all_results.json

{
    "epoch": 0.9880609304240429,
    "eval_logits/chosen": 9.252660751342773,
    "eval_logits/rejected": 10.067312240600586,
    "eval_logps/chosen": -109.64786529541016,
    "eval_logps/ref_chosen": -106.16099548339844,
    "eval_logps/ref_rejected": -119.78473663330078,
    "eval_logps/rejected": -124.11076354980469,
    "eval_loss": 0.6891200542449951,
    "eval_rewards/accuracies": 0.604629635810852,
    "eval_rewards/chosen": -0.034868817776441574,
    "eval_rewards/margins": 0.008391385897994041,
    "eval_rewards/rejected": -0.043260201811790466,
    "eval_runtime": 161.4409,
    "eval_samples": 1080,
    "eval_samples_per_second": 6.69,
    "eval_steps_per_second": 1.672,
    "total_flos": 0.0,
    "train_loss": 0.6916100986798605,
    "train_runtime": 8570.6786,
    "train_samples": 9715,
    "train_samples_per_second": 1.134,
    "train_steps_per_second": 0.009
}