{ "epoch": 1.0, "eval_logits/chosen": -0.043021149933338165, "eval_logits/rejected": 4.756043434143066, "eval_logps/chosen": -587.93701171875, "eval_logps/rejected": -9136.2529296875, "eval_loss": 0.0007141114911064506, "eval_rewards/accuracies": 1.0, "eval_rewards/chosen": -2.2095537185668945, "eval_rewards/margins": 82.74508666992188, "eval_rewards/rejected": -84.95464324951172, "eval_runtime": 192.6276, "eval_samples": 3905, "eval_samples_per_second": 20.272, "eval_steps_per_second": 0.322, "total_flos": 0.0, "train_loss": 0.08915322791869006, "train_runtime": 10109.537, "train_samples": 45000, "train_samples_per_second": 4.451, "train_steps_per_second": 0.035 }