504 lines
12 KiB
JSON
504 lines
12 KiB
JSON
|
{
|
||
|
|
"best_global_step": null,
|
||
|
|
"best_metric": null,
|
||
|
|
"best_model_checkpoint": null,
|
||
|
|
"epoch": 10.0,
|
||
|
|
"eval_steps": 500,
|
||
|
|
"global_step": 250,
|
||
|
|
"is_hyper_param_search": false,
|
||
|
|
"is_local_process_zero": true,
|
||
|
|
"is_world_process_zero": true,
|
||
|
|
"log_history": [
|
||
|
|
{
|
||
|
|
"epoch": 0,
|
||
|
|
"extraction_strength": 0.05119585323242784,
|
||
|
|
"forget_Q_A_Prob": 0.10887884757248685,
|
||
|
|
"forget_Q_A_ROUGE": 0.27890823225307393,
|
||
|
|
"forget_truth_ratio": 0.6448834990772232,
|
||
|
|
"model_utility": 0.2185219733034594,
|
||
|
|
"privleak": -47.96999999040602,
|
||
|
|
"step": 0
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 0.2,
|
||
|
|
"grad_norm": 0.33203125,
|
||
|
|
"learning_rate": 1.6000000000000001e-06,
|
||
|
|
"loss": 0.0302,
|
||
|
|
"step": 5
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 0.4,
|
||
|
|
"grad_norm": 0.31640625,
|
||
|
|
"learning_rate": 3.6000000000000003e-06,
|
||
|
|
"loss": 0.0295,
|
||
|
|
"step": 10
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 0.6,
|
||
|
|
"grad_norm": 0.2890625,
|
||
|
|
"learning_rate": 5.600000000000001e-06,
|
||
|
|
"loss": 0.0286,
|
||
|
|
"step": 15
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 0.8,
|
||
|
|
"grad_norm": 0.3203125,
|
||
|
|
"learning_rate": 7.600000000000001e-06,
|
||
|
|
"loss": 0.0293,
|
||
|
|
"step": 20
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.0,
|
||
|
|
"grad_norm": 0.1494140625,
|
||
|
|
"learning_rate": 9.600000000000001e-06,
|
||
|
|
"loss": 0.0271,
|
||
|
|
"step": 25
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.0,
|
||
|
|
"extraction_strength": 0.04014118794649353,
|
||
|
|
"forget_Q_A_Prob": 0.044040033797500655,
|
||
|
|
"forget_Q_A_ROUGE": 0.19825707891118127,
|
||
|
|
"forget_truth_ratio": 0.6614321776157142,
|
||
|
|
"model_utility": 0.08605862105946929,
|
||
|
|
"privleak": -22.12749999557451,
|
||
|
|
"step": 25
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.2,
|
||
|
|
"grad_norm": 0.11083984375,
|
||
|
|
"learning_rate": 9.822222222222223e-06,
|
||
|
|
"loss": 0.0254,
|
||
|
|
"step": 30
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.4,
|
||
|
|
"grad_norm": 0.10400390625,
|
||
|
|
"learning_rate": 9.600000000000001e-06,
|
||
|
|
"loss": 0.0254,
|
||
|
|
"step": 35
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.6,
|
||
|
|
"grad_norm": 0.08984375,
|
||
|
|
"learning_rate": 9.377777777777779e-06,
|
||
|
|
"loss": 0.024,
|
||
|
|
"step": 40
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.8,
|
||
|
|
"grad_norm": 0.10205078125,
|
||
|
|
"learning_rate": 9.155555555555557e-06,
|
||
|
|
"loss": 0.0246,
|
||
|
|
"step": 45
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.0,
|
||
|
|
"grad_norm": 0.08544921875,
|
||
|
|
"learning_rate": 8.933333333333333e-06,
|
||
|
|
"loss": 0.0241,
|
||
|
|
"step": 50
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.0,
|
||
|
|
"extraction_strength": 0.03555271291935598,
|
||
|
|
"forget_Q_A_Prob": 0.028171376134705498,
|
||
|
|
"forget_Q_A_ROUGE": 0.18136216120754584,
|
||
|
|
"forget_truth_ratio": 0.6646114327873448,
|
||
|
|
"model_utility": 0.07766678967490327,
|
||
|
|
"privleak": -16.837499996632495,
|
||
|
|
"step": 50
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.2,
|
||
|
|
"grad_norm": 0.087890625,
|
||
|
|
"learning_rate": 8.711111111111111e-06,
|
||
|
|
"loss": 0.0242,
|
||
|
|
"step": 55
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.4,
|
||
|
|
"grad_norm": 0.10595703125,
|
||
|
|
"learning_rate": 8.48888888888889e-06,
|
||
|
|
"loss": 0.0243,
|
||
|
|
"step": 60
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.6,
|
||
|
|
"grad_norm": 0.09033203125,
|
||
|
|
"learning_rate": 8.266666666666667e-06,
|
||
|
|
"loss": 0.024,
|
||
|
|
"step": 65
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.8,
|
||
|
|
"grad_norm": 0.0859375,
|
||
|
|
"learning_rate": 8.044444444444444e-06,
|
||
|
|
"loss": 0.0236,
|
||
|
|
"step": 70
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.0,
|
||
|
|
"grad_norm": 0.0791015625,
|
||
|
|
"learning_rate": 7.822222222222224e-06,
|
||
|
|
"loss": 0.0228,
|
||
|
|
"step": 75
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.0,
|
||
|
|
"extraction_strength": 0.03542379317218397,
|
||
|
|
"forget_Q_A_Prob": 0.025599835189350415,
|
||
|
|
"forget_Q_A_ROUGE": 0.17665323136017774,
|
||
|
|
"forget_truth_ratio": 0.666956633608444,
|
||
|
|
"model_utility": 0.025169849004668194,
|
||
|
|
"privleak": -15.60999999687799,
|
||
|
|
"step": 75
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.2,
|
||
|
|
"grad_norm": 0.083984375,
|
||
|
|
"learning_rate": 7.600000000000001e-06,
|
||
|
|
"loss": 0.023,
|
||
|
|
"step": 80
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.4,
|
||
|
|
"grad_norm": 0.07958984375,
|
||
|
|
"learning_rate": 7.377777777777778e-06,
|
||
|
|
"loss": 0.0228,
|
||
|
|
"step": 85
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.6,
|
||
|
|
"grad_norm": 0.08544921875,
|
||
|
|
"learning_rate": 7.155555555555556e-06,
|
||
|
|
"loss": 0.0236,
|
||
|
|
"step": 90
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.8,
|
||
|
|
"grad_norm": 0.07763671875,
|
||
|
|
"learning_rate": 6.9333333333333344e-06,
|
||
|
|
"loss": 0.0235,
|
||
|
|
"step": 95
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.0,
|
||
|
|
"grad_norm": 0.08251953125,
|
||
|
|
"learning_rate": 6.711111111111111e-06,
|
||
|
|
"loss": 0.0233,
|
||
|
|
"step": 100
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.0,
|
||
|
|
"extraction_strength": 0.034881082155764,
|
||
|
|
"forget_Q_A_Prob": 0.024368888099124887,
|
||
|
|
"forget_Q_A_ROUGE": 0.17340090189465512,
|
||
|
|
"forget_truth_ratio": 0.6650891487506957,
|
||
|
|
"model_utility": 0.07572111776600261,
|
||
|
|
"privleak": -15.054999996988997,
|
||
|
|
"step": 100
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.2,
|
||
|
|
"grad_norm": 0.08544921875,
|
||
|
|
"learning_rate": 6.488888888888889e-06,
|
||
|
|
"loss": 0.0243,
|
||
|
|
"step": 105
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.4,
|
||
|
|
"grad_norm": 0.08154296875,
|
||
|
|
"learning_rate": 6.266666666666668e-06,
|
||
|
|
"loss": 0.0231,
|
||
|
|
"step": 110
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.6,
|
||
|
|
"grad_norm": 0.08056640625,
|
||
|
|
"learning_rate": 6.044444444444445e-06,
|
||
|
|
"loss": 0.0231,
|
||
|
|
"step": 115
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.8,
|
||
|
|
"grad_norm": 0.080078125,
|
||
|
|
"learning_rate": 5.822222222222223e-06,
|
||
|
|
"loss": 0.0231,
|
||
|
|
"step": 120
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.0,
|
||
|
|
"grad_norm": 0.0810546875,
|
||
|
|
"learning_rate": 5.600000000000001e-06,
|
||
|
|
"loss": 0.0225,
|
||
|
|
"step": 125
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.0,
|
||
|
|
"extraction_strength": 0.03492969775585019,
|
||
|
|
"forget_Q_A_Prob": 0.023090821445148322,
|
||
|
|
"forget_Q_A_ROUGE": 0.17290647721627245,
|
||
|
|
"forget_truth_ratio": 0.6653264816436281,
|
||
|
|
"model_utility": 0.061282254233379956,
|
||
|
|
"privleak": -14.518749997096252,
|
||
|
|
"step": 125
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.2,
|
||
|
|
"grad_norm": 0.0810546875,
|
||
|
|
"learning_rate": 5.3777777777777784e-06,
|
||
|
|
"loss": 0.0235,
|
||
|
|
"step": 130
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.4,
|
||
|
|
"grad_norm": 0.08203125,
|
||
|
|
"learning_rate": 5.155555555555556e-06,
|
||
|
|
"loss": 0.0227,
|
||
|
|
"step": 135
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.6,
|
||
|
|
"grad_norm": 0.09619140625,
|
||
|
|
"learning_rate": 4.933333333333334e-06,
|
||
|
|
"loss": 0.0235,
|
||
|
|
"step": 140
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.8,
|
||
|
|
"grad_norm": 0.0947265625,
|
||
|
|
"learning_rate": 4.711111111111111e-06,
|
||
|
|
"loss": 0.0245,
|
||
|
|
"step": 145
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.0,
|
||
|
|
"grad_norm": 0.0849609375,
|
||
|
|
"learning_rate": 4.488888888888889e-06,
|
||
|
|
"loss": 0.0225,
|
||
|
|
"step": 150
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.0,
|
||
|
|
"extraction_strength": 0.034710650136802565,
|
||
|
|
"forget_Q_A_Prob": 0.022637412079275235,
|
||
|
|
"forget_Q_A_ROUGE": 0.16969368902404242,
|
||
|
|
"forget_truth_ratio": 0.665408632731608,
|
||
|
|
"model_utility": 0.07293884758887383,
|
||
|
|
"privleak": -14.298749997140254,
|
||
|
|
"step": 150
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.2,
|
||
|
|
"grad_norm": 0.07861328125,
|
||
|
|
"learning_rate": 4.266666666666668e-06,
|
||
|
|
"loss": 0.0239,
|
||
|
|
"step": 155
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.4,
|
||
|
|
"grad_norm": 0.08447265625,
|
||
|
|
"learning_rate": 4.044444444444445e-06,
|
||
|
|
"loss": 0.0234,
|
||
|
|
"step": 160
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.6,
|
||
|
|
"grad_norm": 0.0986328125,
|
||
|
|
"learning_rate": 3.8222222222222224e-06,
|
||
|
|
"loss": 0.0236,
|
||
|
|
"step": 165
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.8,
|
||
|
|
"grad_norm": 0.091796875,
|
||
|
|
"learning_rate": 3.6000000000000003e-06,
|
||
|
|
"loss": 0.023,
|
||
|
|
"step": 170
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.0,
|
||
|
|
"grad_norm": 0.0791015625,
|
||
|
|
"learning_rate": 3.377777777777778e-06,
|
||
|
|
"loss": 0.0231,
|
||
|
|
"step": 175
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.0,
|
||
|
|
"extraction_strength": 0.03454297434412678,
|
||
|
|
"forget_Q_A_Prob": 0.022689154595791478,
|
||
|
|
"forget_Q_A_ROUGE": 0.17071292253126477,
|
||
|
|
"forget_truth_ratio": 0.6661571899518759,
|
||
|
|
"model_utility": 0.024938334488339972,
|
||
|
|
"privleak": -14.50124999709975,
|
||
|
|
"step": 175
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.2,
|
||
|
|
"grad_norm": 0.0859375,
|
||
|
|
"learning_rate": 3.1555555555555555e-06,
|
||
|
|
"loss": 0.0234,
|
||
|
|
"step": 180
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.4,
|
||
|
|
"grad_norm": 0.083984375,
|
||
|
|
"learning_rate": 2.9333333333333338e-06,
|
||
|
|
"loss": 0.0229,
|
||
|
|
"step": 185
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.6,
|
||
|
|
"grad_norm": 0.0849609375,
|
||
|
|
"learning_rate": 2.7111111111111116e-06,
|
||
|
|
"loss": 0.0231,
|
||
|
|
"step": 190
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.8,
|
||
|
|
"grad_norm": 0.080078125,
|
||
|
|
"learning_rate": 2.488888888888889e-06,
|
||
|
|
"loss": 0.0234,
|
||
|
|
"step": 195
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.0,
|
||
|
|
"grad_norm": 0.08056640625,
|
||
|
|
"learning_rate": 2.266666666666667e-06,
|
||
|
|
"loss": 0.023,
|
||
|
|
"step": 200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.0,
|
||
|
|
"extraction_strength": 0.03481065013680257,
|
||
|
|
"forget_Q_A_Prob": 0.022478839480972967,
|
||
|
|
"forget_Q_A_ROUGE": 0.17469059451335844,
|
||
|
|
"forget_truth_ratio": 0.6657672475603854,
|
||
|
|
"model_utility": 0.02489956345780598,
|
||
|
|
"privleak": -13.848749997230259,
|
||
|
|
"step": 200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.2,
|
||
|
|
"grad_norm": 0.08251953125,
|
||
|
|
"learning_rate": 2.0444444444444447e-06,
|
||
|
|
"loss": 0.0234,
|
||
|
|
"step": 205
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.4,
|
||
|
|
"grad_norm": 0.0771484375,
|
||
|
|
"learning_rate": 1.8222222222222225e-06,
|
||
|
|
"loss": 0.0243,
|
||
|
|
"step": 210
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.6,
|
||
|
|
"grad_norm": 0.0869140625,
|
||
|
|
"learning_rate": 1.6000000000000001e-06,
|
||
|
|
"loss": 0.0231,
|
||
|
|
"step": 215
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.8,
|
||
|
|
"grad_norm": 0.080078125,
|
||
|
|
"learning_rate": 1.377777777777778e-06,
|
||
|
|
"loss": 0.0227,
|
||
|
|
"step": 220
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.0,
|
||
|
|
"grad_norm": 0.07177734375,
|
||
|
|
"learning_rate": 1.1555555555555556e-06,
|
||
|
|
"loss": 0.0229,
|
||
|
|
"step": 225
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.0,
|
||
|
|
"extraction_strength": 0.03500545533160776,
|
||
|
|
"forget_Q_A_Prob": 0.02249195015472651,
|
||
|
|
"forget_Q_A_ROUGE": 0.16997095669432055,
|
||
|
|
"forget_truth_ratio": 0.6661828793220076,
|
||
|
|
"model_utility": 0.034445471917355024,
|
||
|
|
"privleak": -13.99624999720075,
|
||
|
|
"step": 225
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.2,
|
||
|
|
"grad_norm": 0.076171875,
|
||
|
|
"learning_rate": 9.333333333333334e-07,
|
||
|
|
"loss": 0.0223,
|
||
|
|
"step": 230
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.4,
|
||
|
|
"grad_norm": 0.0791015625,
|
||
|
|
"learning_rate": 7.111111111111112e-07,
|
||
|
|
"loss": 0.0232,
|
||
|
|
"step": 235
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.6,
|
||
|
|
"grad_norm": 0.0849609375,
|
||
|
|
"learning_rate": 4.88888888888889e-07,
|
||
|
|
"loss": 0.023,
|
||
|
|
"step": 240
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.8,
|
||
|
|
"grad_norm": 0.0849609375,
|
||
|
|
"learning_rate": 2.666666666666667e-07,
|
||
|
|
"loss": 0.0235,
|
||
|
|
"step": 245
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 10.0,
|
||
|
|
"grad_norm": 0.0830078125,
|
||
|
|
"learning_rate": 4.444444444444445e-08,
|
||
|
|
"loss": 0.0232,
|
||
|
|
"step": 250
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 10.0,
|
||
|
|
"extraction_strength": 0.034667360093512525,
|
||
|
|
"forget_Q_A_Prob": 0.022312334740927326,
|
||
|
|
"forget_Q_A_ROUGE": 0.16821955009936282,
|
||
|
|
"forget_truth_ratio": 0.6682139209275343,
|
||
|
|
"model_utility": 0.024945798930326533,
|
||
|
|
"privleak": -14.171249997165742,
|
||
|
|
"step": 250
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 10.0,
|
||
|
|
"step": 250,
|
||
|
|
"total_flos": 0.0,
|
||
|
|
"train_loss": 0.0240108642578125,
|
||
|
|
"train_runtime": 5314.6353,
|
||
|
|
"train_samples_per_second": 0.753,
|
||
|
|
"train_steps_per_second": 0.047
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"logging_steps": 5,
|
||
|
|
"max_steps": 250,
|
||
|
|
"num_input_tokens_seen": 0,
|
||
|
|
"num_train_epochs": 10,
|
||
|
|
"save_steps": 500,
|
||
|
|
"stateful_callbacks": {
|
||
|
|
"TrainerControl": {
|
||
|
|
"args": {
|
||
|
|
"should_epoch_stop": false,
|
||
|
|
"should_evaluate": false,
|
||
|
|
"should_log": false,
|
||
|
|
"should_save": false,
|
||
|
|
"should_training_stop": false
|
||
|
|
},
|
||
|
|
"attributes": {}
|
||
|
|
}
|
||
|
|
},
|
||
|
|
"total_flos": 0.0,
|
||
|
|
"train_batch_size": 4,
|
||
|
|
"trial_name": null,
|
||
|
|
"trial_params": null
|
||
|
|
}
|