GaetanMichelet's picture
Model save
2eca717 verified
{
"best_metric": 0.36825016140937805,
"best_model_checkpoint": "data/Gemma-2-2B_task-3_120-samples_config-1_auto/checkpoint-55",
"epoch": 12.0,
"eval_steps": 500,
"global_step": 132,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.09090909090909091,
"grad_norm": 11.006781578063965,
"learning_rate": 1.818181818181818e-06,
"loss": 4.3957,
"step": 1
},
{
"epoch": 0.18181818181818182,
"grad_norm": 6.361999034881592,
"learning_rate": 3.636363636363636e-06,
"loss": 3.1845,
"step": 2
},
{
"epoch": 0.36363636363636365,
"grad_norm": 6.361944198608398,
"learning_rate": 7.272727272727272e-06,
"loss": 3.3559,
"step": 4
},
{
"epoch": 0.5454545454545454,
"grad_norm": 4.936976909637451,
"learning_rate": 1.0909090909090909e-05,
"loss": 3.1351,
"step": 6
},
{
"epoch": 0.7272727272727273,
"grad_norm": 8.912232398986816,
"learning_rate": 1.4545454545454545e-05,
"loss": 3.0486,
"step": 8
},
{
"epoch": 0.9090909090909091,
"grad_norm": 7.448002815246582,
"learning_rate": 1.8181818181818182e-05,
"loss": 2.9033,
"step": 10
},
{
"epoch": 1.0,
"eval_loss": 2.6048784255981445,
"eval_runtime": 7.7434,
"eval_samples_per_second": 3.099,
"eval_steps_per_second": 3.099,
"step": 11
},
{
"epoch": 1.0909090909090908,
"grad_norm": 6.16674280166626,
"learning_rate": 2.1818181818181818e-05,
"loss": 2.6991,
"step": 12
},
{
"epoch": 1.2727272727272727,
"grad_norm": 5.293051242828369,
"learning_rate": 2.5454545454545454e-05,
"loss": 2.1676,
"step": 14
},
{
"epoch": 1.4545454545454546,
"grad_norm": 3.995722532272339,
"learning_rate": 2.909090909090909e-05,
"loss": 2.153,
"step": 16
},
{
"epoch": 1.6363636363636362,
"grad_norm": 3.376025676727295,
"learning_rate": 3.272727272727273e-05,
"loss": 1.5524,
"step": 18
},
{
"epoch": 1.8181818181818183,
"grad_norm": 2.412623167037964,
"learning_rate": 3.6363636363636364e-05,
"loss": 1.3212,
"step": 20
},
{
"epoch": 2.0,
"grad_norm": 3.236947774887085,
"learning_rate": 4e-05,
"loss": 0.9959,
"step": 22
},
{
"epoch": 2.0,
"eval_loss": 0.9210925102233887,
"eval_runtime": 7.4972,
"eval_samples_per_second": 3.201,
"eval_steps_per_second": 3.201,
"step": 22
},
{
"epoch": 2.1818181818181817,
"grad_norm": 2.2208316326141357,
"learning_rate": 4.3636363636363636e-05,
"loss": 0.7936,
"step": 24
},
{
"epoch": 2.3636363636363638,
"grad_norm": 1.0128132104873657,
"learning_rate": 4.7272727272727275e-05,
"loss": 0.7579,
"step": 26
},
{
"epoch": 2.5454545454545454,
"grad_norm": 1.316047191619873,
"learning_rate": 5.090909090909091e-05,
"loss": 0.6118,
"step": 28
},
{
"epoch": 2.7272727272727275,
"grad_norm": 0.9991164803504944,
"learning_rate": 5.4545454545454546e-05,
"loss": 0.5135,
"step": 30
},
{
"epoch": 2.909090909090909,
"grad_norm": 0.6279535293579102,
"learning_rate": 5.818181818181818e-05,
"loss": 0.3036,
"step": 32
},
