tinyllama-1.1b-chat-dpo-qlora / eval_results.json
martimfasantos's picture
End of training
bc5dc81 verified
{
"epoch": 1.0,
"eval_logits/chosen": -2.642491102218628,
"eval_logits/rejected": -2.609631299972534,
"eval_logps/chosen": -504.97137451171875,
"eval_logps/rejected": -490.8393249511719,
"eval_loss": 0.6084445714950562,
"eval_rewards/accuracies": 0.6579999923706055,
"eval_rewards/chosen": -1.0874665975570679,
"eval_rewards/margins": 0.3040919303894043,
"eval_rewards/rejected": -1.3915585279464722,
"eval_runtime": 347.6592,
"eval_samples": 2000,
"eval_samples_per_second": 5.753,
"eval_steps_per_second": 0.719
}