trainer_state.json

{
  "best_metric": 0.9645333333333334,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0349/checkpoints/checkpoint-2664",
  "epoch": 8.0,
  "eval_steps": 500,
  "global_step": 2664,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.32852640748023987,
      "learning_rate": 6.788924172750679e-05,
      "loss": 0.633,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.9458666666666666,
      "eval_loss": 0.19485706090927124,
      "eval_runtime": 11.886,
      "eval_samples_per_second": 315.498,
      "eval_steps_per_second": 4.964,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.6569105386734009,
      "learning_rate": 6.181155550916422e-05,
      "loss": 0.0846,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.9544,
      "eval_loss": 0.1778106391429901,
      "eval_runtime": 10.4692,
      "eval_samples_per_second": 358.193,
      "eval_steps_per_second": 5.636,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.030999889597296715,
      "learning_rate": 5.2499999999999995e-05,
      "loss": 0.0299,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9533333333333334,
      "eval_loss": 0.18227095901966095,
      "eval_runtime": 10.2858,
      "eval_samples_per_second": 364.58,
      "eval_steps_per_second": 5.736,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.04916199669241905,
      "learning_rate": 4.1077686218342566e-05,
      "loss": 0.0097,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.96,
      "eval_loss": 0.16996051371097565,
      "eval_runtime": 11.0808,
      "eval_samples_per_second": 338.422,
      "eval_steps_per_second": 5.325,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.007740108296275139,
      "learning_rate": 2.8922313781657438e-05,
      "loss": 0.0026,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9629333333333333,
      "eval_loss": 0.15865787863731384,
      "eval_runtime": 11.772,
      "eval_samples_per_second": 318.554,
      "eval_steps_per_second": 5.012,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.01661488227546215,
      "learning_rate": 1.7500000000000005e-05,
      "loss": 0.0012,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9642666666666667,
      "eval_loss": 0.1568000316619873,
      "eval_runtime": 10.2726,
      "eval_samples_per_second": 365.05,
      "eval_steps_per_second": 5.743,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": NaN,
      "learning_rate": 8.212042196399656e-06,
      "loss": 0.0009,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9637333333333333,
      "eval_loss": 0.15844319760799408,
      "eval_runtime": 10.2849,
      "eval_samples_per_second": 364.611,
      "eval_steps_per_second": 5.737,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.01459230575710535,
      "learning_rate": 2.1233245810799825e-06,
      "loss": 0.0008,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9645333333333334,
      "eval_loss": 0.15783073008060455,
      "eval_runtime": 10.0762,
      "eval_samples_per_second": 372.163,
      "eval_steps_per_second": 5.855,
      "step": 2664
    }
  ],
  "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": false
      },
      "attributes": {}
    }
  },
  "total_flos": 1.317930564980736e+19,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}