759125,
|
|
"step": 13130,
|
|
"valid_targets_mean": 2130.1,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 6.880565741225773,
|
|
"grad_norm": 0.3969544010234322,
|
|
"learning_rate": 3.577657721768546e-08,
|
|
"loss": 0.0844,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041403673589229584,
|
|
"step": 13135,
|
|
"valid_targets_mean": 3661.6,
|
|
"valid_targets_min": 2759
|
|
},
|
|
{
|
|
"epoch": 6.883184913567312,
|
|
"grad_norm": 0.4335928920568747,
|
|
"learning_rate": 3.423177785900844e-08,
|
|
"loss": 0.0843,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04310900717973709,
|
|
"step": 13140,
|
|
"valid_targets_mean": 3909.5,
|
|
"valid_targets_min": 3391
|
|
},
|
|
{
|
|
"epoch": 6.885804085908853,
|
|
"grad_norm": 0.3516525920530094,
|
|
"learning_rate": 3.272104153358324e-08,
|
|
"loss": 0.069,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03035726025700569,
|
|
"step": 13145,
|
|
"valid_targets_mean": 3319.9,
|
|
"valid_targets_min": 1099
|
|
},
|
|
{
|
|
"epoch": 6.888423258250393,
|
|
"grad_norm": 0.38123268468604504,
|
|
"learning_rate": 3.12443708188348e-08,
|
|
"loss": 0.0712,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0375220850110054,
|
|
"step": 13150,
|
|
"valid_targets_mean": 3717.0,
|
|
"valid_targets_min": 1874
|
|
},
|
|
{
|
|
"epoch": 6.891042430591933,
|
|
"grad_norm": 0.3370066351747024,
|
|
"learning_rate": 2.9801768234065663e-08,
|
|
"loss": 0.0807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030697906389832497,
|
|
"step": 13155,
|
|
"valid_targets_mean": 3263.5,
|
|
"valid_targets_min": 724
|
|
},
|
|
{
|
|
"epoch": 6.893661602933473,
|
|
"grad_norm": 0.31056030103191984,
|
|
"learning_rate": 2.8393236240464862e-08,
|
|
"loss": 0.0595,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027445081621408463,
|
|
"step": 13160,
|
|
"valid_targets_mean": 3707.5,
|
|
"valid_targets_min": 2769
|
|
},
|
|
{
|
|
"epoch": 6.896280775275013,
|
|
"grad_norm": 0.4350193105458767,
|
|
"learning_rate": 2.7018777241083484e-08,
|
|
"loss": 0.078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03260853886604309,
|
|
"step": 13165,
|
|
"valid_targets_mean": 2815.8,
|
|
"valid_targets_min": 645
|
|
},
|
|
{
|
|
"epoch": 6.898899947616553,
|
|
"grad_norm": 0.3672746943249268,
|
|
"learning_rate": 2.5678393580852446e-08,
|
|
"loss": 0.0733,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02962132729589939,
|
|
"step": 13170,
|
|
"valid_targets_mean": 3488.8,
|
|
"valid_targets_min": 600
|
|
},
|
|
{
|
|
"epoch": 6.901519119958094,
|
|
"grad_norm": 0.47663563451660984,
|
|
"learning_rate": 2.4372087546558064e-08,
|
|
"loss": 0.0882,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03674224019050598,
|
|
"step": 13175,
|
|
"valid_targets_mean": 3373.4,
|
|
"valid_targets_min": 1648
|
|
},
|
|
{
|
|
"epoch": 6.904138292299633,
|
|
"grad_norm": 0.46794766227137546,
|
|
"learning_rate": 2.309986136685316e-08,
|
|
"loss": 0.0759,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04254136234521866,
|
|
"step": 13180,
|
|
"valid_targets_mean": 2555.0,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 6.906757464641173,
|
|
"grad_norm": 0.8481521508164337,
|
|
"learning_rate": 2.186171721225039e-08,
|
|
"loss": 0.0858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05657971277832985,
|
|
"step": 13185,
|
|
"valid_targets_mean": 867.1,
|
|
"valid_targets_min": 612
|
|
},
|
|
{
|
|
"epoch": 6.909376636982714,
|
|
"grad_norm": 0.5203667342062566,
|
|
"learning_rate": 2.0657657195106706e-08,
|
|
"loss": 0.0754,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055201150476932526,
|
|
"step": 13190,
|
|
"valid_targets_mean": 2202.1,
|
|
"valid_targets_min": 830
|
|
},
