[ Misc ] Refactor MoE to isolate Fp8 From Mixtral#5970

Merged

robertgshaw2-redhat merged 48 commits intovllm-project:mainfrom

neuralmagic:refactor-moe

Jul 2, 2024

Commits on Jun 28, 2024

moe refactor
Robert Shaw
committed
end-to-end working!
Robert Shaw
committed
Update linear.py
robertgshaw2-redhat
authored
added moe to fp8
Robert Shaw
committed
added fp8
Robert Shaw
committed

Commits on Jun 30, 2024

Commits on Jul 1, 2024

Commits on Jul 2, 2024

added fused_moe
Robert Shaw
committed
refactor to have separate MoEMethod
Robert Shaw
committed
move fused moe logic into fused_moe directory
Robert Shaw
committed
refactored
Robert Shaw
committed
remove moe
Robert Shaw
committed
remove spurious change
Robert Shaw
committed
cleanup fp8.py
Robert Shaw
committed
Update vllm/model_executor/models/qwen2_moe.py

robertgshaw2-redhat
and
mgoin
authored
revert ray
Robert Shaw
committed
Merge branch 'refactor-moe' of https://github.com/neuralmagic/nm-vllm into refactor-moe
Robert Shaw
committed
added the important file :)
Robert Shaw
committed
format cleanup
Robert Shaw
committed
update Qwen for re-refactor
Robert Shaw
committed
format
Robert Shaw
committed
fix qwen loading
Robert Shaw
committed
format
Robert Shaw
committed
Merge branch 'main' into refactor-moe
Robert Shaw
committed
fix tests
Robert Shaw
committed
Update test_moe.py
robertgshaw2-redhat
authored
format
Robert Shaw
committed
fix experts
Robert Shaw
committed
Merge branch 'main' into refactor-moe
robertgshaw2-redhat
authored