|
{ |
|
"best_metric": 0.22263146936893463, |
|
"best_model_checkpoint": "ai_vs_real_image_detection/checkpoint-550", |
|
"epoch": 50.0, |
|
"eval_steps": 500, |
|
"global_step": 550, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.5714285714285714, |
|
"eval_loss": 0.7273932695388794, |
|
"eval_runtime": 0.2528, |
|
"eval_samples_per_second": 55.374, |
|
"eval_steps_per_second": 3.955, |
|
"step": 11 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.5714285714285714, |
|
"eval_loss": 0.6871033906936646, |
|
"eval_runtime": 0.2522, |
|
"eval_samples_per_second": 55.519, |
|
"eval_steps_per_second": 3.966, |
|
"step": 22 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.6428571428571429, |
|
"eval_loss": 0.628114640712738, |
|
"eval_runtime": 0.2547, |
|
"eval_samples_per_second": 54.961, |
|
"eval_steps_per_second": 3.926, |
|
"step": 33 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.6428571428571429, |
|
"eval_loss": 0.5625258088111877, |
|
"eval_runtime": 0.2502, |
|
"eval_samples_per_second": 55.949, |
|
"eval_steps_per_second": 3.996, |
|
"step": 44 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.6428571428571429, |
|
"eval_loss": 0.5028806328773499, |
|
"eval_runtime": 0.2628, |
|
"eval_samples_per_second": 53.267, |
|
"eval_steps_per_second": 3.805, |
|
"step": 55 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"eval_accuracy": 0.7142857142857143, |
|
"eval_loss": 0.460813969373703, |
|
"eval_runtime": 0.2498, |
|
"eval_samples_per_second": 56.045, |
|
"eval_steps_per_second": 4.003, |
|
"step": 66 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.4366808235645294, |
|
"eval_runtime": 0.2541, |
|
"eval_samples_per_second": 55.086, |
|
"eval_steps_per_second": 3.935, |
|
"step": 77 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"eval_accuracy": 0.7857142857142857, |
|
"eval_loss": 0.4217378795146942, |
|
"eval_runtime": 0.2616, |
|
"eval_samples_per_second": 53.517, |
|
"eval_steps_per_second": 3.823, |
|
"step": 88 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"eval_accuracy": 0.7857142857142857, |
|
"eval_loss": 0.41016918420791626, |
|
"eval_runtime": 0.2591, |
|
"eval_samples_per_second": 54.031, |
|
"eval_steps_per_second": 3.859, |
|
"step": 99 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"eval_accuracy": 0.7857142857142857, |
|
"eval_loss": 0.40068450570106506, |
|
"eval_runtime": 0.3007, |
|
"eval_samples_per_second": 46.553, |
|
"eval_steps_per_second": 3.325, |
|
"step": 110 |
|
}, |
|
{ |
|
"epoch": 11.0, |
|
"eval_accuracy": 0.7857142857142857, |
|
"eval_loss": 0.3925851285457611, |
|
"eval_runtime": 0.2562, |
|
"eval_samples_per_second": 54.651, |
|
"eval_steps_per_second": 3.904, |
|
"step": 121 |
|
}, |
|
{ |
|
"epoch": 12.0, |
|
"eval_accuracy": 0.7857142857142857, |
|
"eval_loss": 0.38272687792778015, |
|
"eval_runtime": 0.2492, |
|
"eval_samples_per_second": 56.182, |
|
"eval_steps_per_second": 4.013, |
|
"step": 132 |
|
}, |
|
{ |
|
"epoch": 13.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.3719024062156677, |
|
"eval_runtime": 0.2771, |
|
"eval_samples_per_second": 50.518, |
|
"eval_steps_per_second": 3.608, |
|
"step": 143 |
|
}, |
|
{ |
|
"epoch": 14.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.36182349920272827, |
|
"eval_runtime": 0.2615, |
|
"eval_samples_per_second": 53.532, |
|
"eval_steps_per_second": 3.824, |
|
"step": 154 |
|
}, |
|
{ |
|
"epoch": 15.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.3521384596824646, |
|
"eval_runtime": 0.2511, |
|
"eval_samples_per_second": 55.754, |
|
"eval_steps_per_second": 3.982, |
|
"step": 165 |
|
}, |
|
{ |
|
"epoch": 16.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.34298914670944214, |
|
"eval_runtime": 0.2598, |
|
"eval_samples_per_second": 53.888, |
|
"eval_steps_per_second": 3.849, |
|
"step": 176 |
|
}, |
|
{ |
|
"epoch": 17.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.33336836099624634, |
