trainer_state.json

{
  "best_metric": 0.9149333333333334,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0438/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": 3.3952715396881104,
      "learning_rate": 8.895562228895563e-05,
      "loss": 0.9621,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8277333333333333,
      "eval_loss": 0.5887923240661621,
      "eval_runtime": 12.1602,
      "eval_samples_per_second": 308.383,
      "eval_steps_per_second": 4.852,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.1953756958246231,
      "learning_rate": 7.784451117784451e-05,
      "loss": 0.3467,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8530666666666666,
      "eval_loss": 0.48693951964378357,
      "eval_runtime": 12.1219,
      "eval_samples_per_second": 309.356,
      "eval_steps_per_second": 4.867,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": Infinity,
      "learning_rate": 6.676676676676677e-05,
      "loss": 0.1836,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8690666666666667,
      "eval_loss": 0.502564013004303,
      "eval_runtime": 12.0971,
      "eval_samples_per_second": 309.991,
      "eval_steps_per_second": 4.877,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.11381039768457413,
      "learning_rate": 5.565565565565566e-05,
      "loss": 0.1026,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8738666666666667,
      "eval_loss": 0.47562655806541443,
      "eval_runtime": 12.094,
      "eval_samples_per_second": 310.07,
      "eval_steps_per_second": 4.878,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.012065846472978592,
      "learning_rate": 4.4544544544544546e-05,
      "loss": 0.0628,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8725333333333334,
      "eval_loss": 0.48403459787368774,
      "eval_runtime": 12.2154,
      "eval_samples_per_second": 306.989,
      "eval_steps_per_second": 4.83,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.22982530295848846,
      "learning_rate": 3.3433433433433435e-05,
      "loss": 0.0311,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8818666666666667,
      "eval_loss": 0.4638214707374573,
      "eval_runtime": 12.0601,
      "eval_samples_per_second": 310.942,
      "eval_steps_per_second": 4.892,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.11877978593111038,
      "learning_rate": 2.2322322322322323e-05,
      "loss": 0.0133,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.8992,
      "eval_loss": 0.4416622519493103,
      "eval_runtime": 12.3331,
      "eval_samples_per_second": 304.06,
      "eval_steps_per_second": 4.784,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.7936831712722778,
      "learning_rate": 1.1211211211211212e-05,
      "loss": 0.0035,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9149333333333334,
      "eval_loss": 0.3849431574344635,
      "eval_runtime": 12.028,
      "eval_samples_per_second": 311.773,
      "eval_steps_per_second": 4.905,
      "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
}