qwen_uncCPO_entropy / all_results.json
yakazimir's picture
End of training
eabe967 verified
raw
history blame contribute delete
786 Bytes
{
"epoch": 2.999297541394882,
"eval_logits/chosen": 7.6457295417785645,
"eval_logits/rejected": 7.3215250968933105,
"eval_logps/chosen": -46.314910888671875,
"eval_logps/rejected": -47.3421516418457,
"eval_loss": 2.0841993375597667e-07,
"eval_rewards/accuracies": 0.5615726709365845,
"eval_rewards/chosen": -46.314910888671875,
"eval_rewards/margins": 1.0272419452667236,
"eval_rewards/rejected": -47.3421516418457,
"eval_runtime": 40.0403,
"eval_samples": 1345,
"eval_samples_per_second": 33.591,
"eval_steps_per_second": 8.417,
"total_flos": 0.0,
"train_loss": 0.008184677160495138,
"train_runtime": 30264.8442,
"train_samples": 59790,
"train_samples_per_second": 5.927,
"train_steps_per_second": 0.185
}