{
"epoch": 3.0,
"eval_loss": 0.4791443645954132,
"eval_runtime": 7.5164,
"eval_samples_per_second": 3.193,
"eval_steps_per_second": 3.193,
"step": 33
},
{
"epoch": 3.090909090909091,
"grad_norm": 0.6234776377677917,
"learning_rate": 6.181818181818182e-05,
"loss": 0.5341,
"step": 34
},
{
"epoch": 3.2727272727272725,
"grad_norm": 0.43922051787376404,
"learning_rate": 6.545454545454546e-05,
"loss": 0.1294,
"step": 36
},
{
"epoch": 3.4545454545454546,
"grad_norm": 0.5349417328834534,
"learning_rate": 6.90909090909091e-05,
"loss": 0.427,
"step": 38
},
{
"epoch": 3.6363636363636362,
"grad_norm": 0.663760781288147,
"learning_rate": 7.272727272727273e-05,
"loss": 0.414,
"step": 40
},
{
"epoch": 3.8181818181818183,
"grad_norm": 0.6541147232055664,
"learning_rate": 7.636363636363637e-05,
"loss": 0.5232,
"step": 42
},
{
"epoch": 4.0,
"grad_norm": 0.539898157119751,
"learning_rate": 8e-05,
"loss": 0.3507,
"step": 44
},
{
"epoch": 4.0,
"eval_loss": 0.4003179967403412,
"eval_runtime": 7.5046,
"eval_samples_per_second": 3.198,
"eval_steps_per_second": 3.198,
"step": 44
},
{
"epoch": 4.181818181818182,
"grad_norm": 0.37961480021476746,
"learning_rate": 8.363636363636364e-05,
"loss": 0.2693,
"step": 46
},
{
"epoch": 4.363636363636363,
"grad_norm": 0.6985395550727844,
"learning_rate": 8.727272727272727e-05,
"loss": 0.4497,
"step": 48
},
{
"epoch": 4.545454545454545,
"grad_norm": 0.4456002712249756,
"learning_rate": 9.090909090909092e-05,
"loss": 0.2553,
"step": 50
},
{
"epoch": 4.7272727272727275,
"grad_norm": 0.4576929211616516,
"learning_rate": 9.454545454545455e-05,
"loss": 0.2604,
"step": 52
},
{
"epoch": 4.909090909090909,
"grad_norm": 0.6171721816062927,
"learning_rate": 9.818181818181818e-05,
"loss": 0.2977,
"step": 54
},
{
"epoch": 5.0,
"eval_loss": 0.36825016140937805,
"eval_runtime": 7.4926,
"eval_samples_per_second": 3.203,
"eval_steps_per_second": 3.203,
"step": 55
},
{
"epoch": 5.090909090909091,
"grad_norm": 0.5098957419395447,
"learning_rate": 9.999899300364532e-05,
"loss": 0.3141,
"step": 56
},
{
"epoch": 5.2727272727272725,
"grad_norm": 0.5279871821403503,
"learning_rate": 9.99909372761763e-05,
"loss": 0.2913,
"step": 58
},
{
"epoch": 5.454545454545454,
"grad_norm": 0.4445846378803253,
"learning_rate": 9.997482711915927e-05,
"loss": 0.158,
"step": 60
},
{
"epoch": 5.636363636363637,
"grad_norm": 0.36122986674308777,
"learning_rate": 9.99506651282272e-05,
"loss": 0.2206,
"step": 62
},
{
"epoch": 5.818181818181818,
"grad_norm": 0.5653752088546753,
"learning_rate": 9.991845519630678e-05,
"loss": 0.2277,
"step": 64
},
{
"epoch": 6.0,
"grad_norm": 0.731476902961731,
"learning_rate": 9.987820251299122e-05,
"loss": 0.2739,
"step": 66
},
{
"epoch": 6.0,
"eval_loss": 0.37520870566368103,
"eval_runtime": 7.4907,
"eval_samples_per_second": 3.204,
"eval_steps_per_second": 3.204,
"step": 66
},
{
"epoch": 6.181818181818182,
"grad_norm": 0.5517078638076782,
"learning_rate": 9.982991356370404e-05,
"loss": 0.1487,
"step": 68
},
{
"epoch": 6.363636363636363,
"grad_norm": 0.5384305119514465,
"learning_rate": 9.977359612865423e-05,
"loss": 0.1944,
"step": 70
},
{
"epoch": 6.545454545454545,
"grad_norm": 0.8201135993003845,
"learning_rate": 9.970925928158274e-05,
"loss": 0.2923,
"step": 72
},
{
"epoch": 6.7272727272727275,
"grad_norm": 0.7586239576339722,
"learning_rate": 9.963691338830044e-05,
"loss": 0.1532,
"step": 74
},
{
"epoch": 6.909090909090909,
"grad_norm": 1.3042950630187988,
"learning_rate": 9.955657010501806e-05,
"loss": 0.1787,
"step": 76
},
{
"epoch": 7.0,
"eval_loss": 0.44048985838890076,
"eval_runtime": 7.4988,
"eval_samples_per_second": 3.201,
"eval_steps_per_second": 3.201,
"step": 77
},
{
"epoch": 7.090909090909091,
"grad_norm": 0.9256752729415894,
"learning_rate": 9.946824237646824e-05,
"loss": 0.0959,
"step": 78
},
{
"epoch": 7.2727272727272725,
"grad_norm": 1.0876535177230835,
"learning_rate": 9.937194443381972e-05,
"loss": 0.1981,
"step": 80
},
{
"epoch": 7.454545454545454,
"grad_norm": 0.6428214311599731,
"learning_rate": 9.926769179238466e-05,
"loss": 0.0554,
"step": 82
},
{
"epoch": 7.636363636363637,
"grad_norm": 0.6934786438941956,
"learning_rate": 9.915550124911866e-05,
"loss": 0.0871,
"step": 84
},
{
"epoch": 7.818181818181818,
"grad_norm": 0.7447779178619385,
"learning_rate": 9.903539087991462e-05,
"loss": 0.1279,
"step": 86
},
{
"epoch": 8.0,
"grad_norm": 1.1104106903076172,
"learning_rate": 9.890738003669029e-05,
"loss": 0.1167,
"step": 88
},
{
"epoch": 8.0,
"eval_loss": 0.5092648267745972,
"eval_runtime": 7.4939,
"eval_samples_per_second": 3.203,
"eval_steps_per_second": 3.203,
"step": 88
},
{
"epoch": 8.181818181818182,
"grad_norm": 0.7919206619262695,
"learning_rate": 9.877148934427037e-05,
"loss": 0.0769,
"step": 90
},
{
"epoch": 8.363636363636363,
"grad_norm": 0.772554874420166,
"learning_rate": 9.862774069706346e-05,
"loss": 0.0515,
"step": 92
},
{
"epoch": 8.545454545454545,
"grad_norm": 0.5378653407096863,
"learning_rate": 9.847615725553456e-05,
"loss": 0.038,
"step": 94
},
{
"epoch": 8.727272727272727,
"grad_norm": 0.684589147567749,
"learning_rate": 9.831676344247342e-05,
"loss": 0.0876,
"step": 96
},
{
"epoch": 8.909090909090908,
"grad_norm": 1.2758557796478271,
"learning_rate": 9.814958493905963e-05,
"loss": 0.0928,
"step": 98
},
{
"epoch": 9.0,
"eval_loss": 0.6302943825721741,
"eval_runtime": 7.4945,
"eval_samples_per_second": 3.202,
"eval_steps_per_second": 3.202,
"step": 99
},
{
"epoch": 9.090909090909092,
"grad_norm": 0.6050943732261658,
"learning_rate": 9.797464868072488e-05,
"loss": 0.0531,
"step": 100
},
{
"epoch": 9.272727272727273,
"grad_norm": 1.1192340850830078,
"learning_rate": 9.779198285281325e-05,
