trainer_state.json

{
  "best_metric": 0.9133333333333333,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/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": 2.0641701221466064,
      "learning_rate": 9.700253128109275e-05,
      "loss": 0.9869,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8272,
      "eval_loss": 0.5782424211502075,
      "eval_runtime": 11.13,
      "eval_samples_per_second": 336.926,
      "eval_steps_per_second": 5.301,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": 3.48075008392334,
      "learning_rate": 8.833589107668211e-05,
      "loss": 0.3785,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8325333333333333,
      "eval_loss": 0.6065165400505066,
      "eval_runtime": 11.8001,
      "eval_samples_per_second": 317.795,
      "eval_steps_per_second": 5.0,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.5759434700012207,
      "learning_rate": 7.504537663092833e-05,
      "loss": 0.2174,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8493333333333334,
      "eval_loss": 0.5362476110458374,
      "eval_runtime": 11.6998,
      "eval_samples_per_second": 320.519,
      "eval_steps_per_second": 5.043,
      "step": 999
    },
    {
      "epoch": 4.0,
      "grad_norm": 4.361108779907227,
      "learning_rate": 5.873402013309226e-05,
      "loss": 0.1271,
      "step": 1332
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8693333333333333,
      "eval_loss": 0.5005639791488647,
      "eval_runtime": 14.1467,
      "eval_samples_per_second": 265.08,
      "eval_steps_per_second": 4.171,
      "step": 1332
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.043982088565826416,
      "learning_rate": 4.136921190679637e-05,
      "loss": 0.0672,
      "step": 1665
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8794666666666666,
      "eval_loss": 0.49878165125846863,
      "eval_runtime": 14.7991,
      "eval_samples_per_second": 253.394,
      "eval_steps_per_second": 3.987,
      "step": 1665
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.038176968693733215,
      "learning_rate": 2.5045404101407076e-05,
      "loss": 0.0278,
      "step": 1998
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8946666666666667,
      "eval_loss": 0.44131216406822205,
      "eval_runtime": 11.0215,
      "eval_samples_per_second": 340.244,
      "eval_steps_per_second": 5.353,
      "step": 1998
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.00500944210216403,
      "learning_rate": 1.1731488851999511e-05,
      "loss": 0.0084,
      "step": 2331
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9098666666666667,
      "eval_loss": 0.36991825699806213,
      "eval_runtime": 11.4103,
      "eval_samples_per_second": 328.652,
      "eval_steps_per_second": 5.171,
      "step": 2331
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.00012373868958093226,
      "learning_rate": 3.033320830114261e-06,
      "loss": 0.0022,
      "step": 2664
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9133333333333333,
      "eval_loss": 0.37062686681747437,
      "eval_runtime": 10.9763,
      "eval_samples_per_second": 341.644,
      "eval_steps_per_second": 5.375,
      "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
}