| { | |
| "best_metric": 0.9426666666666667, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0436/checkpoints/checkpoint-2664", | |
| "epoch": 8.0, | |
| "eval_steps": 500, | |
| "global_step": 2664, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 0.0374147854745388, | |
| "learning_rate": 4.446112779446113e-05, | |
| "loss": 0.7152, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8954666666666666, | |
| "eval_loss": 0.36535754799842834, | |
| "eval_runtime": 7.2277, | |
| "eval_samples_per_second": 518.836, | |
| "eval_steps_per_second": 8.163, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.15005812048912048, | |
| "learning_rate": 3.8905572238905574e-05, | |
| "loss": 0.1692, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8941333333333333, | |
| "eval_loss": 0.3577668070793152, | |
| "eval_runtime": 7.103, | |
| "eval_samples_per_second": 527.948, | |
| "eval_steps_per_second": 8.306, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 2.49857497215271, | |
| "learning_rate": 3.335001668335002e-05, | |
| "loss": 0.0707, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9, | |
| "eval_loss": 0.3871614634990692, | |
| "eval_runtime": 7.1294, | |
| "eval_samples_per_second": 525.993, | |
| "eval_steps_per_second": 8.276, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.00977238081395626, | |
| "learning_rate": 2.7794461127794462e-05, | |
| "loss": 0.0365, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9205333333333333, | |
| "eval_loss": 0.29927074909210205, | |
| "eval_runtime": 7.0217, | |
| "eval_samples_per_second": 534.059, | |
| "eval_steps_per_second": 8.403, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.05136696621775627, | |
| "learning_rate": 2.2238905572238907e-05, | |
| "loss": 0.0187, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9173333333333333, | |
| "eval_loss": 0.3406911790370941, | |
| "eval_runtime": 6.9528, | |
| "eval_samples_per_second": 539.352, | |
| "eval_steps_per_second": 8.486, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.002223671181127429, | |
| "learning_rate": 1.668335001668335e-05, | |
| "loss": 0.0067, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9325333333333333, | |
| "eval_loss": 0.291949599981308, | |
| "eval_runtime": 7.0114, | |
| "eval_samples_per_second": 534.843, | |
| "eval_steps_per_second": 8.415, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.042941153049468994, | |
| "learning_rate": 1.1127794461127795e-05, | |
| "loss": 0.0032, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9389333333333333, | |
| "eval_loss": 0.27105972170829773, | |
| "eval_runtime": 8.8133, | |
| "eval_samples_per_second": 425.493, | |
| "eval_steps_per_second": 6.694, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.0003100693866144866, | |
| "learning_rate": 5.572238905572239e-06, | |
| "loss": 0.0012, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9426666666666667, | |
| "eval_loss": 0.26802629232406616, | |
| "eval_runtime": 6.9182, | |
| "eval_samples_per_second": 542.047, | |
| "eval_steps_per_second": 8.528, | |
| "step": 2664 | |
| } | |
| ], | |
| "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": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 1.317930564980736e+19, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |