Modelpublic

hkr04/distill-1.5b-grpo-minmax

sourceHugging Faceupdated 2mo agoView on Hugging Face

0likes4downloads

Model Card

Trained on DAPO-Math-17k using GRPO implemented by VeRL.

—Batch Size: 32
—Group Size: 8
—Step: 280
—Max Response Length: 8192