|
|
{
|
|
"epoch": 6.911995809324253,
|
|
"grad_norm": 0.4604582895283477,
|
|
"learning_rate": 1.948768336963891e-08,
|
|
"loss": 0.0868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033401697874069214,
|
|
"step": 13195,
|
|
"valid_targets_mean": 3027.0,
|
|
"valid_targets_min": 1146
|
|
},
|
|
{
|
|
"epoch": 6.914614981665793,
|
|
"grad_norm": 0.4239609510185353,
|
|
"learning_rate": 1.8351797731908093e-08,
|
|
"loss": 0.075,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043332524597644806,
|
|
"step": 13200,
|
|
"valid_targets_mean": 3296.6,
|
|
"valid_targets_min": 2760
|
|
},
|
|
{
|
|
"epoch": 6.917234154007334,
|
|
"grad_norm": 0.5935804080113309,
|
|
"learning_rate": 1.725000221981077e-08,
|
|
"loss": 0.0689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04814621061086655,
|
|
"step": 13205,
|
|
"valid_targets_mean": 1469.5,
|
|
"valid_targets_min": 628
|
|
},
|
|
{
|
|
"epoch": 6.919853326348874,
|
|
"grad_norm": 0.6134461274577795,
|
|
"learning_rate": 1.61822987130944e-08,
|
|
"loss": 0.0897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07188950479030609,
|
|
"step": 13210,
|
|
"valid_targets_mean": 1752.1,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 6.922472498690414,
|
|
"grad_norm": 0.3179609176982656,
|
|
"learning_rate": 1.5148689033333e-08,
|
|
"loss": 0.0684,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0196039117872715,
|
|
"step": 13215,
|
|
"valid_targets_mean": 3863.9,
|
|
"valid_targets_min": 3613
|
|
},
|
|
{
|
|
"epoch": 6.925091671031954,
|
|
"grad_norm": 0.3614387988412621,
|
|
"learning_rate": 1.4149174943944854e-08,
|
|
"loss": 0.0663,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03423801809549332,
|
|
"step": 13220,
|
|
"valid_targets_mean": 4304.2,
|
|
"valid_targets_min": 1773
|
|
},
|
|
{
|
|
"epoch": 6.927710843373494,
|
|
"grad_norm": 0.374604024416944,
|
|
"learning_rate": 1.3183758150165926e-08,
|
|
"loss": 0.0728,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039071012288331985,
|
|
"step": 13225,
|
|
"valid_targets_mean": 1971.9,
|
|
"valid_targets_min": 762
|
|
},
|
|
{
|
|
"epoch": 6.930330015715034,
|
|
"grad_norm": 0.3937368208843535,
|
|
"learning_rate": 1.2252440299072022e-08,
|
|
"loss": 0.0634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04523179307579994,
|
|
"step": 13230,
|
|
"valid_targets_mean": 4554.1,
|
|
"valid_targets_min": 946
|
|
},
|
|
{
|
|
"epoch": 6.932949188056574,
|
|
"grad_norm": 0.3459868864418305,
|
|
"learning_rate": 1.1355222979556601e-08,
|
|
"loss": 0.061,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04321659356355667,
|
|
"step": 13235,
|
|
"valid_targets_mean": 5015.5,
|
|
"valid_targets_min": 4104
|
|
},
|
|
{
|
|
"epoch": 6.935568360398114,
|
|
"grad_norm": 0.37124590537719093,
|
|
"learning_rate": 1.049210772233522e-08,
|
|
"loss": 0.08,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03389116749167442,
|
|
"step": 13240,
|
|
"valid_targets_mean": 3064.4,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 6.938187532739654,
|
|
"grad_norm": 0.6654117129434783,
|
|
"learning_rate": 9.663095999947746e-09,
|
|
"loss": 0.0895,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03874023258686066,
|
|
"step": 13245,
|
|
"valid_targets_mean": 3863.6,
|
|
"valid_targets_min": 3550
|
|
},
|
|
{
|
|
"epoch": 6.940806705081195,
|
|
"grad_norm": 0.4001454604435511,
|
|
"learning_rate": 8.868189226742818e-09,
|
|
"loss": 0.0715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04071683809161186,
|
|
"step": 13250,
|
|
"valid_targets_mean": 3474.9,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 6.943425877422735,
|
|
"grad_norm": 0.37231918968783745,
|
|
"learning_rate": 8.107388758893386e-09,
|
|
"loss": 0.0773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030651133507490158,
|
|
"step": 13255,
|
|
"valid_targets_mean": 2568.8,
|
|
"valid_targets_min": 774
|
|
},
|
|
{
|
|
"epoch": 6.946045049764274,
|
|
"grad_norm": 0.3178243411639312,
|
|
"learning_rate": 7.380695894376733e-09,
|
|
"loss": 0.0906,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027471695095300674,
|
|
"step": 13260,
|
|
"valid_targets_mean": 3486.2,
|
|
"valid_targets_min": 888
|
|
},
|
|
{
|
|
"epoch": 6.948664222105815,
|
|
"grad_norm": 0.4390643538547763,
|
|
"learning_rate": 6.688111872983349e-09,
|
|
"loss": 0.0686,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03612307459115982,
|
|
"step": 13265,
|
|
"valid_targets_mean": 3371.6,
|
|
"valid_targets_min": 1048
|
|
},
|
|
{
|
|
"epoch": 6.951283394447355,
|
|
"grad_norm": 0.38238054940342986,
|
|
"learning_rate": 6.0296378763125e-09,
|
|
"loss": 0.0767,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0332840159535408,
|
|
"step": 13270,
|
|
"valid_targets_mean": 2941.5,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 6.953902566788894,
|
|
"grad_norm": 0.4548474365698735,
|
|
"learning_rate": 5.405275027765555e-09,
|
|
"loss": 0.0739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04574700817465782,
|
|
"step": 13275,
|
|
"valid_targets_mean": 2359.2,
|
|
"valid_targets_min": 865
|
|
},
|
|
{
|
|
"epoch": 6.956521739130435,
|
|
"grad_norm": 0.48640228468480634,
|
|
"learning_rate": 4.815024392552659e-09,
|
|
"loss": 0.1072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030867867171764374,
|
|
"step": 13280,
|
|
"valid_targets_mean": 1747.0,
|
|
"valid_targets_min": 570
|
|
},
|
|
{
|
|
"epoch": 6.959140911471975,
|
|
"grad_norm": 0.41893656806678314,
|
|
"learning_rate": 4.258886977683841e-09,
|
|
"loss": 0.0894,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04409385845065117,
|
|
"step": 13285,
|
|
"valid_targets_mean": 3704.2,
|
|
"valid_targets_min": 2811
|
|
},
|
|
{
|
|
"epoch": 6.961760083813515,
|
|
"grad_norm": 0.42153506890974424,
|
|
"learning_rate": 3.7368637319690204e-09,
|
|
"loss": 0.0712,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031005047261714935,
|
|
"step": 13290,
|
|
"valid_targets_mean": 1749.1,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 6.9643792561550555,
|
|
"grad_norm": 0.44756068678422617,
|
|
"learning_rate": 3.2489555460157863e-09,
|
|
"loss": 0.0699,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0354573056101799,
|
|
"step": 13295,
|
|
"valid_targets_mean": 2985.1,
|
|
"valid_targets_min": 1591
|
|
},
|
|
{
|
|
"epoch": 6.966998428496595,
|
|
"grad_norm": 0.4244297106292505,
|
|
"learning_rate": 2.795163252231614e-09,
|
|
"loss": 0.0686,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030690640211105347,
|
|
"step": 13300,
|
|
"valid_targets_mean": 2912.6,
|
|
"valid_targets_min": 800
|
|
},
|
|
{
|
|
"epoch": 6.969617600838135,
|
|
"grad_norm": 0.36870229760270784,
|
|
"learning_rate": 2.3754876248216484e-09,
|
|
"loss": 0.0621,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033027928322553635,
|
|
"step": 13305,
|
|
"valid_targets_mean": 3084.9,
|
|
"valid_targets_min": 860
|
|
},
|
|
{
|
|
"epoch": 6.9722367731796755,
|
|
"grad_norm": 0.5218797174304519,
|
|
"learning_rate": 1.9899293797753792e-09,
|
|
"loss": 0.0865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03889650106430054,
|
|
"step": 13310,
|
|
"valid_targets_mean": 2482.6,
|
|
"valid_targets_min": 878
|
|
},
|
|
{
|
|
"epoch": 6.974855945521215,
|
|
"grad_norm": 0.3670369079177163,
|
|
"learning_rate": 1.6384891748910669e-09,
|
|
"loss": 0.0615,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03365497291088104,
