Skip to content

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

Open
bigximik wants to merge 9 commits intogrpo-metricsfrom
gspo
Open

gspo: GSPO loss + DeepSpeed parity fixes (loss/grad divisors, SDP, fp32_lm_head, docs_per_step, temperature)#502
bigximik wants to merge 9 commits intogrpo-metricsfrom
gspo

Commits

Commits on Apr 28, 2026

Commits on Apr 29, 2026

Commits on May 4, 2026

Commits on May 5, 2026