| { | |
| "best_metric": 2.598254919052124, | |
| "best_model_checkpoint": "cat_breed_image_detection/checkpoint-7622", | |
| "epoch": 2.0, | |
| "eval_steps": 500, | |
| "global_step": 7622, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.8217115689381934e-07, | |
| "loss": 2.6163, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6236133122028525e-07, | |
| "loss": 2.6213, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.425515055467512e-07, | |
| "loss": 2.6131, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.2274167987321712e-07, | |
| "loss": 2.6089, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.0293185419968305e-07, | |
| "loss": 2.6016, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8312202852614895e-07, | |
| "loss": 2.5924, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.6331220285261489e-07, | |
| "loss": 2.5995, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.4030853262474166, | |
| "eval_loss": 2.609971761703491, | |
| "eval_runtime": 309.6807, | |
| "eval_samples_per_second": 87.497, | |
| "eval_steps_per_second": 2.735, | |
| "step": 3811 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.4350237717908082e-07, | |
| "loss": 2.5833, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.2369255150554675e-07, | |
| "loss": 2.592, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.0388272583201267e-07, | |
| "loss": 2.5807, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 8.40729001584786e-08, | |
| "loss": 2.5804, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 6.426307448494454e-08, | |
| "loss": 2.5852, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 4.445324881141046e-08, | |
| "loss": 2.5733, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 2.4643423137876387e-08, | |
| "loss": 2.5765, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 4.833597464342313e-09, | |
| "loss": 2.5837, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.40445084145261295, | |
| "eval_loss": 2.598254919052124, | |
| "eval_runtime": 319.6545, | |
| "eval_samples_per_second": 84.767, | |
| "eval_steps_per_second": 2.65, | |
| "step": 7622 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 7622, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 2, | |
| "save_steps": 500, | |
| "total_flos": 3.781059536914691e+19, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |