Skinny gemm - #71

Draft

tjtanaa wants to merge 15 commits into

Commits on Jul 3, 2025

add skinnnmy-gemm
vllmellm
committed
b1d470b
enable aiter mla on v1 engine
vllmellm
committed
a478a58
enable code path to be executed
vllmellm
committed
027c38d
fix timout issues
vllmellm
committed
58d8092
return back sampler to match the vllm version
vllmellm
committed
ba6c976
add cu kernel
vllmellm
committed
dd10ced
format
vllmellm
committed
643e586
revert back to old version where qk projections were not merged
vllmellm
committed
cc44106
bugfix
vllmellm
committed
ea74777
remove print statement
vllmellm
committed
c49aec0
unquantized tgemm
vllmellm
committed
7ad868e
quantized tgemm
vllmellm
committed
2dccf5f
Merge remote-tracking branch 'origin/deepseek00-mla' into skinny-gemm
vllmellm
committed
d85cef5
revert unnecessary changes
vllmellm
committed
e037c4e
Merge remote-tracking branch 'origin/deepseek00-mla' into skinny-gemm
vllmellm
committed
b319d46