{
  "best_metric": 0.3883848786354065,
  "best_model_checkpoint": "./model/checkpoint-4252",
  "epoch": 4.999412110523222,
  "eval_steps": 500,
  "global_step": 4252,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 0.0002850705882352941,
      "loss": 2.6138,
      "step": 850
    },
    {
      "epoch": 1.0,
      "eval_cer": 0.27268008948545863,
      "eval_cer_best": 0.7273199105145414,
      "eval_loss": 0.7238138318061829,
      "eval_runtime": 95.5312,
      "eval_samples_per_second": 31.665,
      "eval_steps_per_second": 3.967,
      "step": 850
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0002700705882352941,
      "loss": 1.0825,
      "step": 1701
    },
    {
      "epoch": 2.0,
      "eval_cer": 0.2253959731543624,
      "eval_cer_best": 0.7746040268456376,
      "eval_loss": 0.5289922952651978,
      "eval_runtime": 95.2118,
      "eval_samples_per_second": 31.771,
      "eval_steps_per_second": 3.981,
      "step": 1701
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0002550882352941176,
      "loss": 0.9076,
      "step": 2551
    },
    {
      "epoch": 3.0,
      "eval_cer": 0.20298881431767338,
      "eval_cer_best": 0.7970111856823267,
      "eval_loss": 0.46624556183815,
      "eval_runtime": 94.6631,
      "eval_samples_per_second": 31.955,
      "eval_steps_per_second": 4.004,
      "step": 2551
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00024008823529411761,
      "loss": 0.804,
      "step": 3402
    },
    {
      "epoch": 4.0,
      "eval_cer": 0.18314093959731545,
      "eval_cer_best": 0.8168590604026845,
      "eval_loss": 0.42872071266174316,
      "eval_runtime": 94.4283,
      "eval_samples_per_second": 32.035,
      "eval_steps_per_second": 4.014,
      "step": 3402
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.00022510588235294114,
      "loss": 0.7448,
      "step": 4252
    },
    {
      "epoch": 5.0,
      "eval_cer": 0.1720626398210291,
      "eval_cer_best": 0.8279373601789709,
      "eval_loss": 0.3883848786354065,
      "eval_runtime": 94.0362,
      "eval_samples_per_second": 32.168,
      "eval_steps_per_second": 4.03,
      "step": 4252
    }
  ],
  "logging_steps": 500,
  "max_steps": 17000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 6.943311650462615e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}