trainer_state.json

{
  "best_metric": 0.9477333333333333,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0177/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": 0.6684163808822632,
      "learning_rate": 4.4444444444444447e-05,
      "loss": 0.8461,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.9290666666666667,
      "eval_loss": 0.2558877170085907,
      "eval_runtime": 7.3092,
      "eval_samples_per_second": 513.055,
      "eval_steps_per_second": 8.072,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": Infinity,
      "learning_rate": 3.8905572238905574e-05,
      "loss": 0.1437,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.9306666666666666,
      "eval_loss": 0.23281769454479218,
      "eval_runtime": 6.8963,
      "eval_samples_per_second": 543.771,
      "eval_steps_per_second": 8.555,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.1854402720928192,
      "learning_rate": 3.335001668335002e-05,
      "loss": 0.0615,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9378666666666666,
      "eval_loss": 0.24537737667560577,
      "eval_runtime": 7.2331,
      "eval_samples_per_second": 518.452,
      "eval_steps_per_second": 8.157,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.04183875769376755,
      "learning_rate": 2.7794461127794462e-05,
      "loss": 0.0303,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.9429333333333333,
      "eval_loss": 0.23711465299129486,
      "eval_runtime": 7.1918,
      "eval_samples_per_second": 521.427,
      "eval_steps_per_second": 8.204,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.05658993497490883,
      "learning_rate": 2.2238905572238907e-05,
      "loss": 0.0126,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9445333333333333,
      "eval_loss": 0.22240392863750458,
      "eval_runtime": 7.1072,
      "eval_samples_per_second": 527.634,
      "eval_steps_per_second": 8.301,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.04011552780866623,
      "learning_rate": 1.668335001668335e-05,
      "loss": 0.0065,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9445333333333333,
      "eval_loss": 0.25139573216438293,
      "eval_runtime": 6.9312,
      "eval_samples_per_second": 541.034,
      "eval_steps_per_second": 8.512,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.01762446016073227,
      "learning_rate": 1.1127794461127795e-05,
      "loss": 0.0043,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.944,
      "eval_loss": 0.23845496773719788,
      "eval_runtime": 7.0232,
      "eval_samples_per_second": 533.948,
      "eval_steps_per_second": 8.401,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.013406623154878616,
      "learning_rate": 5.572238905572239e-06,
      "loss": 0.003,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9477333333333333,
      "eval_loss": 0.24203184247016907,
      "eval_runtime": 7.576,
      "eval_samples_per_second": 494.985,
      "eval_steps_per_second": 7.788,
      "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
}