|
"eval_runtime": 0.2633, |
|
"eval_samples_per_second": 53.172, |
|
"eval_steps_per_second": 3.798, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 18.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.32352977991104126, |
|
"eval_runtime": 0.2595, |
|
"eval_samples_per_second": 53.958, |
|
"eval_steps_per_second": 3.854, |
|
"step": 198 |
|
}, |
|
{ |
|
"epoch": 19.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.31632325053215027, |
|
"eval_runtime": 0.2585, |
|
"eval_samples_per_second": 54.152, |
|
"eval_steps_per_second": 3.868, |
|
"step": 209 |
|
}, |
|
{ |
|
"epoch": 20.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.309923380613327, |
|
"eval_runtime": 0.2589, |
|
"eval_samples_per_second": 54.079, |
|
"eval_steps_per_second": 3.863, |
|
"step": 220 |
|
}, |
|
{ |
|
"epoch": 21.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.303012877702713, |
|
"eval_runtime": 0.2582, |
|
"eval_samples_per_second": 54.229, |
|
"eval_steps_per_second": 3.874, |
|
"step": 231 |
|
}, |
|
{ |
|
"epoch": 22.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.29638054966926575, |
|
"eval_runtime": 0.2582, |
|
"eval_samples_per_second": 54.214, |
|
"eval_steps_per_second": 3.872, |
|
"step": 242 |
|
}, |
|
{ |
|
"epoch": 23.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.2909241318702698, |
|
"eval_runtime": 0.2567, |
|
"eval_samples_per_second": 54.546, |
|
"eval_steps_per_second": 3.896, |
|
"step": 253 |
|
}, |
|
{ |
|
"epoch": 24.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.28674623370170593, |
|
"eval_runtime": 0.2689, |
|
"eval_samples_per_second": 52.072, |
|
"eval_steps_per_second": 3.719, |
|
"step": 264 |
|
}, |
|
{ |
|
"epoch": 25.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.2813883125782013, |
|
"eval_runtime": 0.2652, |
|
"eval_samples_per_second": 52.793, |
|
"eval_steps_per_second": 3.771, |
|
"step": 275 |
|
}, |
|
{ |
|
"epoch": 26.0, |
|
"eval_accuracy": 0.8571428571428571, |
|
"eval_loss": 0.27675989270210266, |
|
"eval_runtime": 0.2594, |
|
"eval_samples_per_second": 53.96, |
|
"eval_steps_per_second": 3.854, |
|
"step": 286 |
|
}, |
|
{ |
|
"epoch": 27.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.2722492516040802, |
|
"eval_runtime": 0.2583, |
|
"eval_samples_per_second": 54.192, |
|
"eval_steps_per_second": 3.871, |
|
"step": 297 |
|
}, |
|
{ |
|
"epoch": 28.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.26842695474624634, |
|
"eval_runtime": 0.255, |
|
"eval_samples_per_second": 54.892, |
|
"eval_steps_per_second": 3.921, |
|
"step": 308 |
|
}, |
|
{ |
|
"epoch": 29.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.264372318983078, |
|
"eval_runtime": 0.2548, |
|
"eval_samples_per_second": 54.953, |
|
"eval_steps_per_second": 3.925, |
|
"step": 319 |
|
}, |
|
{ |
|
"epoch": 30.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.2603602111339569, |
|
"eval_runtime": 0.2561, |
|
"eval_samples_per_second": 54.677, |
|
"eval_steps_per_second": 3.905, |
|
"step": 330 |
|
}, |
|
{ |
|
"epoch": 31.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.25613969564437866, |
|
"eval_runtime": 0.2548, |
|
"eval_samples_per_second": 54.937, |
|
"eval_steps_per_second": 3.924, |
|
"step": 341 |
|
}, |
|
{ |
|
"epoch": 32.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.25249889492988586, |
|
"eval_runtime": 0.2546, |
|
"eval_samples_per_second": 54.993, |
|
"eval_steps_per_second": 3.928, |
|
"step": 352 |
|
}, |
|
{ |
|
"epoch": 33.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.24893353879451752, |
|
"eval_runtime": 0.2556, |
|
"eval_samples_per_second": 54.783, |
|
"eval_steps_per_second": 3.913, |
|
"step": 363 |
|
}, |
|
{ |
|
"epoch": 34.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.24575325846672058, |
|
"eval_runtime": 0.2547, |
|
"eval_samples_per_second": 54.976, |
|
"eval_steps_per_second": 3.927, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 35.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.24278688430786133, |
|
"eval_runtime": 0.255, |
|
"eval_samples_per_second": 54.895, |
|
"eval_steps_per_second": 3.921, |
|
"step": 385 |
|
}, |
|
{ |
|
"epoch": 36.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.24021367728710175, |
|
"eval_runtime": 0.2517, |
|
"eval_samples_per_second": 55.616, |
|
"eval_steps_per_second": 3.973, |
|
"step": 396 |
|
}, |
|
{ |
|
"epoch": 37.0, |
|
"eval_accuracy": 0.9285714285714286, |
|
"eval_loss": 0.23749400675296783, |
|
"eval_runtime": 0.2567, |
|
"eval_samples_per_second": 54.541, |
|
"eval_steps_per_second": 3.896, |
|
"step": 407 |
|
}, |
|
{ |
|
"epoch": 38.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.23497438430786133, |
|
"eval_runtime": 0.2516, |
|
"eval_samples_per_second": 55.641, |
|
"eval_steps_per_second": 3.974, |
|
"step": 418 |
|
}, |
|
{ |
|
"epoch": 39.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.2329796999692917, |
|
"eval_runtime": 0.256, |
|
"eval_samples_per_second": 54.696, |
|
"eval_steps_per_second": 3.907, |
|
"step": 429 |
|
}, |
|
{ |
|
"epoch": 40.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.23115703463554382, |
|
"eval_runtime": 0.2554, |
|
"eval_samples_per_second": 54.819, |
|
"eval_steps_per_second": 3.916, |
|
"step": 440 |
|
}, |
|
{ |
|
"epoch": 41.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.2296319305896759, |
|
"eval_runtime": 0.2638, |
|
"eval_samples_per_second": 53.075, |
|
"eval_steps_per_second": 3.791, |
|
"step": 451 |
|
}, |
|
{ |
|
"epoch": 42.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22797557711601257, |
|
"eval_runtime": 0.2566, |
|
"eval_samples_per_second": 54.55, |
|
"eval_steps_per_second": 3.896, |
|
"step": 462 |
|
}, |
|
{ |
|
"epoch": 43.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22679831087589264, |
|
"eval_runtime": 0.2528, |
|
"eval_samples_per_second": 55.371, |
|
"eval_steps_per_second": 3.955, |
|
"step": 473 |
|
}, |
|
{ |
|
"epoch": 44.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22579309344291687, |
|
"eval_runtime": 0.2533, |
|
"eval_samples_per_second": 55.279, |
|
"eval_steps_per_second": 3.949, |
|
"step": 484 |
|
}, |
|
{ |
|
"epoch": 45.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22489233314990997, |
|
"eval_runtime": 0.252, |
|
"eval_samples_per_second": 55.566, |
|
"eval_steps_per_second": 3.969, |
|
"step": 495 |
|
}, |
|
{ |
|
"epoch": 45.45, |
|
"grad_norm": 1.2250463962554932, |
|
"learning_rate": 1e-07, |
|
"loss": 0.4654, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 46.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22412164509296417, |
|
"eval_runtime": 0.251, |
|
"eval_samples_per_second": 55.767, |
|
"eval_steps_per_second": 3.983, |
|
"step": 506 |
|
}, |
|
{ |
|
"epoch": 47.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22347252070903778, |
|
"eval_runtime": 0.2544, |
|
"eval_samples_per_second": 55.029, |
|
"eval_steps_per_second": 3.931, |
|
"step": 517 |
|
}, |
|
{ |
|
"epoch": 48.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22297818958759308, |
|
"eval_runtime": 0.2524, |
|
"eval_samples_per_second": 55.461, |
|
"eval_steps_per_second": 3.962, |
|
"step": 528 |
|
}, |
|
{ |
|
"epoch": 49.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22271637618541718, |
|
"eval_runtime": 0.2493, |
|
"eval_samples_per_second": 56.167, |
|
"eval_steps_per_second": 4.012, |
|
"step": 539 |
|
}, |
|
{ |
|
"epoch": 50.0, |
|
"eval_accuracy": 1.0, |
|
"eval_loss": 0.22263146936893463, |
|
"eval_runtime": 0.2541, |
|
"eval_samples_per_second": 55.097, |
|
"eval_steps_per_second": 3.935, |
|
"step": 550 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 550, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 50, |
|
"save_steps": 500, |
|
"total_flos": 5.114471314563072e+18, |
|
"train_batch_size": 128, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|