trainer_state.json

{
  "best_metric": 0.9194666666666667,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0547/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": 8.006006006006006e-05,
      "loss": 0.8999,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8442666666666667,
      "eval_loss": 0.5259901881217957,
      "eval_runtime": 6.1571,
      "eval_samples_per_second": 609.058,
      "eval_steps_per_second": 9.583,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 50.85334014892578,
      "learning_rate": 7.006006006006006e-05,
      "loss": 0.3078,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8658666666666667,
      "eval_loss": 0.46754375100135803,
      "eval_runtime": 6.1434,
      "eval_samples_per_second": 610.409,
      "eval_steps_per_second": 9.604,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.7194852828979492,
      "learning_rate": 6.006006006006006e-05,
      "loss": 0.1647,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8709333333333333,
      "eval_loss": 0.4819689393043518,
      "eval_runtime": 6.1779,
      "eval_samples_per_second": 607.006,
      "eval_steps_per_second": 9.55,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 26.931772232055664,
      "learning_rate": 5.0060060060060066e-05,
      "loss": 0.0972,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8842666666666666,
      "eval_loss": 0.47622746229171753,
      "eval_runtime": 7.009,
      "eval_samples_per_second": 535.026,
      "eval_steps_per_second": 8.418,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 8.519587516784668,
      "learning_rate": 4.006006006006007e-05,
      "loss": 0.0535,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8786666666666667,
      "eval_loss": 0.4686780273914337,
      "eval_runtime": 7.2363,
      "eval_samples_per_second": 518.217,
      "eval_steps_per_second": 8.153,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 1.569627046585083,
      "learning_rate": 3.0060060060060065e-05,
      "loss": 0.0309,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8922666666666667,
      "eval_loss": 0.46777641773223877,
      "eval_runtime": 6.0002,
      "eval_samples_per_second": 624.984,
      "eval_steps_per_second": 9.833,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.032112717628479004,
      "learning_rate": 2.0060060060060062e-05,
      "loss": 0.0102,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9088,
      "eval_loss": 0.42521968483924866,
      "eval_runtime": 6.0679,
      "eval_samples_per_second": 618.002,
      "eval_steps_per_second": 9.723,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.020256856456398964,
      "learning_rate": 1.0060060060060061e-05,
      "loss": 0.0042,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9194666666666667,
      "eval_loss": 0.3713622987270355,
      "eval_runtime": 6.0342,
      "eval_samples_per_second": 621.462,
      "eval_steps_per_second": 9.778,
      "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
}