run-8/checkpoint-1070/trainer_state.json

{
  "best_metric": 0.19721417077066145,
  "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-8/checkpoint-1070",
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 1070,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_loss": 0.6094028353691101,
      "eval_matthews_correlation": 0.0,
      "eval_runtime": 0.7233,
      "eval_samples_per_second": 1442.045,
      "eval_steps_per_second": 91.251,
      "step": 214
    },
    {
      "epoch": 2.0,
      "eval_loss": 0.5870223045349121,
      "eval_matthews_correlation": 0.0,
      "eval_runtime": 0.7787,
      "eval_samples_per_second": 1339.416,
      "eval_steps_per_second": 84.757,
      "step": 428
    },
    {
      "epoch": 2.34,
      "grad_norm": 8.846563339233398,
      "learning_rate": 2.2213050925129666e-06,
      "loss": 0.5957,
      "step": 500
    },
    {
      "epoch": 3.0,
      "eval_loss": 0.5888075232505798,
      "eval_matthews_correlation": 0.11382192951310593,
      "eval_runtime": 0.8014,
      "eval_samples_per_second": 1301.501,
      "eval_steps_per_second": 82.358,
      "step": 642
    },
    {
      "epoch": 4.0,
      "eval_loss": 0.6100687384605408,
      "eval_matthews_correlation": 0.17858768573654893,
      "eval_runtime": 0.7597,
      "eval_samples_per_second": 1372.896,
      "eval_steps_per_second": 86.876,
      "step": 856
    },
    {
      "epoch": 4.67,
      "grad_norm": 7.084113121032715,
      "learning_rate": 2.7279185346650466e-07,
      "loss": 0.4788,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "eval_loss": 0.6058104038238525,
      "eval_matthews_correlation": 0.19721417077066145,
      "eval_runtime": 0.7784,
      "eval_samples_per_second": 1340.012,
      "eval_steps_per_second": 84.795,
      "step": 1070
    }
  ],
  "logging_steps": 500,
  "max_steps": 1070,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 16595681287872.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 4.169818331559429e-06,
    "num_train_epochs": 5,
    "per_device_train_batch_size": 4,
    "seed": 38
  }
}