| { | |
| "best_metric": 0.8997333333333334, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0596/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": 47.80226516723633, | |
| "learning_rate": 0.0004849231551964771, | |
| "loss": 1.4293, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8154666666666667, | |
| "eval_loss": 0.6522212028503418, | |
| "eval_runtime": 8.1881, | |
| "eval_samples_per_second": 457.981, | |
| "eval_steps_per_second": 7.206, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 45.42972946166992, | |
| "learning_rate": 0.0004415111107797445, | |
| "loss": 0.5262, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8538666666666667, | |
| "eval_loss": 0.47904345393180847, | |
| "eval_runtime": 7.9809, | |
| "eval_samples_per_second": 469.87, | |
| "eval_steps_per_second": 7.393, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 16.38282012939453, | |
| "learning_rate": 0.000375, | |
| "loss": 0.3346, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8618666666666667, | |
| "eval_loss": 0.4891296327114105, | |
| "eval_runtime": 7.8845, | |
| "eval_samples_per_second": 475.619, | |
| "eval_steps_per_second": 7.483, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 21.163249969482422, | |
| "learning_rate": 0.00029341204441673266, | |
| "loss": 0.2196, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8658666666666667, | |
| "eval_loss": 0.46806982159614563, | |
| "eval_runtime": 7.9217, | |
| "eval_samples_per_second": 473.385, | |
| "eval_steps_per_second": 7.448, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 23.878019332885742, | |
| "learning_rate": 0.00020658795558326743, | |
| "loss": 0.1398, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8826666666666667, | |
| "eval_loss": 0.4209830164909363, | |
| "eval_runtime": 7.915, | |
| "eval_samples_per_second": 473.782, | |
| "eval_steps_per_second": 7.454, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 36.033164978027344, | |
| "learning_rate": 0.00012500000000000006, | |
| "loss": 0.0856, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8877333333333334, | |
| "eval_loss": 0.44882020354270935, | |
| "eval_runtime": 7.9742, | |
| "eval_samples_per_second": 470.266, | |
| "eval_steps_per_second": 7.399, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 36.23210525512695, | |
| "learning_rate": 5.848888922025553e-05, | |
| "loss": 0.059, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8914666666666666, | |
| "eval_loss": 0.4339054822921753, | |
| "eval_runtime": 7.8511, | |
| "eval_samples_per_second": 477.642, | |
| "eval_steps_per_second": 7.515, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 28.535015106201172, | |
| "learning_rate": 1.5076844803522921e-05, | |
| "loss": 0.0378, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8933333333333333, | |
| "eval_loss": 0.43278878927230835, | |
| "eval_runtime": 7.8672, | |
| "eval_samples_per_second": 476.664, | |
| "eval_steps_per_second": 7.5, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 32.43698501586914, | |
| "learning_rate": 0.0, | |
| "loss": 0.0312, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8997333333333334, | |
| "eval_loss": 0.4177744686603546, | |
| "eval_runtime": 7.7631, | |
| "eval_samples_per_second": 483.056, | |
| "eval_steps_per_second": 7.6, | |
| "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": 7.3587863155968e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |