| { | |
| "best_metric": 0.9037333333333334, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0971/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": 2.0280227661132812, | |
| "learning_rate": 9.698463103929542e-05, | |
| "loss": 1.492, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7757333333333334, | |
| "eval_loss": 0.7875099778175354, | |
| "eval_runtime": 8.2973, | |
| "eval_samples_per_second": 451.954, | |
| "eval_steps_per_second": 7.111, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 4.885120391845703, | |
| "learning_rate": 8.83022221559489e-05, | |
| "loss": 0.5376, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.84, | |
| "eval_loss": 0.5496315360069275, | |
| "eval_runtime": 7.4187, | |
| "eval_samples_per_second": 505.477, | |
| "eval_steps_per_second": 7.953, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 2.7939510345458984, | |
| "learning_rate": 7.504537663092833e-05, | |
| "loss": 0.3392, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8557333333333333, | |
| "eval_loss": 0.4865073561668396, | |
| "eval_runtime": 7.4076, | |
| "eval_samples_per_second": 506.238, | |
| "eval_steps_per_second": 7.965, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.10549061745405197, | |
| "learning_rate": 5.873402013309226e-05, | |
| "loss": 0.2066, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8752, | |
| "eval_loss": 0.43492552638053894, | |
| "eval_runtime": 8.4699, | |
| "eval_samples_per_second": 442.744, | |
| "eval_steps_per_second": 6.966, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 5.698217391967773, | |
| "learning_rate": 4.136921190679637e-05, | |
| "loss": 0.1227, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8866666666666667, | |
| "eval_loss": 0.4146025478839874, | |
| "eval_runtime": 7.3763, | |
| "eval_samples_per_second": 508.388, | |
| "eval_steps_per_second": 7.999, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.10362745076417923, | |
| "learning_rate": 2.5045404101407076e-05, | |
| "loss": 0.0607, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8890666666666667, | |
| "eval_loss": 0.4123982787132263, | |
| "eval_runtime": 7.4828, | |
| "eval_samples_per_second": 501.15, | |
| "eval_steps_per_second": 7.885, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.08338188380002975, | |
| "learning_rate": 1.1731488851999511e-05, | |
| "loss": 0.0295, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8992, | |
| "eval_loss": 0.383831650018692, | |
| "eval_runtime": 7.5391, | |
| "eval_samples_per_second": 497.404, | |
| "eval_steps_per_second": 7.826, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.05393170192837715, | |
| "learning_rate": 3.033320830114261e-06, | |
| "loss": 0.0141, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9016, | |
| "eval_loss": 0.3772144317626953, | |
| "eval_runtime": 7.5629, | |
| "eval_samples_per_second": 495.839, | |
| "eval_steps_per_second": 7.801, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.23172830045223236, | |
| "learning_rate": 2.7470478758395702e-11, | |
| "loss": 0.0099, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9037333333333334, | |
| "eval_loss": 0.38102009892463684, | |
| "eval_runtime": 7.8832, | |
| "eval_samples_per_second": 475.693, | |
| "eval_steps_per_second": 7.484, | |
| "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 | |
| } | |