trainer_state.json

{
  "best_metric": 0.9202666666666667,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0943/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": Infinity,
      "learning_rate": 4.811948852959657e-05,
      "loss": 0.7763,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8717333333333334,
      "eval_loss": 0.415194571018219,
      "eval_runtime": 13.177,
      "eval_samples_per_second": 284.587,
      "eval_steps_per_second": 4.478,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.20256295800209045,
      "learning_rate": 4.271931404387096e-05,
      "loss": 0.2474,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8762666666666666,
      "eval_loss": 0.3985234498977661,
      "eval_runtime": 13.1693,
      "eval_samples_per_second": 284.754,
      "eval_steps_per_second": 4.48,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 49.82950973510742,
      "learning_rate": 3.4621534620947785e-05,
      "loss": 0.1235,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.884,
      "eval_loss": 0.39801737666130066,
      "eval_runtime": 13.132,
      "eval_samples_per_second": 285.562,
      "eval_steps_per_second": 4.493,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.1702968031167984,
      "learning_rate": 2.5058963771413844e-05,
      "loss": 0.0667,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.9010666666666667,
      "eval_loss": 0.35740700364112854,
      "eval_runtime": 13.2505,
      "eval_samples_per_second": 283.009,
      "eval_steps_per_second": 4.453,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.005936861038208008,
      "learning_rate": 1.5487416222190068e-05,
      "loss": 0.0276,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9069333333333334,
      "eval_loss": 0.3710463345050812,
      "eval_runtime": 13.0365,
      "eval_samples_per_second": 287.654,
      "eval_steps_per_second": 4.526,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.5627461075782776,
      "learning_rate": 7.36407332135117e-06,
      "loss": 0.0105,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9162666666666667,
      "eval_loss": 0.335354208946228,
      "eval_runtime": 13.1591,
      "eval_samples_per_second": 284.975,
      "eval_steps_per_second": 4.484,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.21349656581878662,
      "learning_rate": 1.925640387263111e-06,
      "loss": 0.0035,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9173333333333333,
      "eval_loss": 0.32604408264160156,
      "eval_runtime": 13.0728,
      "eval_samples_per_second": 286.855,
      "eval_steps_per_second": 4.513,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.01123800314962864,
      "learning_rate": 6.953462348691098e-11,
      "loss": 0.0015,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9202666666666667,
      "eval_loss": 0.3300507664680481,
      "eval_runtime": 14.2571,
      "eval_samples_per_second": 263.026,
      "eval_steps_per_second": 4.138,
      "step": 2664
    }
  ],
  "logging_steps": 500,
  "max_steps": 2664,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 8,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.317930564980736e+19,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}