"loss": 0.0141,
"step": 102
},
{
"epoch": 9.454545454545455,
"grad_norm": 1.3488664627075195,
"learning_rate": 9.760161688604008e-05,
"loss": 0.0797,
"step": 104
},
{
"epoch": 9.636363636363637,
"grad_norm": 0.8850768804550171,
"learning_rate": 9.740358145174998e-05,
"loss": 0.0414,
"step": 106
},
{
"epoch": 9.818181818181818,
"grad_norm": 1.1459277868270874,
"learning_rate": 9.719790845697533e-05,
"loss": 0.0287,
"step": 108
},
{
"epoch": 10.0,
"grad_norm": 0.6032407283782959,
"learning_rate": 9.698463103929542e-05,
"loss": 0.0316,
"step": 110
},
{
"epoch": 10.0,
"eval_loss": 0.7081627249717712,
"eval_runtime": 7.4983,
"eval_samples_per_second": 3.201,
"eval_steps_per_second": 3.201,
"step": 110
},
{
"epoch": 10.181818181818182,
"grad_norm": 0.7567160725593567,
"learning_rate": 9.676378356149734e-05,
"loss": 0.0287,
"step": 112
},
{
"epoch": 10.363636363636363,
"grad_norm": 0.6921805739402771,
"learning_rate": 9.653540160603956e-05,
"loss": 0.0389,
"step": 114
},
{
"epoch": 10.545454545454545,
"grad_norm": 0.9014925360679626,
"learning_rate": 9.629952196931901e-05,
"loss": 0.0288,
"step": 116
},
{
"epoch": 10.727272727272727,
"grad_norm": 1.5081769227981567,
"learning_rate": 9.60561826557425e-05,
"loss": 0.055,
"step": 118
},
{
"epoch": 10.909090909090908,
"grad_norm": 2.0981285572052,
"learning_rate": 9.580542287160348e-05,
"loss": 0.0231,
"step": 120
},
{
"epoch": 11.0,
"eval_loss": 0.6809744834899902,
"eval_runtime": 7.4929,
"eval_samples_per_second": 3.203,
"eval_steps_per_second": 3.203,
"step": 121
},
{
"epoch": 11.090909090909092,
"grad_norm": 1.2533338069915771,
"learning_rate": 9.554728301876526e-05,
"loss": 0.0251,
"step": 122
},
{
"epoch": 11.272727272727273,
"grad_norm": 0.9695553779602051,
"learning_rate": 9.528180468815155e-05,
"loss": 0.0261,
"step": 124
},
{
"epoch": 11.454545454545455,
"grad_norm": 0.31058475375175476,
"learning_rate": 9.50090306530454e-05,
"loss": 0.0225,
"step": 126
},
{
"epoch": 11.636363636363637,
"grad_norm": 0.6186215877532959,
"learning_rate": 9.472900486219769e-05,
"loss": 0.0216,
"step": 128
},
{
"epoch": 11.818181818181818,
"grad_norm": 0.3455003798007965,
"learning_rate": 9.444177243274618e-05,
"loss": 0.0124,
"step": 130
},
{
"epoch": 12.0,
"grad_norm": 0.8163161873817444,
"learning_rate": 9.414737964294636e-05,
"loss": 0.0257,
"step": 132
},
{
"epoch": 12.0,
"eval_loss": 0.8246378898620605,
"eval_runtime": 7.4958,
"eval_samples_per_second": 3.202,
"eval_steps_per_second": 3.202,
"step": 132
},
{
"epoch": 12.0,
"step": 132,
"total_flos": 2.8581641706274816e+16,
"train_loss": 0.5765298042643928,
"train_runtime": 1206.9458,
"train_samples_per_second": 3.646,
"train_steps_per_second": 0.456
}
],
"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": 2.8581641706274816e+16,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}