GaetanMichelet's picture
Model save
aa850c4 verified
{
"best_metric": 0.7838263511657715,
"best_model_checkpoint": "data/Llama-31-8B_task-1_120-samples_config-1_full_auto/checkpoint-88",
"epoch": 15.0,
"eval_steps": 500,
"global_step": 165,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.09090909090909091,
"grad_norm": 1.8327940702438354,
"learning_rate": 1.818181818181818e-06,
"loss": 2.1624,
"step": 1
},
{
"epoch": 0.18181818181818182,
"grad_norm": 1.6178241968154907,
"learning_rate": 3.636363636363636e-06,
"loss": 2.1438,
"step": 2
},
{
"epoch": 0.36363636363636365,
"grad_norm": 1.587786316871643,
"learning_rate": 7.272727272727272e-06,
"loss": 2.2061,
"step": 4
},
{
"epoch": 0.5454545454545454,
"grad_norm": 1.5586106777191162,
"learning_rate": 1.0909090909090909e-05,
"loss": 2.1749,
"step": 6
},
{
"epoch": 0.7272727272727273,
"grad_norm": 1.3981194496154785,
"learning_rate": 1.4545454545454545e-05,
"loss": 2.0761,
"step": 8
},
{
"epoch": 0.9090909090909091,
"grad_norm": 1.656570315361023,
"learning_rate": 1.8181818181818182e-05,
"loss": 2.1444,
"step": 10
},
{
"epoch": 1.0,
"eval_loss": 1.9951225519180298,
"eval_runtime": 9.8854,
"eval_samples_per_second": 2.428,
"eval_steps_per_second": 2.428,
"step": 11
},
{
"epoch": 1.0909090909090908,
"grad_norm": 1.2477445602416992,
"learning_rate": 2.1818181818181818e-05,
"loss": 2.0133,
"step": 12
},
{
"epoch": 1.2727272727272727,
"grad_norm": 1.0866979360580444,
"learning_rate": 2.5454545454545454e-05,
"loss": 1.982,
"step": 14
},
{
"epoch": 1.4545454545454546,
"grad_norm": 1.0384626388549805,
"learning_rate": 2.909090909090909e-05,
"loss": 1.792,
"step": 16
},
{
"epoch": 1.6363636363636362,
"grad_norm": 1.6297812461853027,
"learning_rate": 3.272727272727273e-05,
"loss": 1.8401,
"step": 18
},
{
"epoch": 1.8181818181818183,
"grad_norm": 0.8939803838729858,
"learning_rate": 3.6363636363636364e-05,
"loss": 1.7569,
"step": 20
},
{
"epoch": 2.0,
"grad_norm": 0.6418306231498718,
"learning_rate": 4e-05,
"loss": 1.5597,
"step": 22
},
{
"epoch": 2.0,
"eval_loss": 1.5854285955429077,
"eval_runtime": 9.8853,
"eval_samples_per_second": 2.428,
"eval_steps_per_second": 2.428,
"step": 22
},
{
"epoch": 2.1818181818181817,
"grad_norm": 0.7771486043930054,
"learning_rate": 4.3636363636363636e-05,
"loss": 1.5096,
"step": 24
},
{
"epoch": 2.3636363636363638,
"grad_norm": 0.8559293746948242,
"learning_rate": 4.7272727272727275e-05,
"loss": 1.4798,
"step": 26
},
{
"epoch": 2.5454545454545454,
"grad_norm": 0.8326192498207092,
"learning_rate": 5.090909090909091e-05,
"loss": 1.3439,
"step": 28
},
{
"epoch": 2.7272727272727275,
"grad_norm": 0.9386058449745178,
"learning_rate": 5.4545454545454546e-05,
"loss": 1.2474,
"step": 30
},
{
"epoch": 2.909090909090909,
"grad_norm": 0.9360014200210571,
"learning_rate": 5.818181818181818e-05,
"loss": 1.1135,
"step": 32
},
{
"epoch": 3.0,
"eval_loss": 1.0142416954040527,
"eval_runtime": 9.882,
"eval_samples_per_second": 2.429,
"eval_steps_per_second": 2.429,
"step": 33
},
{
"epoch": 3.090909090909091,
"grad_norm": 0.6011117696762085,
"learning_rate": 6.181818181818182e-05,
"loss": 1.0217,
"step": 34
},
{
"epoch": 3.2727272727272725,
"grad_norm": 0.5689023733139038,
"learning_rate": 6.545454545454546e-05,
"loss": 0.9049,
"step": 36
},
{
"epoch": 3.4545454545454546,
"grad_norm": 0.5020841956138611,
"learning_rate": 6.90909090909091e-05,
"loss": 0.8946,
"step": 38
},
{
"epoch": 3.6363636363636362,
"grad_norm": 0.4844783842563629,
"learning_rate": 7.272727272727273e-05,
"loss": 0.8966,
"step": 40
},
{
"epoch": 3.8181818181818183,
"grad_norm": 0.4568497836589813,
"learning_rate": 7.636363636363637e-05,
"loss": 0.9219,
"step": 42
},
{
"epoch": 4.0,
"grad_norm": 0.4025140106678009,
"learning_rate": 8e-05,
"loss": 0.8595,
"step": 44
},
{
"epoch": 4.0,
"eval_loss": 0.879399299621582,
"eval_runtime": 9.8938,
"eval_samples_per_second": 2.426,
"eval_steps_per_second": 2.426,
"step": 44
},
{
"epoch": 4.181818181818182,
"grad_norm": 0.45321401953697205,
"learning_rate": 8.363636363636364e-05,
"loss": 0.836,
"step": 46
},
{
"epoch": 4.363636363636363,
"grad_norm": 0.5086842775344849,
"learning_rate": 8.727272727272727e-05,
"loss": 0.8666,
"step": 48
},
{
"epoch": 4.545454545454545,
"grad_norm": 0.48371073603630066,
"learning_rate": 9.090909090909092e-05,
"loss": 0.8232,
"step": 50
},
{
"epoch": 4.7272727272727275,
"grad_norm": 0.44395631551742554,
"learning_rate": 9.454545454545455e-05,
"loss": 0.8118,
"step": 52
},
{
"epoch": 4.909090909090909,
"grad_norm": 0.4474985897541046,
"learning_rate": 9.818181818181818e-05,
"loss": 0.7701,
"step": 54
},
{
"epoch": 5.0,
"eval_loss": 0.8356377482414246,
"eval_runtime": 9.8781,
"eval_samples_per_second": 2.43,
"eval_steps_per_second": 2.43,
"step": 55
},
{
"epoch": 5.090909090909091,
"grad_norm": 0.4774426519870758,
"learning_rate": 9.999899300364532e-05,
"loss": 0.7538,
"step": 56
},
{
"epoch": 5.2727272727272725,
"grad_norm": 0.5100274085998535,
"learning_rate": 9.99909372761763e-05,
"loss": 0.7634,
"step": 58
},
{
"epoch": 5.454545454545454,
"grad_norm": 0.5279253125190735,
"learning_rate": 9.997482711915927e-05,
"loss": 0.705,
"step": 60
},
{
"epoch": 5.636363636363637,
"grad_norm": 0.5918782353401184,
"learning_rate": 9.99506651282272e-05,
"loss": 0.773,
"step": 62
},
{
"epoch": 5.818181818181818,
"grad_norm": 0.6757959127426147,
"learning_rate": 9.991845519630678e-05,
"loss": 0.7146,
"step": 64
},
{
"epoch": 6.0,
"grad_norm": 0.6711525321006775,
"learning_rate": 9.987820251299122e-05,
"loss": 0.7434,
"step": 66
},
{
"epoch": 6.0,
"eval_loss": 0.8024230003356934,
"eval_runtime": 9.8828,
"eval_samples_per_second": 2.428,
"eval_steps_per_second": 2.428,
"step": 66
},
{
