checkpoint-50/trainer_state.json

{
  "best_metric": 5.878845691680908,
  "best_model_checkpoint": "./Qwen1-5-4B-Chat-hindi-sft/checkpoint-50",
  "epoch": 0.007214486689272059,
  "eval_steps": 25,
  "global_step": 50,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "grad_norm": 4.076544761657715,
      "learning_rate": 4.807692307692308e-06,
      "loss": 6.0722,
      "step": 25
    },
    {
      "epoch": 0.0,
      "eval_loss": 6.239687919616699,
      "eval_runtime": 258.5805,
      "eval_samples_per_second": 1.087,
      "eval_steps_per_second": 1.087,
      "step": 25
    },
    {
      "epoch": 0.01,
      "grad_norm": 2.1901891231536865,
      "learning_rate": 9.615384615384616e-06,
      "loss": 6.1105,
      "step": 50
    },
    {
      "epoch": 0.01,
      "eval_loss": 5.878845691680908,
      "eval_runtime": 259.0078,
      "eval_samples_per_second": 1.085,
      "eval_steps_per_second": 1.085,
      "step": 50
    }
  ],
  "logging_steps": 25,
  "max_steps": 34650,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 25,
  "total_flos": 1263381335884800.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}