checkpoint-1500/trainer_state.json

{
  "best_global_step": 1500,
  "best_metric": 1.6885799534410963,
  "best_model_checkpoint": "./SALAMA_NEW1/checkpoint-1500",
  "epoch": 0.8987417615338527,
  "eval_steps": 1500,
  "global_step": 1500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8987417615338527,
      "grad_norm": 3.4230892658233643,
      "learning_rate": 5.679431747992589e-06,
      "loss": 0.0461,
      "step": 1500
    },
    {
      "epoch": 0.8987417615338527,
      "eval_loss": 0.022678062319755554,
      "eval_runtime": 4849.8512,
      "eval_samples_per_second": 2.753,
      "eval_steps_per_second": 0.344,
      "eval_wer": 1.6885799534410963,
      "step": 1500
    }
  ],
  "logging_steps": 1500,
  "max_steps": 3338,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 1500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 1.385209921536e+19,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}