Skip to content

[JAX] Add an MoE Block (Layer) that compound router, permutation, groupedGEMM and communication#2912

Draft
tdophung wants to merge 2 commits intoNVIDIA:mainfrom
tdophung:teddy/moe_block
Draft

[JAX] Add an MoE Block (Layer) that compound router, permutation, groupedGEMM and communication#2912
tdophung wants to merge 2 commits intoNVIDIA:mainfrom
tdophung:teddy/moe_block

Conversation

@tdophung
Copy link
Copy Markdown
Collaborator

Description

Most of MoE building blocks integration work has been deeply coupled with Maxtext development. Now creating this MoE block to isolate the work from Maxtext and provide more room for experimentation with different sequence order of computation and communication, as well as different sharding rules and data layout storage.

This first iteration will start with ring-of-experts EP, sharding on batch dimention for FSDP, CUBLASLt groupedGEMM,
option to choose between Triton kernels or pure JAX permutation implementation, and fused router kernel.

Fixes # (issue)

Type of change

  • Documentation change (change only to the documentation, either a fix or a new content)
  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to not work as expected)
  • Infra/Build change
  • Code refactoring

Changes

  • