trainer_state.json

{
  "best_metric": 0.9218666666666666,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0093/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": 1.080172061920166,
      "learning_rate": 0.0002909538931178862,
      "loss": 0.5942,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8632,
      "eval_loss": 0.45586028695106506,
      "eval_runtime": 10.7965,
      "eval_samples_per_second": 347.334,
      "eval_steps_per_second": 5.465,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": Infinity,
      "learning_rate": 0.0002650076732300463,
      "loss": 0.2796,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8754666666666666,
      "eval_loss": 0.4188299775123596,
      "eval_runtime": 10.3893,
      "eval_samples_per_second": 360.948,
      "eval_steps_per_second": 5.679,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 40.78641128540039,
      "learning_rate": 0.00022513612989278497,
      "loss": 0.1803,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8794666666666666,
      "eval_loss": 0.39748480916023254,
      "eval_runtime": 10.8902,
      "eval_samples_per_second": 344.347,
      "eval_steps_per_second": 5.418,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.011747262440621853,
      "learning_rate": 0.00017620206039927676,
      "loss": 0.0982,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8845333333333333,
      "eval_loss": 0.43565303087234497,
      "eval_runtime": 10.9957,
      "eval_samples_per_second": 341.044,
      "eval_steps_per_second": 5.366,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.07822491228580475,
      "learning_rate": 0.00012410763572038907,
      "loss": 0.0502,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9042666666666667,
      "eval_loss": 0.3838825523853302,
      "eval_runtime": 10.5276,
      "eval_samples_per_second": 356.206,
      "eval_steps_per_second": 5.604,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.542050302028656,
      "learning_rate": 7.513621230422122e-05,
      "loss": 0.0184,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9090666666666667,
      "eval_loss": 0.3779655694961548,
      "eval_runtime": 10.5679,
      "eval_samples_per_second": 354.847,
      "eval_steps_per_second": 5.583,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.009499728679656982,
      "learning_rate": 3.519446655599853e-05,
      "loss": 0.0055,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9184,
      "eval_loss": 0.354938805103302,
      "eval_runtime": 10.6151,
      "eval_samples_per_second": 353.271,
      "eval_steps_per_second": 5.558,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.0032351624686270952,
      "learning_rate": 9.099962490342783e-06,
      "loss": 0.0017,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9218666666666666,
      "eval_loss": 0.3458367586135864,
      "eval_runtime": 10.5448,
      "eval_samples_per_second": 355.625,
      "eval_steps_per_second": 5.595,
      "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
}