|
{ |
|
"best_metric": 3.0114543437957764, |
|
"best_model_checkpoint": "car_brands_image_detection/checkpoint-22335", |
|
"epoch": 15.0, |
|
"eval_steps": 500, |
|
"global_step": 22335, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 9.798070450975994e-06, |
|
"loss": 5.6594, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 9.573704285393764e-06, |
|
"loss": 5.4474, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.053250461951957, |
|
"eval_loss": 5.233431816101074, |
|
"eval_runtime": 123.0141, |
|
"eval_samples_per_second": 96.786, |
|
"eval_steps_per_second": 12.104, |
|
"step": 1489 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 9.349338119811532e-06, |
|
"loss": 5.2977, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 1.34, |
|
"learning_rate": 9.124971954229302e-06, |
|
"loss": 5.1487, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 1.68, |
|
"learning_rate": 8.900605788647072e-06, |
|
"loss": 5.0336, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.11061649588442803, |
|
"eval_loss": 4.870540142059326, |
|
"eval_runtime": 113.0841, |
|
"eval_samples_per_second": 105.285, |
|
"eval_steps_per_second": 13.167, |
|
"step": 2978 |
|
}, |
|
{ |
|
"epoch": 2.01, |
|
"learning_rate": 8.676239623064842e-06, |
|
"loss": 4.9069, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 2.35, |
|
"learning_rate": 8.451873457482612e-06, |
|
"loss": 4.7895, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 2.69, |
|
"learning_rate": 8.227507291900382e-06, |
|
"loss": 4.6794, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.15723164790861752, |
|
"eval_loss": 4.573022365570068, |
|
"eval_runtime": 112.9135, |
|
"eval_samples_per_second": 105.444, |
|
"eval_steps_per_second": 13.187, |
|
"step": 4467 |
|
}, |
|
{ |
|
"epoch": 3.02, |
|
"learning_rate": 8.003141126318152e-06, |
|
"loss": 4.5832, |
|
"step": 4500 |
|
}, |
|
{ |
|
"epoch": 3.36, |
|
"learning_rate": 7.778774960735922e-06, |
|
"loss": 4.4765, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 3.69, |
|
"learning_rate": 7.5544087951536915e-06, |
|
"loss": 4.3855, |
|
"step": 5500 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.19032420628254662, |
|
"eval_loss": 4.311987400054932, |
|
"eval_runtime": 114.7324, |
|
"eval_samples_per_second": 103.772, |
|
"eval_steps_per_second": 12.978, |
|
"step": 5956 |
|
}, |
|
{ |
|
"epoch": 4.03, |
|
"learning_rate": 7.3300426295714614e-06, |
|
"loss": 4.3036, |
|
"step": 6000 |
|
}, |
|
{ |
|
"epoch": 4.37, |
|
"learning_rate": 7.1056764639892306e-06, |
|
"loss": 4.1922, |
|
"step": 6500 |
|
}, |
|
{ |
|
"epoch": 4.7, |
|
"learning_rate": 6.881310298407001e-06, |
|
"loss": 4.1169, |
|
"step": 7000 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.22736435410717284, |
|
"eval_loss": 4.080084800720215, |
|
"eval_runtime": 111.3941, |
|
"eval_samples_per_second": 106.882, |
|
"eval_steps_per_second": 13.367, |
|
"step": 7445 |
|
}, |
|
{ |
|
"epoch": 5.04, |
|
"learning_rate": 6.6569441328247705e-06, |
|
"loss": 4.0398, |
|
"step": 7500 |
|
}, |
|
{ |
|
"epoch": 5.37, |
|
"learning_rate": 6.4325779672425405e-06, |
|
"loss": 3.9461, |
|
"step": 8000 |
|
}, |
|
{ |
|
"epoch": 5.71, |
|
"learning_rate": 6.20821180166031e-06, |
|
"loss": 3.8711, |
|
"step": 8500 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"eval_accuracy": 0.2560893667058626, |
|
"eval_loss": 3.882530450820923, |
|
"eval_runtime": 111.8089, |
|
"eval_samples_per_second": 106.485, |
|
"eval_steps_per_second": 13.317, |
|
"step": 8934 |
|
}, |
|
{ |
|
"epoch": 6.04, |
|
"learning_rate": 5.98384563607808e-06, |
|
"loss": 3.7993, |
|
"step": 9000 |
|
}, |
|
{ |
|
"epoch": 6.38, |
|
"learning_rate": 5.7594794704958505e-06, |
|
"loss": 3.721, |
|
"step": 9500 |
|
}, |
|
{ |
|
"epoch": 6.72, |
|
"learning_rate": 5.53511330491362e-06, |
|
"loss": 3.6598, |
|
"step": 10000 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"eval_accuracy": 0.28674617839744665, |
|
"eval_loss": 3.685960292816162, |
|
"eval_runtime": 111.0451, |
|
"eval_samples_per_second": 107.218, |
|
"eval_steps_per_second": 13.409, |
|
"step": 10423 |
|
}, |
|
{ |
|
"epoch": 7.05, |
|
"learning_rate": 5.31074713933139e-06, |
|
"loss": 3.5779, |
|
"step": 10500 |
|
}, |
|
{ |
|
"epoch": 7.39, |
|
"learning_rate": 5.086380973749159e-06, |
|
"loss": 3.5094, |
|
"step": 11000 |
|
}, |
|
{ |
|
"epoch": 7.72, |
|
"learning_rate": 4.862014808166929e-06, |
|
"loss": 3.4613, |
|
"step": 11500 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"eval_accuracy": 0.31169158407525616, |
|
"eval_loss": 3.5292482376098633, |
|
"eval_runtime": 112.2315, |
|
"eval_samples_per_second": 106.084, |
|
"eval_steps_per_second": 13.267, |
|
"step": 11912 |
|
}, |
|
{ |
|
"epoch": 8.06, |
|
"learning_rate": 4.637648642584699e-06, |
|
"loss": 3.4116, |
|
"step": 12000 |
|
}, |
|
{ |
|
"epoch": 8.39, |
|
"learning_rate": 4.413282477002469e-06, |
|
"loss": 3.3376, |
|
"step": 12500 |
|
}, |
|
{ |
|
"epoch": 8.73, |
|
"learning_rate": 4.188916311420238e-06, |
|
"loss": 3.2839, |
|
"step": 13000 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"eval_accuracy": 0.3312615487989249, |
|
"eval_loss": 3.40043044090271, |
|
"eval_runtime": 122.6619, |
|
"eval_samples_per_second": 97.064, |
|
"eval_steps_per_second": 12.139, |
|
"step": 13401 |
|
}, |
|
{ |
|
"epoch": 9.07, |
|
"learning_rate": 3.964550145838008e-06, |
|
"loss": 3.2336, |
|
"step": 13500 |
|
}, |
|
{ |
|
"epoch": 9.4, |
|
"learning_rate": 3.740183980255778e-06, |
|
"loss": 3.1897, |
|
"step": 14000 |
|
}, |
|
{ |
|
"epoch": 9.74, |
|
"learning_rate": 3.515817814673548e-06, |
|
"loss": 3.1272, |
|
"step": 14500 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"eval_accuracy": 0.3625062993448681, |
|
"eval_loss": 3.284891128540039, |
|
"eval_runtime": 111.9781, |
|
"eval_samples_per_second": 106.324, |
|
"eval_steps_per_second": 13.297, |
|
"step": 14890 |
|
}, |
|
{ |
|
"epoch": 10.07, |
|
"learning_rate": 3.2914516490913174e-06, |
|
"loss": 3.0939, |
|
"step": 15000 |
