checkpoint-20/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.005234231876472127,
  "eval_steps": 20,
  "global_step": 20,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "grad_norm": 14.02274227142334,
      "learning_rate": 0.0,
      "log_odds_chosen": 0.0881616473197937,
      "log_odds_ratio": -0.6889785528182983,
      "logits/chosen": -2.7954044342041016,
      "logits/rejected": -2.7376980781555176,
      "logps/chosen": -1.0155786275863647,
      "logps/rejected": -1.0806159973144531,
      "loss": 2.5353,
      "nll_loss": 2.466395854949951,
      "rewards/accuracies": 0.5093749761581421,
      "rewards/chosen": -0.10155787318944931,
      "rewards/margins": 0.0065037282183766365,
      "rewards/rejected": -0.10806159675121307,
      "step": 20
    },
    {
      "epoch": 0.01,
      "eval_log_odds_chosen": 0.1270170509815216,
      "eval_log_odds_ratio": -0.6689257621765137,
      "eval_logits/chosen": -2.936189889907837,
      "eval_logits/rejected": -2.9139583110809326,
      "eval_logps/chosen": -1.0113662481307983,
      "eval_logps/rejected": -1.0981656312942505,
      "eval_loss": 1.7056713104248047,
      "eval_nll_loss": 1.638778567314148,
      "eval_rewards/accuracies": 0.5649999976158142,
      "eval_rewards/chosen": -0.10113661736249924,
      "eval_rewards/margins": 0.008679931052029133,
      "eval_rewards/rejected": -0.10981655865907669,
      "eval_runtime": 4862.3267,
      "eval_samples_per_second": 0.411,
      "eval_steps_per_second": 0.103,
      "step": 20
    }
  ],
  "logging_steps": 20,
  "max_steps": 20,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 10,
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}