Qwen2.5-7B-DPO-main / eval_results.json
qgallouedec's picture
qgallouedec HF staff
Training in progress, step 500
f775025 verified
raw
history blame
495 Bytes
{
"epoch": 0.02059732234809475,
"eval_logits/chosen": -0.6875,
"eval_logits/rejected": -0.73046875,
"eval_logps/chosen": -284.0,
"eval_logps/rejected": -308.0,
"eval_loss": 0.695437490940094,
"eval_rewards/accuracies": 0.296875,
"eval_rewards/chosen": 0.0113525390625,
"eval_rewards/margins": -0.0233154296875,
"eval_rewards/rejected": 0.03466796875,
"eval_runtime": 8.5701,
"eval_samples_per_second": 116.685,
"eval_steps_per_second": 1.867
}