| { | |
| "best_metric": 0.40453333333333336, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0779/checkpoints/checkpoint-2997", | |
| "epoch": 9.0, | |
| "eval_steps": 500, | |
| "global_step": 2997, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 17.57228660583496, | |
| "learning_rate": 0.00048519088401615274, | |
| "loss": 3.4703, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.18666666666666668, | |
| "eval_loss": 3.0453500747680664, | |
| "eval_runtime": 11.1833, | |
| "eval_samples_per_second": 335.323, | |
| "eval_steps_per_second": 5.276, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 17.13241195678711, | |
| "learning_rate": 0.0004420155125428249, | |
| "loss": 2.9464, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.22453333333333333, | |
| "eval_loss": 2.8640048503875732, | |
| "eval_runtime": 11.0446, | |
| "eval_samples_per_second": 339.532, | |
| "eval_steps_per_second": 5.342, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 16.1562557220459, | |
| "learning_rate": 0.00037568023640968044, | |
| "loss": 2.7554, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.24266666666666667, | |
| "eval_loss": 2.771890640258789, | |
| "eval_runtime": 11.0091, | |
| "eval_samples_per_second": 340.627, | |
| "eval_steps_per_second": 5.359, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 12.122749328613281, | |
| "learning_rate": 0.00029418606892278544, | |
| "loss": 2.5659, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.28026666666666666, | |
| "eval_loss": 2.6315364837646484, | |
| "eval_runtime": 10.8308, | |
| "eval_samples_per_second": 346.234, | |
| "eval_steps_per_second": 5.447, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 22.77225685119629, | |
| "learning_rate": 0.0002073624094068776, | |
| "loss": 2.4043, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.32293333333333335, | |
| "eval_loss": 2.4572322368621826, | |
| "eval_runtime": 11.0614, | |
| "eval_samples_per_second": 339.016, | |
| "eval_steps_per_second": 5.334, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 16.535472869873047, | |
| "learning_rate": 0.00012568147258031897, | |
| "loss": 2.2037, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.3528, | |
| "eval_loss": 2.3550593852996826, | |
| "eval_runtime": 10.929, | |
| "eval_samples_per_second": 343.123, | |
| "eval_steps_per_second": 5.398, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 15.841681480407715, | |
| "learning_rate": 5.899518490663266e-05, | |
| "loss": 2.0058, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.3874666666666667, | |
| "eval_loss": 2.2249364852905273, | |
| "eval_runtime": 10.9031, | |
| "eval_samples_per_second": 343.938, | |
| "eval_steps_per_second": 5.411, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 15.18927001953125, | |
| "learning_rate": 1.534689686405272e-05, | |
| "loss": 1.8444, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.39866666666666667, | |
| "eval_loss": 2.2154488563537598, | |
| "eval_runtime": 10.7521, | |
| "eval_samples_per_second": 348.768, | |
| "eval_steps_per_second": 5.487, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 6.721107006072998, | |
| "learning_rate": 1.2361706385188854e-09, | |
| "loss": 1.7252, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.40453333333333336, | |
| "eval_loss": 2.229093313217163, | |
| "eval_runtime": 11.2377, | |
| "eval_samples_per_second": 333.697, | |
| "eval_steps_per_second": 5.25, | |
| "step": 2997 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2997, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 9, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 1.482671885603328e+19, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |