trainer_state.json

{
  "best_metric": 0.9514666666666667,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0793/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": 0.490887314081192,
      "learning_rate": 4.4444444444444447e-05,
      "loss": 0.8303,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.9266666666666666,
      "eval_loss": 0.25579723715782166,
      "eval_runtime": 12.5104,
      "eval_samples_per_second": 299.751,
      "eval_steps_per_second": 4.716,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.7050927877426147,
      "learning_rate": 3.888888888888889e-05,
      "loss": 0.138,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.936,
      "eval_loss": 0.21508587896823883,
      "eval_runtime": 12.6174,
      "eval_samples_per_second": 297.208,
      "eval_steps_per_second": 4.676,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.0586836114525795,
      "learning_rate": 3.3333333333333335e-05,
      "loss": 0.0586,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9341333333333334,
      "eval_loss": 0.23093150556087494,
      "eval_runtime": 12.3516,
      "eval_samples_per_second": 303.604,
      "eval_steps_per_second": 4.777,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.020262569189071655,
      "learning_rate": 2.777777777777778e-05,
      "loss": 0.026,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.948,
      "eval_loss": 0.19068004190921783,
      "eval_runtime": 11.4633,
      "eval_samples_per_second": 327.132,
      "eval_steps_per_second": 5.147,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.04259951040148735,
      "learning_rate": 2.2222222222222223e-05,
      "loss": 0.0113,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9458666666666666,
      "eval_loss": 0.19669321179389954,
      "eval_runtime": 12.19,
      "eval_samples_per_second": 307.63,
      "eval_steps_per_second": 4.84,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.021843507885932922,
      "learning_rate": 1.6666666666666667e-05,
      "loss": 0.0048,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9474666666666667,
      "eval_loss": 0.20668993890285492,
      "eval_runtime": 12.2602,
      "eval_samples_per_second": 305.867,
      "eval_steps_per_second": 4.812,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": NaN,
      "learning_rate": 1.1127794461127795e-05,
      "loss": 0.0028,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9477333333333333,
      "eval_loss": 0.20840458571910858,
      "eval_runtime": 12.0888,
      "eval_samples_per_second": 310.204,
      "eval_steps_per_second": 4.881,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.012012549676001072,
      "learning_rate": 5.572238905572239e-06,
      "loss": 0.002,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9514666666666667,
      "eval_loss": 0.2093937247991562,
      "eval_runtime": 13.5419,
      "eval_samples_per_second": 276.919,
      "eval_steps_per_second": 4.357,
      "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
}