trainer_state.json

{
  "best_metric": 0.9013333333333333,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0755/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": 7.878378378378379e-05,
      "loss": 1.486,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8165333333333333,
      "eval_loss": 0.6536134481430054,
      "eval_runtime": 5.7275,
      "eval_samples_per_second": 654.737,
      "eval_steps_per_second": 10.301,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 8.028653144836426,
      "learning_rate": 6.753378378378378e-05,
      "loss": 0.414,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8618666666666667,
      "eval_loss": 0.4709874391555786,
      "eval_runtime": 5.7511,
      "eval_samples_per_second": 652.053,
      "eval_steps_per_second": 10.259,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.216451406478882,
      "learning_rate": 5.628378378378378e-05,
      "loss": 0.2073,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8549333333333333,
      "eval_loss": 0.49632787704467773,
      "eval_runtime": 5.9368,
      "eval_samples_per_second": 631.649,
      "eval_steps_per_second": 9.938,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": Infinity,
      "learning_rate": 4.5067567567567564e-05,
      "loss": 0.1081,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8701333333333333,
      "eval_loss": 0.451526939868927,
      "eval_runtime": 5.6072,
      "eval_samples_per_second": 668.782,
      "eval_steps_per_second": 10.522,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 3.262497663497925,
      "learning_rate": 3.381756756756757e-05,
      "loss": 0.0473,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8850666666666667,
      "eval_loss": 0.4380522668361664,
      "eval_runtime": 5.7643,
      "eval_samples_per_second": 650.555,
      "eval_steps_per_second": 10.235,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.04452311247587204,
      "learning_rate": 2.256756756756757e-05,
      "loss": 0.0158,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8885333333333333,
      "eval_loss": 0.42894694209098816,
      "eval_runtime": 5.6925,
      "eval_samples_per_second": 658.761,
      "eval_steps_per_second": 10.365,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.0880967229604721,
      "learning_rate": 1.1317567567567568e-05,
      "loss": 0.0061,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.8965333333333333,
      "eval_loss": 0.42244744300842285,
      "eval_runtime": 5.7857,
      "eval_samples_per_second": 648.15,
      "eval_steps_per_second": 10.198,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.030032018199563026,
      "learning_rate": 6.756756756756757e-08,
      "loss": 0.0036,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9013333333333333,
      "eval_loss": 0.41273218393325806,
      "eval_runtime": 5.7491,
      "eval_samples_per_second": 652.273,
      "eval_steps_per_second": 10.262,
      "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
}