all_results.json

{
    "epoch": 0.9993060374739764,
    "eval_logits/chosen": -0.16716668009757996,
    "eval_logits/rejected": 0.0005068429163657129,
    "eval_logps/chosen": -725.14599609375,
    "eval_logps/rejected": -973.1326293945312,
    "eval_loss": 0.3478291928768158,
    "eval_rewards/accuracies": 0.8355262875556946,
    "eval_rewards/chosen": -2.6376235485076904,
    "eval_rewards/margins": 2.2410366535186768,
    "eval_rewards/rejected": -4.878660202026367,
    "eval_runtime": 380.6416,
    "eval_samples": 8491,
    "eval_samples_per_second": 22.307,
    "eval_steps_per_second": 0.349,
    "total_flos": 0.0,
    "train_loss": 0.21818815544247627,
    "train_runtime": 20177.9124,
    "train_samples": 184443,
    "train_samples_per_second": 9.141,
    "train_steps_per_second": 0.036
}