trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 9.0,
  "eval_steps": 500,
  "global_step": 1224,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 3.68,
      "learning_rate": 6.323529411764706e-06,
      "logits/chosen": -2.905975580215454,
      "logits/rejected": -3.2132484912872314,
      "logps/chosen": -347.41741943359375,
      "logps/rejected": -173.79278564453125,
      "loss": 0.0303,
      "rewards/accuracies": 0.9965069890022278,
      "rewards/chosen": 2.506096601486206,
      "rewards/margins": 7.654053211212158,
      "rewards/rejected": -5.147956848144531,
      "step": 500
    },
    {
      "epoch": 7.35,
      "learning_rate": 2.647058823529412e-06,
      "logits/chosen": -2.9062390327453613,
      "logits/rejected": -3.1844594478607178,
      "logps/chosen": -346.119873046875,
      "logps/rejected": -186.7269287109375,
      "loss": 0.0034,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.641273260116577,
      "rewards/margins": 9.2114839553833,
      "rewards/rejected": -6.570209980010986,
      "step": 1000
    }
  ],
  "logging_steps": 500,
  "max_steps": 1360,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}