| { | |
| "best_metric": 0.9773333333333334, | |
| "best_model_checkpoint": "vit-base-xray-pneumonia-lcbsi/checkpoint-216", | |
| "epoch": 9.981818181818182, | |
| "global_step": 270, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.7037037037037037e-06, | |
| "loss": 1.6187, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.4074074074074075e-06, | |
| "loss": 1.5887, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_accuracy": 0.44133333333333336, | |
| "eval_loss": 1.4948956966400146, | |
| "eval_runtime": 52.7762, | |
| "eval_samples_per_second": 14.211, | |
| "eval_steps_per_second": 0.455, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.876543209876543e-06, | |
| "loss": 1.5811, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 9.465020576131688e-06, | |
| "loss": 1.4024, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 9.053497942386832e-06, | |
| "loss": 1.3065, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "eval_accuracy": 0.888, | |
| "eval_loss": 1.1940162181854248, | |
| "eval_runtime": 51.505, | |
| "eval_samples_per_second": 14.562, | |
| "eval_steps_per_second": 0.466, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.641975308641975e-06, | |
| "loss": 1.236, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 8.23045267489712e-06, | |
| "loss": 1.0602, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 7.818930041152263e-06, | |
| "loss": 0.9621, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "eval_accuracy": 0.9493333333333334, | |
| "eval_loss": 0.9100235104560852, | |
| "eval_runtime": 52.3044, | |
| "eval_samples_per_second": 14.339, | |
| "eval_steps_per_second": 0.459, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 7.4074074074074075e-06, | |
| "loss": 0.9084, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 6.9958847736625525e-06, | |
| "loss": 0.792, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "eval_accuracy": 0.9653333333333334, | |
| "eval_loss": 0.7024865746498108, | |
| "eval_runtime": 52.9227, | |
| "eval_samples_per_second": 14.172, | |
| "eval_steps_per_second": 0.453, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 6.584362139917696e-06, | |
| "loss": 0.7528, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 6.17283950617284e-06, | |
| "loss": 0.6486, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 5.761316872427984e-06, | |
| "loss": 0.5976, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "eval_accuracy": 0.972, | |
| "eval_loss": 0.5612229108810425, | |
| "eval_runtime": 56.6976, | |
| "eval_samples_per_second": 13.228, | |
| "eval_steps_per_second": 0.423, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 5.349794238683128e-06, | |
| "loss": 0.5696, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 5.55, | |
| "learning_rate": 4.938271604938272e-06, | |
| "loss": 0.5075, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 4.526748971193416e-06, | |
| "loss": 0.4804, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "eval_accuracy": 0.9746666666666667, | |
| "eval_loss": 0.47045987844467163, | |
| "eval_runtime": 55.7599, | |
| "eval_samples_per_second": 13.451, | |
| "eval_steps_per_second": 0.43, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 4.11522633744856e-06, | |
| "loss": 0.4573, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 3.7037037037037037e-06, | |
| "loss": 0.4194, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "eval_accuracy": 0.976, | |
| "eval_loss": 0.4130578637123108, | |
| "eval_runtime": 52.5365, | |
| "eval_samples_per_second": 14.276, | |
| "eval_steps_per_second": 0.457, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 3.292181069958848e-06, | |
| "loss": 0.421, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 2.880658436213992e-06, | |
| "loss": 0.3781, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 2.469135802469136e-06, | |
| "loss": 0.373, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "eval_accuracy": 0.9773333333333334, | |
| "eval_loss": 0.3774995803833008, | |
| "eval_runtime": 52.9476, | |
| "eval_samples_per_second": 14.165, | |
| "eval_steps_per_second": 0.453, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 2.05761316872428e-06, | |
| "loss": 0.371, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 1.646090534979424e-06, | |
| "loss": 0.3477, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 1.234567901234568e-06, | |
| "loss": 0.3402, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "eval_accuracy": 0.976, | |
| "eval_loss": 0.36156395077705383, | |
| "eval_runtime": 52.2783, | |
| "eval_samples_per_second": 14.346, | |
| "eval_steps_per_second": 0.459, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 8.23045267489712e-07, | |
| "loss": 0.3483, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 9.62, | |
| "learning_rate": 4.11522633744856e-07, | |
| "loss": 0.3323, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "learning_rate": 0.0, | |
| "loss": 0.3308, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "eval_accuracy": 0.976, | |
| "eval_loss": 0.35380691289901733, | |
| "eval_runtime": 54.0465, | |
| "eval_samples_per_second": 13.877, | |
| "eval_steps_per_second": 0.444, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "step": 270, | |
| "total_flos": 2.7088828235890606e+18, | |
| "train_loss": 0.7456211717040451, | |
| "train_runtime": 3840.345, | |
| "train_samples_per_second": 9.114, | |
| "train_steps_per_second": 0.07 | |
| } | |
| ], | |
| "max_steps": 270, | |
| "num_train_epochs": 10, | |
| "total_flos": 2.7088828235890606e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |