trainer_state.json

{
  "best_metric": 0.9362666666666667,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0895/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": 3.3955466747283936,
      "learning_rate": 0.0002909538931178862,
      "loss": 0.5633,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8778666666666667,
      "eval_loss": 0.42347806692123413,
      "eval_runtime": 12.4187,
      "eval_samples_per_second": 301.963,
      "eval_steps_per_second": 4.751,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.3795112669467926,
      "learning_rate": 0.00026490666646784665,
      "loss": 0.2459,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8837333333333334,
      "eval_loss": 0.38822492957115173,
      "eval_runtime": 12.2733,
      "eval_samples_per_second": 305.54,
      "eval_steps_per_second": 4.807,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.08243769407272339,
      "learning_rate": 0.000225,
      "loss": 0.1486,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9032,
      "eval_loss": 0.36264583468437195,
      "eval_runtime": 12.3425,
      "eval_samples_per_second": 303.827,
      "eval_steps_per_second": 4.78,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 6.132674217224121,
      "learning_rate": 0.00017604722665003956,
      "loss": 0.0928,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.9058666666666667,
      "eval_loss": 0.35461053252220154,
      "eval_runtime": 12.3315,
      "eval_samples_per_second": 304.099,
      "eval_steps_per_second": 4.784,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.0071482909843325615,
      "learning_rate": 0.00012395277334996044,
      "loss": 0.0441,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9181333333333334,
      "eval_loss": 0.3312985301017761,
      "eval_runtime": 12.0059,
      "eval_samples_per_second": 312.345,
      "eval_steps_per_second": 4.914,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.3783445358276367,
      "learning_rate": 7.500000000000002e-05,
      "loss": 0.0131,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9282666666666667,
      "eval_loss": 0.31529131531715393,
      "eval_runtime": 12.2743,
      "eval_samples_per_second": 305.516,
      "eval_steps_per_second": 4.807,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": NaN,
      "learning_rate": 3.519446655599853e-05,
      "loss": 0.0035,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9357333333333333,
      "eval_loss": 0.28041183948516846,
      "eval_runtime": 12.2705,
      "eval_samples_per_second": 305.612,
      "eval_steps_per_second": 4.808,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.004883639980107546,
      "learning_rate": 9.099962490342783e-06,
      "loss": 0.0013,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9362666666666667,
      "eval_loss": 0.28732413053512573,
      "eval_runtime": 12.7072,
      "eval_samples_per_second": 295.109,
      "eval_steps_per_second": 4.643,
      "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
}