Skip to content

perf(ds4): accelerate ROCm decode and long-context selection - #667

Draft
davide221 wants to merge 6 commits into
codex/ds4-pflash-long-contextfrom
codex/ds4-rocm-moe-wave32
Draft

perf(ds4): accelerate ROCm decode and long-context selection#667
davide221 wants to merge 6 commits into
codex/ds4-pflash-long-contextfrom
codex/ds4-rocm-moe-wave32

hip: split indexed MLA decode across KV blocks

0f1264c
Select commit
Loading
Failed to load commit list.

There are no checks for this commit