Added support for quantizing TEGroupedMLP for megatron-lm#403

Merged

kinjalpatel27 merged 28 commits intomainfrom

kinjal/grouped_linear

Oct 17, 2025

Commits on Oct 11, 2025

sync amax in context parallel and awq act scale

jenchen13
authored and
kinjalpatel27
committed
lint

jenchen13
authored and
kinjalpatel27
committed
test weight quantizer too

jenchen13
authored and
kinjalpatel27
committed
fix test

jenchen13
authored and
kinjalpatel27
committed
awq test

jenchen13
authored and
kinjalpatel27
committed
move awq test inside megatron tests

jenchen13
authored and
kinjalpatel27
committed
fix amax tests

jenchen13
authored and
kinjalpatel27
committed
fix awq lite param

jenchen13
authored and
kinjalpatel27
committed
fix test

jenchen13
authored and
kinjalpatel27
committed
add print

jenchen13
authored and
kinjalpatel27
committed
docstring

jenchen13
authored and
kinjalpatel27
committed
fix tests

jenchen13
authored and
kinjalpatel27
committed
fix multiprocess size

jenchen13
authored and
kinjalpatel27
committed
Added quantization support for TEGroupedMoE for megatron-lm
kinjalpatel27
committed
code cleanup
kinjalpatel27
committed
code and test cleanup
kinjalpatel27
committed
Updated moe names in tests
kinjalpatel27
committed
updated parallel state for experts
kinjalpatel27
committed
fixed bug for is_quantized_linear check
kinjalpatel27
committed
code cleanup and bug fixes
kinjalpatel27
committed
rebase bug fixes
kinjalpatel27
committed
fixing test and comments
kinjalpatel27
committed

Commits on Oct 13, 2025

Code cleanup
kinjalpatel27
committed

Commits on Oct 14, 2025

Commits on Oct 16, 2025

Commits on Oct 17, 2025

Addressed MR comments
kinjalpatel27
committed