[WIP] a16w4 scale b16#3390

Open

yadaish wants to merge 28 commits intodevelopfrom

yandai/moe_flatmm_async_scale_b16

Commits on Nov 28, 2025

update
zhiding512
committed
mixed-prec flatmm pipeline improve
yadaish
committed
support a16_wint4 moe
yadaish
committed
fix out of lds
yadaish
committed
update
yadaish
committed
update
yadaish
committed
update
yadaish
committed
change endian seems working
yadaish
committed
Add validity checks for MoE FlatMM scatter and enable bf16 hardware atomic

msaffari-amd
authored and
yadaish
committed
correct clang-format

msaffari-amd
authored and
yadaish
committed
removed unused rtol_atol variable from example code

msaffari-amd
authored and
yadaish
committed
clang format correction

msaffari-amd
authored and
yadaish
committed
remove unused varable max_accumulated_value from example

msaffari-amd
authored and
yadaish
committed
update
yadaish
committed
update
yadaish
committed
update
yadaish
committed
update
yadaish
committed

Commits on Nov 29, 2025

update
yadaish
committed

Commits on Dec 1, 2025

Commits on Dec 4, 2025

Commits on Dec 5, 2025

update
yadaish
committed

Commits on Dec 9, 2025

update
yadaish
committed