|
{ |
|
"best_metric": 0.5748319625854492, |
|
"best_model_checkpoint": "./croupier-creature-classifier/checkpoint-400", |
|
"epoch": 5.0, |
|
"global_step": 455, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.00019560439560439562, |
|
"loss": 1.3464, |
|
"step": 10 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.00019120879120879122, |
|
"loss": 1.2608, |
|
"step": 20 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.00018681318681318683, |
|
"loss": 1.1356, |
|
"step": 30 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0001824175824175824, |
|
"loss": 1.1795, |
|
"step": 40 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.00017802197802197802, |
|
"loss": 0.9829, |
|
"step": 50 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.00017362637362637365, |
|
"loss": 1.049, |
|
"step": 60 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.00016923076923076923, |
|
"loss": 1.0607, |
|
"step": 70 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.00016483516483516484, |
|
"loss": 0.9264, |
|
"step": 80 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.00016043956043956044, |
|
"loss": 0.8931, |
|
"step": 90 |
|
}, |
|
{ |
|
"epoch": 1.1, |
|
"learning_rate": 0.00015604395604395605, |
|
"loss": 0.6663, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 1.1, |
|
"eval_accuracy": 0.5941176470588235, |
|
"eval_loss": 1.0179336071014404, |
|
"eval_runtime": 1.6282, |
|
"eval_samples_per_second": 104.41, |
|
"eval_steps_per_second": 13.512, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 1.21, |
|
"learning_rate": 0.00015164835164835165, |
|
"loss": 0.7923, |
|
"step": 110 |
|
}, |
|
{ |
|
"epoch": 1.32, |
|
"learning_rate": 0.00014725274725274726, |
|
"loss": 0.83, |
|
"step": 120 |
|
}, |
|
{ |
|
"epoch": 1.43, |
|
"learning_rate": 0.00014285714285714287, |
|
"loss": 0.6796, |
|
"step": 130 |
|
}, |
|
{ |
|
"epoch": 1.54, |
|
"learning_rate": 0.00013846153846153847, |
|
"loss": 0.7702, |
|
"step": 140 |
|
}, |
|
{ |
|
"epoch": 1.65, |
|
"learning_rate": 0.00013406593406593405, |
|
"loss": 0.8231, |
|
"step": 150 |
|
}, |
|
{ |
|
"epoch": 1.76, |
|
"learning_rate": 0.0001296703296703297, |
|
"loss": 0.7022, |
|
"step": 160 |
|
}, |
|
{ |
|
"epoch": 1.87, |
|
"learning_rate": 0.00012527472527472527, |
|
"loss": 0.6745, |
|
"step": 170 |
|
}, |
|
{ |
|
"epoch": 1.98, |
|
"learning_rate": 0.00012087912087912087, |
|
"loss": 0.7639, |
|
"step": 180 |
|
}, |
|
{ |
|
"epoch": 2.09, |
|
"learning_rate": 0.0001164835164835165, |
|
"loss": 0.5086, |
|
"step": 190 |
|
}, |
|
{ |
|
"epoch": 2.2, |
|
"learning_rate": 0.0001120879120879121, |
|
"loss": 0.4924, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 2.2, |
|
"eval_accuracy": 0.7529411764705882, |
|
"eval_loss": 0.7035554647445679, |
|
"eval_runtime": 2.1312, |
|
"eval_samples_per_second": 79.768, |
|
"eval_steps_per_second": 10.323, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 2.31, |
|
"learning_rate": 0.0001076923076923077, |
|
"loss": 0.4121, |
|
"step": 210 |
|
}, |
|
{ |
|
"epoch": 2.42, |
|
"learning_rate": 0.00010329670329670331, |
|
"loss": 0.451, |
|
"step": 220 |
|
}, |
|
{ |
|
"epoch": 2.53, |
|
"learning_rate": 9.89010989010989e-05, |
|
"loss": 0.5021, |
|
"step": 230 |
|
}, |
|
{ |
|
"epoch": 2.64, |
|
"learning_rate": 9.450549450549451e-05, |
|
"loss": 0.3728, |
|
"step": 240 |
|
}, |
|
{ |
|
"epoch": 2.75, |
|
"learning_rate": 9.010989010989012e-05, |
|
"loss": 0.5141, |
|
"step": 250 |
|
}, |
|
{ |
|
"epoch": 2.86, |
|
"learning_rate": 8.571428571428571e-05, |
|
"loss": 0.4065, |
|
"step": 260 |
|
}, |
|
{ |
|
"epoch": 2.97, |
|
"learning_rate": 8.131868131868132e-05, |
|
"loss": 0.434, |
|
"step": 270 |
|
}, |
|
{ |
|
"epoch": 3.08, |
|
"learning_rate": 7.692307692307693e-05, |
|
"loss": 0.3047, |
|
"step": 280 |
|
}, |
|
{ |
|
"epoch": 3.19, |
|
"learning_rate": 7.252747252747253e-05, |
|
"loss": 0.2544, |
|
"step": 290 |
|
}, |
|
{ |
|
"epoch": 3.3, |
|
"learning_rate": 6.813186813186814e-05, |
|
"loss": 0.4552, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 3.3, |
|
"eval_accuracy": 0.7823529411764706, |
|
"eval_loss": 0.612335741519928, |
|
"eval_runtime": 1.7201, |
|
"eval_samples_per_second": 98.833, |
|
"eval_steps_per_second": 12.79, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 3.41, |
|
"learning_rate": 6.373626373626373e-05, |
|
"loss": 0.2929, |
|
"step": 310 |
|
}, |
|
{ |
|
"epoch": 3.52, |
|
"learning_rate": 5.9340659340659345e-05, |
|
"loss": 0.297, |
|
"step": 320 |
|
}, |
|
{ |
|
"epoch": 3.63, |
|
"learning_rate": 5.494505494505495e-05, |
|
"loss": 0.2667, |
|
"step": 330 |
|
}, |
|
{ |
|
"epoch": 3.74, |
|
"learning_rate": 5.054945054945055e-05, |
|
"loss": 0.2297, |
|
"step": 340 |
|
}, |
|
{ |
|
"epoch": 3.85, |
|
"learning_rate": 4.615384615384616e-05, |
|
"loss": 0.2997, |
|
"step": 350 |
|
}, |
|
{ |
|
"epoch": 3.96, |
|
"learning_rate": 4.1758241758241765e-05, |
|
"loss": 0.4482, |
|
"step": 360 |
|
}, |
|
{ |
|
"epoch": 4.07, |
|
"learning_rate": 3.7362637362637365e-05, |
|
"loss": 0.2956, |
|
"step": 370 |
|
}, |
|
{ |
|
"epoch": 4.18, |
|
"learning_rate": 3.296703296703297e-05, |
|
"loss": 0.1919, |
|
"step": 380 |
|
}, |
|
{ |
|
"epoch": 4.29, |
|
"learning_rate": 2.857142857142857e-05, |
|
"loss": 0.2353, |
|
"step": 390 |
|
}, |
|
{ |
|
"epoch": 4.4, |
|
"learning_rate": 2.4175824175824177e-05, |
|
"loss": 0.2355, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 4.4, |
|
"eval_accuracy": 0.7647058823529411, |
|
"eval_loss": 0.5748319625854492, |
|
"eval_runtime": 1.7042, |
|
"eval_samples_per_second": 99.752, |
|
"eval_steps_per_second": 12.909, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 4.51, |
|
"learning_rate": 1.978021978021978e-05, |
|
"loss": 0.1942, |
|
"step": 410 |
|
}, |
|
{ |
|
"epoch": 4.62, |
|
"learning_rate": 1.5384615384615387e-05, |
|
"loss": 0.231, |
|
"step": 420 |
|
}, |
|
{ |
|
"epoch": 4.73, |
|
"learning_rate": 1.0989010989010989e-05, |
|
"loss": 0.2007, |
|
"step": 430 |
|
}, |
|
{ |
|
"epoch": 4.84, |
|
"learning_rate": 6.5934065934065935e-06, |
|
"loss": 0.2592, |
|
"step": 440 |
|
}, |
|
{ |
|
"epoch": 4.95, |
|
"learning_rate": 2.197802197802198e-06, |
|
"loss": 0.2764, |
|
"step": 450 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"step": 455, |
|
"total_flos": 5.622144623740109e+17, |
|
"train_loss": 0.5662190929873959, |
|
"train_runtime": 205.6317, |
|
"train_samples_per_second": 35.282, |
|
"train_steps_per_second": 2.213 |
|
} |
|
], |
|
"max_steps": 455, |
|
"num_train_epochs": 5, |
|
"total_flos": 5.622144623740109e+17, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|