| { | |
| "best_metric": 0.884, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0679/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": 103.67074584960938, | |
| "learning_rate": 4.849231551964771e-05, | |
| "loss": 3.717, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.4848, | |
| "eval_loss": 3.133535623550415, | |
| "eval_runtime": 13.3227, | |
| "eval_samples_per_second": 281.475, | |
| "eval_steps_per_second": 4.429, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 165.566162109375, | |
| "learning_rate": 4.415111107797445e-05, | |
| "loss": 1.7478, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7768, | |
| "eval_loss": 0.8910831212997437, | |
| "eval_runtime": 13.2068, | |
| "eval_samples_per_second": 283.946, | |
| "eval_steps_per_second": 4.467, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 130.57664489746094, | |
| "learning_rate": 3.7500000000000003e-05, | |
| "loss": 0.7823, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8186666666666667, | |
| "eval_loss": 0.6578493714332581, | |
| "eval_runtime": 13.2206, | |
| "eval_samples_per_second": 283.648, | |
| "eval_steps_per_second": 4.463, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 258.30218505859375, | |
| "learning_rate": 2.9341204441673266e-05, | |
| "loss": 0.5806, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8557333333333333, | |
| "eval_loss": 0.4865432381629944, | |
| "eval_runtime": 13.2754, | |
| "eval_samples_per_second": 282.476, | |
| "eval_steps_per_second": 4.444, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 164.8311767578125, | |
| "learning_rate": 2.0658795558326743e-05, | |
| "loss": 0.4944, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8666666666666667, | |
| "eval_loss": 0.44439566135406494, | |
| "eval_runtime": 14.732, | |
| "eval_samples_per_second": 254.549, | |
| "eval_steps_per_second": 4.005, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 114.75255584716797, | |
| "learning_rate": 1.2500000000000006e-05, | |
| "loss": 0.4363, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8725333333333334, | |
| "eval_loss": 0.4250194728374481, | |
| "eval_runtime": 13.2186, | |
| "eval_samples_per_second": 283.692, | |
| "eval_steps_per_second": 4.463, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 209.06814575195312, | |
| "learning_rate": 5.848888922025553e-06, | |
| "loss": 0.4088, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8712, | |
| "eval_loss": 0.42266231775283813, | |
| "eval_runtime": 13.2991, | |
| "eval_samples_per_second": 281.975, | |
| "eval_steps_per_second": 4.436, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 134.50389099121094, | |
| "learning_rate": 1.5076844803522922e-06, | |
| "loss": 0.3836, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8789333333333333, | |
| "eval_loss": 0.40383920073509216, | |
| "eval_runtime": 13.3302, | |
| "eval_samples_per_second": 281.316, | |
| "eval_steps_per_second": 4.426, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 144.50161743164062, | |
| "learning_rate": 0.0, | |
| "loss": 0.389, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.884, | |
| "eval_loss": 0.40426984429359436, | |
| "eval_runtime": 13.3157, | |
| "eval_samples_per_second": 281.622, | |
| "eval_steps_per_second": 4.431, | |
| "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 | |
| } | |