checkpoint-25/trainer_state.json

{
  "best_metric": 5.88370418548584,
  "best_model_checkpoint": "./Qwen1-5-4B-Chat-hindi-sft/checkpoint-25",
  "epoch": 0.003574747980267391,
  "eval_steps": 25,
  "global_step": 25,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "grad_norm": 5.597507476806641,
      "learning_rate": 7.936507936507936e-06,
      "loss": 6.5107,
      "step": 25
    },
    {
      "epoch": 0.0,
      "eval_loss": 5.88370418548584,
      "eval_runtime": 27.753,
      "eval_samples_per_second": 1.045,
      "eval_steps_per_second": 1.045,
      "step": 25
    }
  ],
  "logging_steps": 25,
  "max_steps": 20979,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 25,
  "total_flos": 500129031075840.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}