mrpc/roberta-base_lr1e-05/trainer_state.json

{
  "best_metric": 0.84512111394795,
  "best_model_checkpoint": "./save_models/mrpc/roberta-base_lr1e-05/checkpoint-117",
  "epoch": 10.0,
  "eval_steps": 500,
  "global_step": 130,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6839237057220708,
      "eval_averaged_scores": 0.7481107201749513,
      "eval_f1": 0.8122977346278317,
      "eval_loss": 0.6196000576019287,
      "eval_runtime": 0.9135,
      "eval_samples_per_second": 401.752,
      "eval_steps_per_second": 2.189,
      "step": 13
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6839237057220708,
      "eval_averaged_scores": 0.7481107201749513,
      "eval_f1": 0.8122977346278317,
      "eval_loss": 0.5717987418174744,
      "eval_runtime": 0.9146,
      "eval_samples_per_second": 401.283,
      "eval_steps_per_second": 2.187,
      "step": 26
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7084468664850136,
      "eval_averaged_scores": 0.7630820363089463,
      "eval_f1": 0.8177172061328791,
      "eval_loss": 0.5101604461669922,
      "eval_runtime": 0.9281,
      "eval_samples_per_second": 395.44,
      "eval_steps_per_second": 2.155,
      "step": 39
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.7956403269754768,
      "eval_averaged_scores": 0.8287593900070754,
      "eval_f1": 0.861878453038674,
      "eval_loss": 0.45689526200294495,
      "eval_runtime": 0.9188,
      "eval_samples_per_second": 399.445,
      "eval_steps_per_second": 2.177,
      "step": 52
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.771117166212534,
      "eval_averaged_scores": 0.7961968809786075,
      "eval_f1": 0.8212765957446809,
      "eval_loss": 0.42808404564857483,
      "eval_runtime": 0.9515,
      "eval_samples_per_second": 385.719,
      "eval_steps_per_second": 2.102,
      "step": 65
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8092643051771117,
      "eval_averaged_scores": 0.8343510280905638,
      "eval_f1": 0.859437751004016,
      "eval_loss": 0.4162532687187195,
      "eval_runtime": 0.9128,
      "eval_samples_per_second": 402.06,
      "eval_steps_per_second": 2.191,
      "step": 78
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.8147138964577657,
      "eval_averaged_scores": 0.8396278645635442,
      "eval_f1": 0.8645418326693227,
      "eval_loss": 0.41786280274391174,
      "eval_runtime": 1.0619,
      "eval_samples_per_second": 345.592,
      "eval_steps_per_second": 1.883,
      "step": 91
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.8065395095367848,
      "eval_averaged_scores": 0.8318411833398209,
      "eval_f1": 0.8571428571428572,
      "eval_loss": 0.4168005883693695,
      "eval_runtime": 0.9219,
      "eval_samples_per_second": 398.082,
      "eval_steps_per_second": 2.169,
      "step": 104
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.8201634877384196,
      "eval_averaged_scores": 0.84512111394795,
      "eval_f1": 0.8700787401574802,
      "eval_loss": 0.42427295446395874,
      "eval_runtime": 0.9217,
      "eval_samples_per_second": 398.165,
      "eval_steps_per_second": 2.17,
      "step": 117
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.8119891008174387,
      "eval_averaged_scores": 0.8374060812238286,
      "eval_f1": 0.8628230616302187,
      "eval_loss": 0.42201152443885803,
      "eval_runtime": 0.9117,
      "eval_samples_per_second": 402.525,
      "eval_steps_per_second": 2.194,
      "step": 130
    },
    {
      "epoch": 10.0,
      "step": 130,
      "total_flos": 1559281933489620.0,
      "train_loss": 0.44359661982609677,
      "train_runtime": 101.3918,
      "train_samples_per_second": 325.569,
      "train_steps_per_second": 1.282
    }
  ],
  "logging_steps": 500,
  "max_steps": 130,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 1559281933489620.0,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}