"epoch": 6.181818181818182,
"grad_norm": 0.7074957489967346,
"learning_rate": 9.982991356370404e-05,
"loss": 0.6427,
"step": 68
},
{
"epoch": 6.363636363636363,
"grad_norm": 0.7052618265151978,
"learning_rate": 9.977359612865423e-05,
"loss": 0.7065,
"step": 70
},
{
"epoch": 6.545454545454545,
"grad_norm": 0.5625373721122742,
"learning_rate": 9.970925928158274e-05,
"loss": 0.7053,
"step": 72
},
{
"epoch": 6.7272727272727275,
"grad_norm": 0.637117862701416,
"learning_rate": 9.963691338830044e-05,
"loss": 0.6585,
"step": 74
},
{
"epoch": 6.909090909090909,
"grad_norm": 0.5640820860862732,
"learning_rate": 9.955657010501806e-05,
"loss": 0.6039,
"step": 76
},
{
"epoch": 7.0,
"eval_loss": 0.7894789576530457,
"eval_runtime": 9.8849,
"eval_samples_per_second": 2.428,
"eval_steps_per_second": 2.428,
"step": 77
},
{
"epoch": 7.090909090909091,
"grad_norm": 0.5558722019195557,
"learning_rate": 9.946824237646824e-05,
"loss": 0.5665,
"step": 78
},
{
"epoch": 7.2727272727272725,
"grad_norm": 0.6826435923576355,
"learning_rate": 9.937194443381972e-05,
"loss": 0.6311,
"step": 80
},
{
"epoch": 7.454545454545454,
"grad_norm": 0.6229597330093384,
"learning_rate": 9.926769179238466e-05,
"loss": 0.5216,
"step": 82
},
{
"epoch": 7.636363636363637,
"grad_norm": 0.6224767565727234,
"learning_rate": 9.915550124911866e-05,
"loss": 0.5337,
"step": 84
},
{
"epoch": 7.818181818181818,
"grad_norm": 0.6049901247024536,
"learning_rate": 9.903539087991462e-05,
"loss": 0.5987,
"step": 86
},
{
"epoch": 8.0,
"grad_norm": 0.7121050357818604,
"learning_rate": 9.890738003669029e-05,
"loss": 0.5441,
"step": 88
},
{
"epoch": 8.0,
"eval_loss": 0.7838263511657715,
"eval_runtime": 9.8893,
"eval_samples_per_second": 2.427,
"eval_steps_per_second": 2.427,
"step": 88
},
{
"epoch": 8.181818181818182,
"grad_norm": 0.659483015537262,
"learning_rate": 9.877148934427037e-05,
"loss": 0.4877,
"step": 90
},
{
"epoch": 8.363636363636363,
"grad_norm": 0.8029683828353882,
"learning_rate": 9.862774069706346e-05,
"loss": 0.4356,
"step": 92
},
{
"epoch": 8.545454545454545,
"grad_norm": 0.951897382736206,
"learning_rate": 9.847615725553456e-05,
"loss": 0.4593,
"step": 94
},
{
"epoch": 8.727272727272727,
"grad_norm": 0.7811606526374817,
"learning_rate": 9.831676344247342e-05,
"loss": 0.4709,
"step": 96
},
{
"epoch": 8.909090909090908,
"grad_norm": 0.8405193090438843,
"learning_rate": 9.814958493905963e-05,
"loss": 0.4965,
"step": 98
},
{
"epoch": 9.0,
"eval_loss": 0.8283455967903137,
"eval_runtime": 9.8784,
"eval_samples_per_second": 2.43,
"eval_steps_per_second": 2.43,
"step": 99
},
{
"epoch": 9.090909090909092,
"grad_norm": 0.7512415647506714,
"learning_rate": 9.797464868072488e-05,
"loss": 0.4094,
"step": 100
},
{
"epoch": 9.272727272727273,
"grad_norm": 0.803386390209198,
"learning_rate": 9.779198285281325e-05,
"loss": 0.3419,
"step": 102
},
{
"epoch": 9.454545454545455,
"grad_norm": 1.6166610717773438,
"learning_rate": 9.760161688604008e-05,
"loss": 0.385,
"step": 104
},
{
"epoch": 9.636363636363637,
"grad_norm": 1.1257920265197754,
"learning_rate": 9.740358145174998e-05,
"loss": 0.3488,
"step": 106
},
{
"epoch": 9.818181818181818,
"grad_norm": 0.9676744341850281,
"learning_rate": 9.719790845697533e-05,
"loss": 0.3473,
"step": 108
},
{
"epoch": 10.0,
"grad_norm": 0.8668843507766724,
"learning_rate": 9.698463103929542e-05,
"loss": 0.353,
"step": 110
},
{
"epoch": 10.0,
"eval_loss": 0.9091915488243103,
"eval_runtime": 9.8801,
"eval_samples_per_second": 2.429,
"eval_steps_per_second": 2.429,
"step": 110
},
{
"epoch": 10.181818181818182,
"grad_norm": 0.9420936703681946,
"learning_rate": 9.676378356149734e-05,
"loss": 0.2767,
"step": 112
},
{
"epoch": 10.363636363636363,
"grad_norm": 1.1771456003189087,
"learning_rate": 9.653540160603956e-05,
"loss": 0.2792,
"step": 114
},
{
"epoch": 10.545454545454545,
"grad_norm": 1.374729037284851,
"learning_rate": 9.629952196931901e-05,
"loss": 0.2421,
"step": 116
},
{
"epoch": 10.727272727272727,
"grad_norm": 1.7256646156311035,
"learning_rate": 9.60561826557425e-05,
"loss": 0.2812,
"step": 118
},
{
"epoch": 10.909090909090908,
"grad_norm": 1.1676362752914429,
"learning_rate": 9.580542287160348e-05,
"loss": 0.2505,
"step": 120
},
{
"epoch": 11.0,
"eval_loss": 1.0033034086227417,
"eval_runtime": 9.8798,
"eval_samples_per_second": 2.429,
"eval_steps_per_second": 2.429,
"step": 121
},
{
"epoch": 11.090909090909092,
"grad_norm": 0.8704199194908142,
"learning_rate": 9.554728301876526e-05,
"loss": 0.1951,
"step": 122
},
{
"epoch": 11.272727272727273,
"grad_norm": 1.0148502588272095,
"learning_rate": 9.528180468815155e-05,
"loss": 0.1705,
"step": 124
},
{
"epoch": 11.454545454545455,
"grad_norm": 1.4075133800506592,
"learning_rate": 9.50090306530454e-05,
"loss": 0.1857,
"step": 126
},
{
"epoch": 11.636363636363637,
"grad_norm": 1.4041534662246704,
"learning_rate": 9.472900486219769e-05,
"loss": 0.1774,
"step": 128
},
{
"epoch": 11.818181818181818,
"grad_norm": 1.2094171047210693,
"learning_rate": 9.444177243274618e-05,
"loss": 0.1672,
"step": 130
},
{
"epoch": 12.0,
"grad_norm": 1.2141450643539429,
"learning_rate": 9.414737964294636e-05,
"loss": 0.2204,
"step": 132
},
{
"epoch": 12.0,
"eval_loss": 1.1737927198410034,
"eval_runtime": 9.888,
"eval_samples_per_second": 2.427,
"eval_steps_per_second": 2.427,
"step": 132
},
{
"epoch": 12.181818181818182,
"grad_norm": 0.8845875263214111,
"learning_rate": 9.384587392471515e-05,
"loss": 0.1202,
"step": 134
},
{
"epoch": 12.363636363636363,
"grad_norm": 0.9842550158500671,
"learning_rate": 9.353730385598887e-05,
"loss": 0.1311,
"step": 136
},
{
"epoch": 12.545454545454545,
"grad_norm": 1.1591681241989136,
"learning_rate": 9.322171915289635e-05,
"loss": 0.1369,