|
|
"step": 13315,
|
|
"valid_targets_mean": 3403.2,
|
|
"valid_targets_min": 1024
|
|
},
|
|
{
|
|
"epoch": 6.977475117862755,
|
|
"grad_norm": 0.44160164601539337,
|
|
"learning_rate": 1.3211676097446558e-09,
|
|
"loss": 0.0681,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034843601286411285,
|
|
"step": 13320,
|
|
"valid_targets_mean": 2173.1,
|
|
"valid_targets_min": 786
|
|
},
|
|
{
|
|
"epoch": 6.9800942902042955,
|
|
"grad_norm": 0.43451913090030936,
|
|
"learning_rate": 1.0379652257141991e-09,
|
|
"loss": 0.0697,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0458085760474205,
|
|
"step": 13325,
|
|
"valid_targets_mean": 3233.9,
|
|
"valid_targets_min": 2573
|
|
},
|
|
{
|
|
"epoch": 6.982713462545836,
|
|
"grad_norm": 0.3746873455054217,
|
|
"learning_rate": 7.888825059576555e-10,
|
|
"loss": 0.0711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03752672299742699,
|
|
"step": 13330,
|
|
"valid_targets_mean": 3821.9,
|
|
"valid_targets_min": 2274
|
|
},
|
|
{
|
|
"epoch": 6.985332634887375,
|
|
"grad_norm": 0.4241442188246108,
|
|
"learning_rate": 5.739198754350916e-10,
|
|
"loss": 0.0626,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039890218526124954,
|
|
"step": 13335,
|
|
"valid_targets_mean": 3791.2,
|
|
"valid_targets_min": 892
|
|
},
|
|
{
|
|
"epoch": 6.9879518072289155,
|
|
"grad_norm": 0.3639122279373955,
|
|
"learning_rate": 3.9307770088203855e-10,
|
|
"loss": 0.0675,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027569664642214775,
|
|
"step": 13340,
|
|
"valid_targets_mean": 4820.4,
|
|
"valid_targets_min": 4491
|
|
},
|
|
{
|
|
"epoch": 6.990570979570456,
|
|
"grad_norm": 0.6518008964117099,
|
|
"learning_rate": 2.463562908316952e-10,
|
|
"loss": 0.1068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08742758631706238,
|
|
"step": 13345,
|
|
"valid_targets_mean": 2159.1,
|
|
"valid_targets_min": 709
|
|
},
|
|
{
|
|
"epoch": 6.993190151911996,
|
|
"grad_norm": 0.33525197631159753,
|
|
"learning_rate": 1.3375589559938562e-10,
|
|
"loss": 0.1183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.021990729495882988,
|
|
"step": 13350,
|
|
"valid_targets_mean": 3430.8,
|
|
"valid_targets_min": 1097
|
|
},
|
|
{
|
|
"epoch": 6.9958093242535355,
|
|
"grad_norm": 0.3826266404168604,
|
|
"learning_rate": 5.5276707291440637e-11,
|
|
"loss": 0.0714,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029649188742041588,
|
|
"step": 13355,
|
|
"valid_targets_mean": 3032.0,
|
|
"valid_targets_min": 840
|
|
},
|
|
{
|
|
"epoch": 6.998428496595076,
|
|
"grad_norm": 0.3197481698629408,
|
|
"learning_rate": 1.0918859796316127e-11,
|
|
"loss": 0.0653,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026989787817001343,
|
|
"step": 13360,
|
|
"valid_targets_mean": 3812.2,
|
|
"valid_targets_min": 3106
|
|
},
|
|
{
|
|
"epoch": 7.0,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029614586383104324,
|
|
"step": 13363,
|
|
"total_flos": 4.5218040599627694e+18,
|
|
"train_loss": 0.07040505743371078,
|
|
"train_runtime": 148647.7084,
|
|
"train_samples_per_second": 1.438,
|
|
"train_steps_per_second": 0.09,
|
|
"valid_targets_mean": 2983.1,
|
|
"valid_targets_min": 791
|
|
}
|
|
],
|
|
"logging_steps": 5,
|
|
"max_steps": 13363,
|
|
"num_input_tokens_seen": 0,
|
|
"num_train_epochs": 7,
|
|
"save_steps": 200,
|
|
"stateful_callbacks": {
|
|
"TrainerControl": {
|
|
"args": {
|
|
"should_epoch_stop": false,
|
|
"should_evaluate": false,
|
|
"should_log": false,
|
|
"should_save": true,
|
|
"should_training_stop": true
|
|
},
|
|
"attributes": {}
|
|
}
|
|
},
|
|
"total_flos": 4.5218040599627694e+18,
|
|
"train_batch_size": 1,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|