trainer_state.json

{
  "best_metric": 0.9245333333333333,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0859/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": 8.757507507507508e-05,
      "loss": 0.8211,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8346666666666667,
      "eval_loss": 0.5296463370323181,
      "eval_runtime": 11.6661,
      "eval_samples_per_second": 321.445,
      "eval_steps_per_second": 5.057,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.017846249043941498,
      "learning_rate": 7.507507507507507e-05,
      "loss": 0.2853,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8688,
      "eval_loss": 0.4697921872138977,
      "eval_runtime": 11.7582,
      "eval_samples_per_second": 318.928,
      "eval_steps_per_second": 5.018,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.1291416883468628,
      "learning_rate": 6.257507507507508e-05,
      "loss": 0.1457,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8666666666666667,
      "eval_loss": 0.46728795766830444,
      "eval_runtime": 11.728,
      "eval_samples_per_second": 319.746,
      "eval_steps_per_second": 5.031,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.5224069952964783,
      "learning_rate": 5.0075075075075074e-05,
      "loss": 0.0768,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8973333333333333,
      "eval_loss": 0.4047873020172119,
      "eval_runtime": 10.6974,
      "eval_samples_per_second": 350.553,
      "eval_steps_per_second": 5.515,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.2307441383600235,
      "learning_rate": 3.7575075075075075e-05,
      "loss": 0.0374,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8957333333333334,
      "eval_loss": 0.41717272996902466,
      "eval_runtime": 11.564,
      "eval_samples_per_second": 324.283,
      "eval_steps_per_second": 5.102,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.00016157765639945865,
      "learning_rate": 2.5075075075075076e-05,
      "loss": 0.0126,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9122666666666667,
      "eval_loss": 0.3856476843357086,
      "eval_runtime": 10.8889,
      "eval_samples_per_second": 344.388,
      "eval_steps_per_second": 5.418,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.010492057539522648,
      "learning_rate": 1.2575075075075077e-05,
      "loss": 0.0031,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9189333333333334,
      "eval_loss": 0.3648628890514374,
      "eval_runtime": 10.8703,
      "eval_samples_per_second": 344.977,
      "eval_steps_per_second": 5.428,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.010752501897513866,
      "learning_rate": 7.507507507507508e-08,
      "loss": 0.0006,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9245333333333333,
      "eval_loss": 0.34505754709243774,
      "eval_runtime": 10.9686,
      "eval_samples_per_second": 341.884,
      "eval_steps_per_second": 5.379,
      "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
}