perf(cpu-moe): extend expert dedup to ds_fp4 and mxfp4 - #49

Open

gdevenyi wants to merge 3 commits into

FlashML-org:mainfrom

gdevenyi:perf/cpu-moe-dedup-fp4

Commits on Aug 23, 2026

perf(cpu-moe): dedup experts across the batch in pass 1

gdevenyi
and
claude
committed
49d8a0f
perf(cpu-moe): dedup experts in pass 2 as well

gdevenyi
and
claude
committed
a60a654
perf(cpu-moe): extend expert dedup to ds_fp4 and mxfp4

gdevenyi
and
claude
committed
4a6d86b