trainer_state.json

{
  "best_metric": 0.9042666666666667,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0421/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": 4.446112779446113e-05,
      "loss": 1.8615,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7898666666666667,
      "eval_loss": 0.8213516473770142,
      "eval_runtime": 6.0473,
      "eval_samples_per_second": 620.115,
      "eval_steps_per_second": 9.756,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.6264141798019409,
      "learning_rate": 3.8905572238905574e-05,
      "loss": 0.5438,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8410666666666666,
      "eval_loss": 0.5247025489807129,
      "eval_runtime": 5.7436,
      "eval_samples_per_second": 652.901,
      "eval_steps_per_second": 10.272,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.678921103477478,
      "learning_rate": 3.335001668335002e-05,
      "loss": 0.3076,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8658666666666667,
      "eval_loss": 0.4472540020942688,
      "eval_runtime": 5.88,
      "eval_samples_per_second": 637.756,
      "eval_steps_per_second": 10.034,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.4889630973339081,
      "learning_rate": 2.7794461127794462e-05,
      "loss": 0.1891,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8818666666666667,
      "eval_loss": 0.40249159932136536,
      "eval_runtime": 5.8216,
      "eval_samples_per_second": 644.15,
      "eval_steps_per_second": 10.135,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 1.2652322053909302,
      "learning_rate": 2.2238905572238907e-05,
      "loss": 0.1173,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8848,
      "eval_loss": 0.39481320977211,
      "eval_runtime": 5.9995,
      "eval_samples_per_second": 625.054,
      "eval_steps_per_second": 9.834,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.8212692737579346,
      "learning_rate": 1.668335001668335e-05,
      "loss": 0.0749,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8925333333333333,
      "eval_loss": 0.3783826231956482,
      "eval_runtime": 5.7921,
      "eval_samples_per_second": 647.433,
      "eval_steps_per_second": 10.186,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.12563005089759827,
      "learning_rate": 1.1127794461127795e-05,
      "loss": 0.047,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9026666666666666,
      "eval_loss": 0.3663483262062073,
      "eval_runtime": 5.7145,
      "eval_samples_per_second": 656.229,
      "eval_steps_per_second": 10.325,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.07185222208499908,
      "learning_rate": 5.572238905572239e-06,
      "loss": 0.0268,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9042666666666667,
      "eval_loss": 0.36399564146995544,
      "eval_runtime": 6.0531,
      "eval_samples_per_second": 619.519,
      "eval_steps_per_second": 9.747,
      "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
}