Qwen2.5-7B-DPO-2209 / eval_results.json
qgallouedec's picture
qgallouedec HF staff
End of training
14b022b verified
raw
history blame contribute delete
482 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": -0.2890625,
"eval_logits/rejected": -0.1591796875,
"eval_logps/chosen": -254.0,
"eval_logps/rejected": -268.0,
"eval_loss": 0.6815624833106995,
"eval_rewards/accuracies": 0.4453125,
"eval_rewards/chosen": -0.08544921875,
"eval_rewards/margins": 0.01904296875,
"eval_rewards/rejected": -0.1044921875,
"eval_runtime": 8.8106,
"eval_samples_per_second": 113.499,
"eval_steps_per_second": 1.816
}