[None][feat] add CuteDSL FP8/FP16 MLA decode attention fmha lib#15138
Open
haow-nv wants to merge 21 commits into
Open
[None][feat] add CuteDSL FP8/FP16 MLA decode attention fmha lib#15138haow-nv wants to merge 21 commits into
haow-nv wants to merge 21 commits into
Commits
Commits on Jul 23, 2026
- committed
[None][feat] CuteDSL MLA decode: fp8 scaling fix, BF16 support, per-layer paged-KV, multi-layer test
committed- committed
- authored andcommitted
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
[None][chore] CuteDSL MLA decode: drop debug kernel-arg dumps, fix autotuner free dim, trim comments
committed- authored
