| { | |
| "best_metric": 0.8845333333333333, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0121/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": 158.83192443847656, | |
| "learning_rate": 4.849231551964771e-05, | |
| "loss": 3.6037, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.5909333333333333, | |
| "eval_loss": 2.9021124839782715, | |
| "eval_runtime": 14.2295, | |
| "eval_samples_per_second": 263.537, | |
| "eval_steps_per_second": 4.146, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 274.6767578125, | |
| "learning_rate": 4.415111107797445e-05, | |
| "loss": 1.6072, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7984, | |
| "eval_loss": 0.7853330373764038, | |
| "eval_runtime": 14.2323, | |
| "eval_samples_per_second": 263.485, | |
| "eval_steps_per_second": 4.145, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 180.38287353515625, | |
| "learning_rate": 3.7500000000000003e-05, | |
| "loss": 0.6428, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8445333333333334, | |
| "eval_loss": 0.5243738889694214, | |
| "eval_runtime": 14.1906, | |
| "eval_samples_per_second": 264.259, | |
| "eval_steps_per_second": 4.158, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 183.49349975585938, | |
| "learning_rate": 2.9341204441673266e-05, | |
| "loss": 0.4299, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8613333333333333, | |
| "eval_loss": 0.45170897245407104, | |
| "eval_runtime": 14.266, | |
| "eval_samples_per_second": 262.864, | |
| "eval_steps_per_second": 4.136, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 113.89549255371094, | |
| "learning_rate": 2.0658795558326743e-05, | |
| "loss": 0.3341, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8698666666666667, | |
| "eval_loss": 0.4065164625644684, | |
| "eval_runtime": 14.1787, | |
| "eval_samples_per_second": 264.482, | |
| "eval_steps_per_second": 4.161, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 116.1170654296875, | |
| "learning_rate": 1.2500000000000006e-05, | |
| "loss": 0.2803, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8786666666666667, | |
| "eval_loss": 0.3829270303249359, | |
| "eval_runtime": 14.1309, | |
| "eval_samples_per_second": 265.377, | |
| "eval_steps_per_second": 4.175, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 125.80799865722656, | |
| "learning_rate": 5.848888922025553e-06, | |
| "loss": 0.2438, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8813333333333333, | |
| "eval_loss": 0.38604316115379333, | |
| "eval_runtime": 14.0632, | |
| "eval_samples_per_second": 266.653, | |
| "eval_steps_per_second": 4.195, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 225.0499267578125, | |
| "learning_rate": 1.5076844803522922e-06, | |
| "loss": 0.2317, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8813333333333333, | |
| "eval_loss": 0.37400609254837036, | |
| "eval_runtime": 14.1252, | |
| "eval_samples_per_second": 265.483, | |
| "eval_steps_per_second": 4.177, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 114.55079650878906, | |
| "learning_rate": 0.0, | |
| "loss": 0.2218, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8845333333333333, | |
| "eval_loss": 0.37966179847717285, | |
| "eval_runtime": 14.0921, | |
| "eval_samples_per_second": 266.107, | |
| "eval_steps_per_second": 4.187, | |
| "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 | |
| } | |