|
}, |
|
{ |
|
"epoch": 10.41, |
|
"learning_rate": 3.067085483509087e-06, |
|
"loss": 3.06, |
|
"step": 15500 |
|
}, |
|
{ |
|
"epoch": 10.75, |
|
"learning_rate": 2.842719317926857e-06, |
|
"loss": 3.0064, |
|
"step": 16000 |
|
}, |
|
{ |
|
"epoch": 11.0, |
|
"eval_accuracy": 0.3743490676969595, |
|
"eval_loss": 3.1829681396484375, |
|
"eval_runtime": 111.6647, |
|
"eval_samples_per_second": 106.623, |
|
"eval_steps_per_second": 13.335, |
|
"step": 16379 |
|
}, |
|
{ |
|
"epoch": 11.08, |
|
"learning_rate": 2.6183531523446265e-06, |
|
"loss": 2.9652, |
|
"step": 16500 |
|
}, |
|
{ |
|
"epoch": 11.42, |
|
"learning_rate": 2.3939869867623964e-06, |
|
"loss": 2.9325, |
|
"step": 17000 |
|
}, |
|
{ |
|
"epoch": 11.75, |
|
"learning_rate": 2.169620821180166e-06, |
|
"loss": 2.9154, |
|
"step": 17500 |
|
}, |
|
{ |
|
"epoch": 12.0, |
|
"eval_accuracy": 0.3936670586259029, |
|
"eval_loss": 3.108581304550171, |
|
"eval_runtime": 112.5771, |
|
"eval_samples_per_second": 105.759, |
|
"eval_steps_per_second": 13.226, |
|
"step": 17868 |
|
}, |
|
{ |
|
"epoch": 12.09, |
|
"learning_rate": 1.945254655597936e-06, |
|
"loss": 2.8818, |
|
"step": 18000 |
|
}, |
|
{ |
|
"epoch": 12.42, |
|
"learning_rate": 1.7208884900157058e-06, |
|
"loss": 2.8296, |
|
"step": 18500 |
|
}, |
|
{ |
|
"epoch": 12.76, |
|
"learning_rate": 1.4965223244334755e-06, |
|
"loss": 2.8489, |
|
"step": 19000 |
|
}, |
|
{ |
|
"epoch": 13.0, |
|
"eval_accuracy": 0.4065177221568957, |
|
"eval_loss": 3.0557422637939453, |
|
"eval_runtime": 111.394, |
|
"eval_samples_per_second": 106.882, |
|
"eval_steps_per_second": 13.367, |
|
"step": 19357 |
|
}, |
|
{ |
|
"epoch": 13.1, |
|
"learning_rate": 1.2721561588512455e-06, |
|
"loss": 2.816, |
|
"step": 19500 |
|
}, |
|
{ |
|
"epoch": 13.43, |
|
"learning_rate": 1.047789993269015e-06, |
|
"loss": 2.7841, |
|
"step": 20000 |
|
}, |
|
{ |
|
"epoch": 13.77, |
|
"learning_rate": 8.23423827686785e-07, |
|
"loss": 2.7947, |
|
"step": 20500 |
|
}, |
|
{ |
|
"epoch": 14.0, |
|
"eval_accuracy": 0.41323702334957163, |
|
"eval_loss": 3.022327184677124, |
|
"eval_runtime": 111.3923, |
|
"eval_samples_per_second": 106.884, |
|
"eval_steps_per_second": 13.367, |
|
"step": 20846 |
|
}, |
|
{ |
|
"epoch": 14.1, |
|
"learning_rate": 5.990576621045547e-07, |
|
"loss": 2.7577, |
|
"step": 21000 |
|
}, |
|
{ |
|
"epoch": 14.44, |
|
"learning_rate": 3.7469149652232446e-07, |
|
"loss": 2.7593, |
|
"step": 21500 |
|
}, |
|
{ |
|
"epoch": 14.78, |
|
"learning_rate": 1.5032533094009426e-07, |
|
"loss": 2.7426, |
|
"step": 22000 |
|
}, |
|
{ |
|
"epoch": 15.0, |
|
"eval_accuracy": 0.4139089534688392, |
|
"eval_loss": 3.0114543437957764, |
|
"eval_runtime": 110.643, |
|
"eval_samples_per_second": 107.607, |
|
"eval_steps_per_second": 13.458, |
|
"step": 22335 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 22335, |
|
"num_train_epochs": 15, |
|
"save_steps": 500, |
|
"total_flos": 5.551527576563553e+19, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|