GaetanMichelet's picture
Model save
635b3b4 verified
{
"best_metric": 0.8656427264213562,
"best_model_checkpoint": "data/Llama-31-8B_task-1_120-samples_config-1_full/checkpoint-77",
"epoch": 14.0,
"eval_steps": 500,
"global_step": 154,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.09090909090909091,
"grad_norm": 2.0200283527374268,
"learning_rate": 1.818181818181818e-06,
"loss": 2.4844,
"step": 1
},
{
"epoch": 0.18181818181818182,
"grad_norm": 1.8869131803512573,
"learning_rate": 3.636363636363636e-06,
"loss": 2.5082,
"step": 2
},
{
"epoch": 0.36363636363636365,
"grad_norm": 1.758622169494629,
"learning_rate": 7.272727272727272e-06,
"loss": 2.5059,
"step": 4
},
{
"epoch": 0.5454545454545454,
"grad_norm": 1.6956913471221924,
"learning_rate": 1.0909090909090909e-05,
"loss": 2.4567,
"step": 6
},
{
"epoch": 0.7272727272727273,
"grad_norm": 1.5868192911148071,
"learning_rate": 1.4545454545454545e-05,
"loss": 2.3684,
"step": 8
},
{
"epoch": 0.9090909090909091,
"grad_norm": 1.787414312362671,
"learning_rate": 1.8181818181818182e-05,
"loss": 2.3371,
"step": 10
},
{
"epoch": 1.0,
"eval_loss": 2.238572835922241,
"eval_runtime": 9.6136,
"eval_samples_per_second": 2.496,
"eval_steps_per_second": 2.496,
"step": 11
},
{
"epoch": 1.0909090909090908,
"grad_norm": 1.3648227453231812,
"learning_rate": 2.1818181818181818e-05,
"loss": 2.2684,
"step": 12
},
{
"epoch": 1.2727272727272727,
"grad_norm": 1.2888587713241577,
"learning_rate": 2.5454545454545454e-05,
"loss": 2.2457,
"step": 14
},
{
"epoch": 1.4545454545454546,
"grad_norm": 1.0789052248001099,
"learning_rate": 2.909090909090909e-05,
"loss": 2.0599,
"step": 16
},
{
"epoch": 1.6363636363636362,
"grad_norm": 1.7428549528121948,
"learning_rate": 3.272727272727273e-05,
"loss": 2.0556,
"step": 18
},
{
"epoch": 1.8181818181818183,
"grad_norm": 0.905511736869812,
"learning_rate": 3.6363636363636364e-05,
"loss": 1.919,
"step": 20
},
{
"epoch": 2.0,
"grad_norm": 0.7128010988235474,
"learning_rate": 4e-05,
"loss": 1.7745,
"step": 22
},
{
"epoch": 2.0,
"eval_loss": 1.7576960325241089,
"eval_runtime": 9.615,
"eval_samples_per_second": 2.496,
"eval_steps_per_second": 2.496,
"step": 22
},
{
"epoch": 2.1818181818181817,
"grad_norm": 0.8917616605758667,
"learning_rate": 4.3636363636363636e-05,
"loss": 1.6951,
"step": 24
},
{
"epoch": 2.3636363636363638,
"grad_norm": 0.9202045798301697,
"learning_rate": 4.7272727272727275e-05,
"loss": 1.6476,
"step": 26
},
{
"epoch": 2.5454545454545454,
"grad_norm": 1.1835784912109375,
"learning_rate": 5.090909090909091e-05,
"loss": 1.4571,
"step": 28
},
{
"epoch": 2.7272727272727275,
"grad_norm": 1.0429062843322754,
"learning_rate": 5.4545454545454546e-05,
"loss": 1.3454,
"step": 30
},
{
"epoch": 2.909090909090909,
"grad_norm": 0.8713496327400208,
"learning_rate": 5.818181818181818e-05,
"loss": 1.2294,
"step": 32
},
{
"epoch": 3.0,
"eval_loss": 1.1298459768295288,
"eval_runtime": 9.6241,
"eval_samples_per_second": 2.494,
"eval_steps_per_second": 2.494,
"step": 33
},
{
"epoch": 3.090909090909091,
"grad_norm": 0.6041619777679443,
"learning_rate": 6.181818181818182e-05,
"loss": 1.1845,
"step": 34
},
{
"epoch": 3.2727272727272725,
"grad_norm": 0.5593976974487305,
"learning_rate": 6.545454545454546e-05,
"loss": 1.0476,
"step": 36
},
{
"epoch": 3.4545454545454546,
"grad_norm": 0.5526394248008728,
"learning_rate": 6.90909090909091e-05,
"loss": 0.989,
"step": 38
},
{
"epoch": 3.6363636363636362,
"grad_norm": 0.49133172631263733,
"learning_rate": 7.272727272727273e-05,
"loss": 1.0105,
"step": 40
},
{
"epoch": 3.8181818181818183,
"grad_norm": 0.5152122378349304,
"learning_rate": 7.636363636363637e-05,
"loss": 1.0186,
"step": 42
},
{
"epoch": 4.0,
"grad_norm": 0.4263210594654083,
"learning_rate": 8e-05,
"loss": 0.9781,
"step": 44
},
{
"epoch": 4.0,
"eval_loss": 0.9825031757354736,
"eval_runtime": 9.6255,
"eval_samples_per_second": 2.493,
"eval_steps_per_second": 2.493,
"step": 44
},
{
"epoch": 4.181818181818182,
"grad_norm": 0.4594288170337677,
"learning_rate": 8.363636363636364e-05,
"loss": 0.9387,
"step": 46
},
{
"epoch": 4.363636363636363,
"grad_norm": 0.5193449854850769,
"learning_rate": 8.727272727272727e-05,
"loss": 0.9481,
"step": 48
},
{
"epoch": 4.545454545454545,
"grad_norm": 0.5140848159790039,
"learning_rate": 9.090909090909092e-05,
"loss": 0.932,
"step": 50
},
{
"epoch": 4.7272727272727275,
"grad_norm": 0.4637601673603058,
"learning_rate": 9.454545454545455e-05,
"loss": 0.9293,
"step": 52
},
{
"epoch": 4.909090909090909,
"grad_norm": 0.5280166864395142,
"learning_rate": 9.818181818181818e-05,
"loss": 0.8705,
"step": 54
},
{
"epoch": 5.0,
"eval_loss": 0.9252349734306335,
"eval_runtime": 9.6225,
"eval_samples_per_second": 2.494,
"eval_steps_per_second": 2.494,
"step": 55
},
{
"epoch": 5.090909090909091,
"grad_norm": 0.5434139370918274,
"learning_rate": 9.999899300364532e-05,
"loss": 0.8527,
"step": 56
},
{
"epoch": 5.2727272727272725,
"grad_norm": 0.5284571647644043,
"learning_rate": 9.99909372761763e-05,
"loss": 0.8431,
"step": 58
},
{
"epoch": 5.454545454545454,
"grad_norm": 0.5927901268005371,
"learning_rate": 9.997482711915927e-05,
"loss": 0.7975,
"step": 60
},
{
"epoch": 5.636363636363637,
"grad_norm": 0.622176468372345,
"learning_rate": 9.99506651282272e-05,
"loss": 0.8844,
"step": 62
},
{
"epoch": 5.818181818181818,
"grad_norm": 0.7427988648414612,
"learning_rate": 9.991845519630678e-05,
"loss": 0.7823,
"step": 64
},
{
"epoch": 6.0,
"grad_norm": 0.7913317084312439,
"learning_rate": 9.987820251299122e-05,
"loss": 0.8269,
"step": 66
},
{
"epoch": 6.0,
"eval_loss": 0.884376049041748,
"eval_runtime": 9.6171,
"eval_samples_per_second": 2.496,
"eval_steps_per_second": 2.496,
"step": 66
},
{
"epoch": 6.181818181818182,
"grad_norm": 0.7040154337882996,
"learning_rate": 9.982991356370404e-05,
"loss": 0.7398,
"step": 68
},
{
"epoch": 6.363636363636363,
"grad_norm": 0.7483051419258118,
"learning_rate": 9.977359612865423e-05,
"loss": 0.7724,
"step": 70
},
{
"epoch": 6.545454545454545,
"grad_norm": 0.5491539239883423,
"learning_rate": 9.970925928158274e-05,
"loss": 0.7608,
"step": 72
},
{
"epoch": 6.7272727272727275,
"grad_norm": 0.6535021066665649,
"learning_rate": 9.963691338830044e-05,
"loss": 0.7548,
"step": 74
},
{
"epoch": 6.909090909090909,
"grad_norm": 0.686294436454773,
"learning_rate": 9.955657010501806e-05,
"loss": 0.6627,
"step": 76
},
{
"epoch": 7.0,
"eval_loss": 0.8656427264213562,
"eval_runtime": 9.62,
"eval_samples_per_second": 2.495,
"eval_steps_per_second": 2.495,
"step": 77
},
{
"epoch": 7.090909090909091,
"grad_norm": 0.6294690370559692,
"learning_rate": 9.946824237646824e-05,
"loss": 0.6067,
"step": 78
},
{
"epoch": 7.2727272727272725,
"grad_norm": 0.732207179069519,
"learning_rate": 9.937194443381972e-05,
"loss": 0.6866,
"step": 80
},
{
"epoch": 7.454545454545454,
"grad_norm": 0.6583709716796875,
"learning_rate": 9.926769179238466e-05,
"loss": 0.5795,
"step": 82
},
{
"epoch": 7.636363636363637,
"grad_norm": 0.6547601222991943,
"learning_rate": 9.915550124911866e-05,
"loss": 0.5979,
"step": 84
},
{
"epoch": 7.818181818181818,
"grad_norm": 0.6767302751541138,
"learning_rate": 9.903539087991462e-05,
"loss": 0.6686,
"step": 86
},
{
"epoch": 8.0,
"grad_norm": 0.8115944862365723,
"learning_rate": 9.890738003669029e-05,
"loss": 0.6053,
"step": 88
},
{
"epoch": 8.0,
"eval_loss": 0.868882417678833,
"eval_runtime": 9.6154,
"eval_samples_per_second": 2.496,
"eval_steps_per_second": 2.496,
"step": 88
},
{
"epoch": 8.181818181818182,
"grad_norm": 0.7251437902450562,
"learning_rate": 9.877148934427037e-05,
"loss": 0.5444,
"step": 90
},
{
"epoch": 8.363636363636363,
"grad_norm": 0.9013553261756897,
"learning_rate": 9.862774069706346e-05,
"loss": 0.5037,
"step": 92
},
{
"epoch": 8.545454545454545,
"grad_norm": 0.9543907046318054,
"learning_rate": 9.847615725553456e-05,
"loss": 0.4884,
"step": 94
},
{
"epoch": 8.727272727272727,
"grad_norm": 0.8602244853973389,
"learning_rate": 9.831676344247342e-05,
"loss": 0.4928,
"step": 96
},
{
"epoch": 8.909090909090908,
"grad_norm": 0.9116283059120178,
"learning_rate": 9.814958493905963e-05,
"loss": 0.5258,
"step": 98
},
{
"epoch": 9.0,
"eval_loss": 0.930729866027832,
"eval_runtime": 9.6169,
"eval_samples_per_second": 2.496,
"eval_steps_per_second": 2.496,
"step": 99
},
{
"epoch": 9.090909090909092,
"grad_norm": 0.8587744832038879,
"learning_rate": 9.797464868072488e-05,
"loss": 0.4467,
"step": 100
},
{
"epoch": 9.272727272727273,
"grad_norm": 1.011325478553772,
"learning_rate": 9.779198285281325e-05,
"loss": 0.3718,
"step": 102
},
{
"epoch": 9.454545454545455,
"grad_norm": 1.561611294746399,
"learning_rate": 9.760161688604008e-05,
"loss": 0.411,
"step": 104
},
{
"epoch": 9.636363636363637,
"grad_norm": 1.0310885906219482,
"learning_rate": 9.740358145174998e-05,
"loss": 0.3617,
"step": 106
},
{
"epoch": 9.818181818181818,
"grad_norm": 0.9703785181045532,
"learning_rate": 9.719790845697533e-05,
"loss": 0.3618,
"step": 108
},
{
"epoch": 10.0,
"grad_norm": 1.0072264671325684,
"learning_rate": 9.698463103929542e-05,
"loss": 0.3668,
"step": 110
},
{
"epoch": 10.0,
"eval_loss": 1.0742104053497314,
"eval_runtime": 9.6189,
"eval_samples_per_second": 2.495,
"eval_steps_per_second": 2.495,
"step": 110
},
{
"epoch": 10.181818181818182,
"grad_norm": 1.1375013589859009,
"learning_rate": 9.676378356149734e-05,
"loss": 0.2664,
"step": 112
},
{
"epoch": 10.363636363636363,
"grad_norm": 1.377677083015442,
"learning_rate": 9.653540160603956e-05,
"loss": 0.2831,
"step": 114
},
{
"epoch": 10.545454545454545,
"grad_norm": 1.225994348526001,
"learning_rate": 9.629952196931901e-05,
"loss": 0.2456,
"step": 116
},
{
"epoch": 10.727272727272727,
"grad_norm": 1.329970359802246,
"learning_rate": 9.60561826557425e-05,
"loss": 0.2814,
"step": 118
},
{
"epoch": 10.909090909090908,
"grad_norm": 1.3457225561141968,
"learning_rate": 9.580542287160348e-05,
"loss": 0.248,
"step": 120
},
{
"epoch": 11.0,
"eval_loss": 1.2238476276397705,
"eval_runtime": 9.624,
"eval_samples_per_second": 2.494,
"eval_steps_per_second": 2.494,
"step": 121
},
{
"epoch": 11.090909090909092,
"grad_norm": 1.0499111413955688,
"learning_rate": 9.554728301876526e-05,
"loss": 0.1828,
"step": 122
},
{
"epoch": 11.272727272727273,
"grad_norm": 1.2574657201766968,
"learning_rate": 9.528180468815155e-05,
"loss": 0.1455,
"step": 124
},
{
"epoch": 11.454545454545455,
"grad_norm": 1.8661640882492065,
"learning_rate": 9.50090306530454e-05,
"loss": 0.1843,
"step": 126
},
{
"epoch": 11.636363636363637,
"grad_norm": 1.4496153593063354,
"learning_rate": 9.472900486219769e-05,
"loss": 0.181,
"step": 128
},
{
"epoch": 11.818181818181818,
"grad_norm": 1.520456075668335,
"learning_rate": 9.444177243274618e-05,
"loss": 0.1556,
"step": 130
},
{
"epoch": 12.0,
"grad_norm": 1.3242138624191284,
"learning_rate": 9.414737964294636e-05,
"loss": 0.2039,
"step": 132
},
{
"epoch": 12.0,
"eval_loss": 1.3416190147399902,
"eval_runtime": 9.6192,
"eval_samples_per_second": 2.495,
"eval_steps_per_second": 2.495,
"step": 132
},
{
"epoch": 12.181818181818182,
"grad_norm": 1.0514355897903442,
"learning_rate": 9.384587392471515e-05,
"loss": 0.1147,
"step": 134
},
{
"epoch": 12.363636363636363,
"grad_norm": 1.0401465892791748,
"learning_rate": 9.353730385598887e-05,
"loss": 0.1134,
"step": 136
},
{
"epoch": 12.545454545454545,
"grad_norm": 1.2559245824813843,
"learning_rate": 9.322171915289635e-05,
"loss": 0.1241,
"step": 138
},
{
"epoch": 12.727272727272727,
"grad_norm": 1.3274048566818237,
"learning_rate": 9.289917066174886e-05,
"loss": 0.1263,
"step": 140
},
{
"epoch": 12.909090909090908,
"grad_norm": 1.2401025295257568,
"learning_rate": 9.256971035084785e-05,
"loss": 0.1357,
"step": 142
},
{
"epoch": 13.0,
"eval_loss": 1.4582844972610474,
"eval_runtime": 9.6219,
"eval_samples_per_second": 2.494,
"eval_steps_per_second": 2.494,
"step": 143
},
{
"epoch": 13.090909090909092,
"grad_norm": 1.0610545873641968,
"learning_rate": 9.223339130211192e-05,
"loss": 0.0999,
"step": 144
},
{
"epoch": 13.272727272727273,
"grad_norm": 1.0816364288330078,
"learning_rate": 9.189026770252436e-05,
"loss": 0.0877,
"step": 146
},
{
"epoch": 13.454545454545455,
"grad_norm": 0.9884896874427795,
"learning_rate": 9.154039483540273e-05,
"loss": 0.0952,
"step": 148
},
{
"epoch": 13.636363636363637,
"grad_norm": 1.4226878881454468,
"learning_rate": 9.118382907149165e-05,
"loss": 0.0921,
"step": 150
},
{
"epoch": 13.818181818181818,
"grad_norm": 1.2806740999221802,
"learning_rate": 9.082062785988049e-05,
"loss": 0.0858,
"step": 152
},
{
"epoch": 14.0,
"grad_norm": 1.4139348268508911,
"learning_rate": 9.045084971874738e-05,
"loss": 0.1016,
"step": 154
},
{
"epoch": 14.0,
"eval_loss": 1.5564292669296265,
"eval_runtime": 9.6208,
"eval_samples_per_second": 2.495,
"eval_steps_per_second": 2.495,
"step": 154
},
{
"epoch": 14.0,
"step": 154,
"total_flos": 1.8910971254800384e+16,
"train_loss": 0.8254564341593098,
"train_runtime": 1444.0259,
"train_samples_per_second": 3.047,
"train_steps_per_second": 0.381
}
],
"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": 1.8910971254800384e+16,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}