checkpoint-252/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.8064,
  "eval_steps": 63,
  "global_step": 252,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2016,
      "eval_logps/chosen": -106.34315490722656,
      "eval_logps/ref_chosen": -70.265625,
      "eval_logps/ref_rejected": -63.51186752319336,
      "eval_logps/rejected": -106.34315490722656,
      "eval_loss": 0.6557361483573914,
      "eval_rewards/accuracies": 0.6162024772678456,
      "eval_rewards/chosen": -1.8038759231567383,
      "eval_rewards/margins": 0.1433669477701187,
      "eval_rewards/rejected": -1.9472428560256958,
      "eval_runtime": 2599.3441,
      "eval_samples_per_second": 32.24,
      "eval_steps_per_second": 4.03,
      "step": 63
    },
    {
      "epoch": 0.32,
      "grad_norm": 5.536309242248535,
      "learning_rate": 2.3016987391917016e-05,
      "logps/chosen": -187.26882934570312,
      "logps/ref_chosen": -199.84158325195312,
      "logps/ref_rejected": -140.63331604003906,
      "logps/rejected": -187.26882934570312,
      "loss": 0.4864,
      "rewards/accuracies": 0.7975000143051147,
      "rewards/chosen": 0.6286371350288391,
      "rewards/margins": 0.6605393886566162,
      "rewards/rejected": -0.031902212649583817,
      "step": 100
    },
    {
      "epoch": 0.4032,
      "eval_logps/chosen": -123.7810287475586,
      "eval_logps/ref_chosen": -70.265625,
      "eval_logps/ref_rejected": -63.51186752319336,
      "eval_logps/rejected": -123.7810287475586,
      "eval_loss": 0.6727969646453857,
      "eval_rewards/accuracies": 0.5971814515166702,
      "eval_rewards/chosen": -2.6757709980010986,
      "eval_rewards/margins": 0.16077223420143127,
      "eval_rewards/rejected": -2.836543321609497,
      "eval_runtime": 2597.2346,
      "eval_samples_per_second": 32.266,
      "eval_steps_per_second": 4.034,
      "step": 126
    },
    {
      "epoch": 0.6048,
      "eval_logps/chosen": -136.31578063964844,
      "eval_logps/ref_chosen": -70.265625,
      "eval_logps/ref_rejected": -63.51186752319336,
      "eval_logps/rejected": -136.31578063964844,
      "eval_loss": 0.683315098285675,
      "eval_rewards/accuracies": 0.5913224028066156,
      "eval_rewards/chosen": -3.3025076389312744,
      "eval_rewards/margins": 0.17308557033538818,
      "eval_rewards/rejected": -3.475592851638794,
      "eval_runtime": 2597.1774,
      "eval_samples_per_second": 32.267,
      "eval_steps_per_second": 4.034,
      "step": 189
    },
    {
      "epoch": 0.64,
      "grad_norm": 6.213208198547363,
      "learning_rate": 8.569611578954186e-06,
      "logps/chosen": -181.8887176513672,
      "logps/ref_chosen": -198.8422088623047,
      "logps/ref_rejected": -139.47769165039062,
      "logps/rejected": -181.8887176513672,
      "loss": 0.3802,
      "rewards/accuracies": 0.8515625,
      "rewards/chosen": 0.8476755023002625,
      "rewards/margins": 1.189441204071045,
      "rewards/rejected": -0.3417656719684601,
      "step": 200
    },
    {
      "epoch": 0.8064,
      "eval_logps/chosen": -139.97152709960938,
      "eval_logps/ref_chosen": -70.265625,
      "eval_logps/ref_rejected": -63.51186752319336,
      "eval_logps/rejected": -139.97152709960938,
      "eval_loss": 0.6932100653648376,
      "eval_rewards/accuracies": 0.5844013269373046,
      "eval_rewards/chosen": -3.4852941036224365,
      "eval_rewards/margins": 0.16311165690422058,
      "eval_rewards/rejected": -3.6484062671661377,
      "eval_runtime": 2597.2713,
      "eval_samples_per_second": 32.265,
      "eval_steps_per_second": 4.033,
      "step": 252
    }
  ],
  "logging_steps": 100,
  "max_steps": 312,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 63,
  "total_flos": 0.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}