trainer_state.json

{
  "best_metric": 0.9170666666666667,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0229/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": 7.608119487762451,
      "learning_rate": 0.0002666666666666666,
      "loss": 0.6867,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8410666666666666,
      "eval_loss": 0.5290897488594055,
      "eval_runtime": 6.5915,
      "eval_samples_per_second": 568.917,
      "eval_steps_per_second": 8.951,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": Infinity,
      "learning_rate": 0.00023343343343343342,
      "loss": 0.3237,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8664,
      "eval_loss": 0.45978105068206787,
      "eval_runtime": 5.8218,
      "eval_samples_per_second": 644.132,
      "eval_steps_per_second": 10.134,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.614945888519287,
      "learning_rate": 0.00020010010010010007,
      "loss": 0.2072,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8810666666666667,
      "eval_loss": 0.41819220781326294,
      "eval_runtime": 6.3464,
      "eval_samples_per_second": 590.889,
      "eval_steps_per_second": 9.297,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.334083616733551,
      "learning_rate": 0.00016676676676676675,
      "loss": 0.1332,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8869333333333334,
      "eval_loss": 0.4150986075401306,
      "eval_runtime": 7.3794,
      "eval_samples_per_second": 508.174,
      "eval_steps_per_second": 7.995,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.11773234605789185,
      "learning_rate": 0.00013343343343343343,
      "loss": 0.0843,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.888,
      "eval_loss": 0.4493732750415802,
      "eval_runtime": 7.2152,
      "eval_samples_per_second": 519.733,
      "eval_steps_per_second": 8.177,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.5204492211341858,
      "learning_rate": 0.00010010010010010008,
      "loss": 0.0464,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8962666666666667,
      "eval_loss": 0.42368218302726746,
      "eval_runtime": 6.1702,
      "eval_samples_per_second": 607.763,
      "eval_steps_per_second": 9.562,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.012271203100681305,
      "learning_rate": 6.676676676676677e-05,
      "loss": 0.02,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9082666666666667,
      "eval_loss": 0.4151240289211273,
      "eval_runtime": 5.8563,
      "eval_samples_per_second": 640.338,
      "eval_steps_per_second": 10.075,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.006078313570469618,
      "learning_rate": 3.343343343343343e-05,
      "loss": 0.0066,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9170666666666667,
      "eval_loss": 0.37969592213630676,
      "eval_runtime": 5.7707,
      "eval_samples_per_second": 649.833,
      "eval_steps_per_second": 10.224,
      "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
}