"step": 138
},
{
"epoch": 12.727272727272727,
"grad_norm": 1.1487535238265991,
"learning_rate": 9.289917066174886e-05,
"loss": 0.136,
"step": 140
},
{
"epoch": 12.909090909090908,
"grad_norm": 1.2362055778503418,
"learning_rate": 9.256971035084785e-05,
"loss": 0.1355,
"step": 142
},
{
"epoch": 13.0,
"eval_loss": 1.306993842124939,
"eval_runtime": 9.88,
"eval_samples_per_second": 2.429,
"eval_steps_per_second": 2.429,
"step": 143
},
{
"epoch": 13.090909090909092,
"grad_norm": 0.9263333678245544,
"learning_rate": 9.223339130211192e-05,
"loss": 0.1163,
"step": 144
},
{
"epoch": 13.272727272727273,
"grad_norm": 0.86616051197052,
"learning_rate": 9.189026770252436e-05,
"loss": 0.0827,
"step": 146
},
{
"epoch": 13.454545454545455,
"grad_norm": 1.0412770509719849,
"learning_rate": 9.154039483540273e-05,
"loss": 0.0969,
"step": 148
},
{
"epoch": 13.636363636363637,
"grad_norm": 1.0952099561691284,
"learning_rate": 9.118382907149165e-05,
"loss": 0.0962,
"step": 150
},
{
"epoch": 13.818181818181818,
"grad_norm": 1.0400218963623047,
"learning_rate": 9.082062785988049e-05,
"loss": 0.0901,
"step": 152
},
{
"epoch": 14.0,
"grad_norm": 1.0147318840026855,
"learning_rate": 9.045084971874738e-05,
"loss": 0.1041,
"step": 154
},
{
"epoch": 14.0,
"eval_loss": 1.3560261726379395,
"eval_runtime": 9.8831,
"eval_samples_per_second": 2.428,
"eval_steps_per_second": 2.428,
"step": 154
},
{
"epoch": 14.181818181818182,
"grad_norm": 0.6898658275604248,
"learning_rate": 9.007455422593077e-05,
"loss": 0.0707,
"step": 156
},
{
"epoch": 14.363636363636363,
"grad_norm": 0.9911731481552124,
"learning_rate": 8.969180200933047e-05,
"loss": 0.0775,
"step": 158
},
{
"epoch": 14.545454545454545,
"grad_norm": 0.9983232021331787,
"learning_rate": 8.930265473713938e-05,
"loss": 0.0734,
"step": 160
},
{
"epoch": 14.727272727272727,
"grad_norm": 0.9479069709777832,
"learning_rate": 8.890717510790763e-05,
"loss": 0.0782,
"step": 162
},
{
"epoch": 14.909090909090908,
"grad_norm": 0.7364582419395447,
"learning_rate": 8.850542684044078e-05,
"loss": 0.0759,
"step": 164
},
{
"epoch": 15.0,
"eval_loss": 1.3713493347167969,
"eval_runtime": 9.8872,
"eval_samples_per_second": 2.427,
"eval_steps_per_second": 2.427,
"step": 165
},
{
"epoch": 15.0,
"step": 165,
"total_flos": 2.196659233410253e+16,
"train_loss": 0.7007441697247101,
"train_runtime": 1592.0409,
"train_samples_per_second": 2.764,
"train_steps_per_second": 0.345
}
],
"logging_steps": 2,
"max_steps": 550,
"num_input_tokens_seen": 0,
"num_train_epochs": 50,
"save_steps": 25,
"stateful_callbacks": {
"EarlyStoppingCallback": {
"args": {
"early_stopping_patience": 7,
"early_stopping_threshold": 0.0
},
"attributes": {
"early_stopping_patience_counter": 0
}
},
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 2.196659233410253e+16,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}