| { | |
| "best_metric": 0.8333333333333334, | |
| "best_model_checkpoint": "vit-base-patch16-224-U6-10\\checkpoint-160", | |
| "epoch": 10.0, | |
| "eval_steps": 500, | |
| "global_step": 160, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 5.4276315789473686e-05, | |
| "loss": 1.3653, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6333333333333333, | |
| "eval_loss": 1.2199313640594482, | |
| "eval_runtime": 1.1758, | |
| "eval_samples_per_second": 51.028, | |
| "eval_steps_per_second": 1.701, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 5.0657894736842104e-05, | |
| "loss": 1.2317, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 4.703947368421053e-05, | |
| "loss": 1.0932, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.65, | |
| "eval_loss": 1.008633017539978, | |
| "eval_runtime": 0.9786, | |
| "eval_samples_per_second": 61.313, | |
| "eval_steps_per_second": 2.044, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 4.342105263157895e-05, | |
| "loss": 0.9284, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.6666666666666666, | |
| "eval_loss": 0.846645712852478, | |
| "eval_runtime": 0.9829, | |
| "eval_samples_per_second": 61.047, | |
| "eval_steps_per_second": 2.035, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 3.9802631578947374e-05, | |
| "loss": 0.7896, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 3.618421052631579e-05, | |
| "loss": 0.6745, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.7, | |
| "eval_loss": 0.8237163424491882, | |
| "eval_runtime": 0.9433, | |
| "eval_samples_per_second": 63.603, | |
| "eval_steps_per_second": 2.12, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 3.256578947368421e-05, | |
| "loss": 0.5741, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 2.894736842105263e-05, | |
| "loss": 0.4775, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.7666666666666667, | |
| "eval_loss": 0.7472967505455017, | |
| "eval_runtime": 1.018, | |
| "eval_samples_per_second": 58.942, | |
| "eval_steps_per_second": 1.965, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 2.5328947368421052e-05, | |
| "loss": 0.4194, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.7833333333333333, | |
| "eval_loss": 0.6147837042808533, | |
| "eval_runtime": 1.0221, | |
| "eval_samples_per_second": 58.702, | |
| "eval_steps_per_second": 1.957, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 2.1710526315789474e-05, | |
| "loss": 0.3463, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 1.8092105263157896e-05, | |
| "loss": 0.3043, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8166666666666667, | |
| "eval_loss": 0.6220994591712952, | |
| "eval_runtime": 1.1621, | |
| "eval_samples_per_second": 51.63, | |
| "eval_steps_per_second": 1.721, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 1.4473684210526315e-05, | |
| "loss": 0.2947, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.7666666666666667, | |
| "eval_loss": 0.6156017184257507, | |
| "eval_runtime": 1.1636, | |
| "eval_samples_per_second": 51.562, | |
| "eval_steps_per_second": 1.719, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 1.0855263157894737e-05, | |
| "loss": 0.2506, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 7.2368421052631575e-06, | |
| "loss": 0.269, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8166666666666667, | |
| "eval_loss": 0.5700237154960632, | |
| "eval_runtime": 1.0077, | |
| "eval_samples_per_second": 59.539, | |
| "eval_steps_per_second": 1.985, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 9.38, | |
| "learning_rate": 3.6184210526315788e-06, | |
| "loss": 0.232, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.2261, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.8333333333333334, | |
| "eval_loss": 0.5472932457923889, | |
| "eval_runtime": 0.9808, | |
| "eval_samples_per_second": 61.172, | |
| "eval_steps_per_second": 2.039, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 160, | |
| "total_flos": 1.5731155873456128e+18, | |
| "train_loss": 0.5923053041100502, | |
| "train_runtime": 337.9075, | |
| "train_samples_per_second": 60.076, | |
| "train_steps_per_second": 0.474 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 160, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 10, | |
| "save_steps": 500, | |
| "total_flos": 1.5731155873456128e+18, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |