run-1/checkpoint-4210/trainer_state.json

{
  "best_metric": 0.8944954128440367,
  "best_model_checkpoint": "distilbert-base-uncased-finetuned-sst2/run-1/checkpoint-4210",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 4210,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.1187648456057007,
      "grad_norm": 10.556523323059082,
      "learning_rate": 3.3640213484135133e-06,
      "loss": 0.4628,
      "step": 500
    },
    {
      "epoch": 0.2375296912114014,
      "grad_norm": 27.427276611328125,
      "learning_rate": 3.261083118657898e-06,
      "loss": 0.3038,
      "step": 1000
    },
    {
      "epoch": 0.35629453681710216,
      "grad_norm": 10.17439079284668,
      "learning_rate": 3.1581448889022824e-06,
      "loss": 0.2837,
      "step": 1500
    },
    {
      "epoch": 0.4750593824228028,
      "grad_norm": 13.1768159866333,
      "learning_rate": 3.0552066591466673e-06,
      "loss": 0.2588,
      "step": 2000
    },
    {
      "epoch": 0.5938242280285035,
      "grad_norm": 7.38680362701416,
      "learning_rate": 2.952268429391051e-06,
      "loss": 0.2554,
      "step": 2500
    },
    {
      "epoch": 0.7125890736342043,
      "grad_norm": 13.285799980163574,
      "learning_rate": 2.849330199635436e-06,
      "loss": 0.2433,
      "step": 3000
    },
    {
      "epoch": 0.831353919239905,
      "grad_norm": 10.867403984069824,
      "learning_rate": 2.7463919698798204e-06,
      "loss": 0.2491,
      "step": 3500
    },
    {
      "epoch": 0.9501187648456056,
      "grad_norm": 8.815483093261719,
      "learning_rate": 2.643453740124205e-06,
      "loss": 0.232,
      "step": 4000
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8944954128440367,
      "eval_loss": 0.2890743613243103,
      "eval_runtime": 1.2151,
      "eval_samples_per_second": 717.616,
      "eval_steps_per_second": 45.262,
      "step": 4210
    }
  ],
  "logging_steps": 500,
  "max_steps": 16840,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 582173519062080.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 3.466959578169129e-06,
    "num_train_epochs": 4,
    "per_device_train_batch_size": 16,
    "seed": 22
  }
}