Skip to content

gspo: dispatch via self._forward = <kernel> instead of wrapper method

0d0185c
Select commit
Loading
Failed to load commit list.
Open

gspo: GSPO loss + DeepSpeed parity fixes (loss/grad divisors, SDP, fp32_lm_head, docs_per_step, temperature) #502

gspo: dispatch via self._forward = <kernel> instead of wrapper method
0d0185c
Select commit
Loading
Failed to load commit list.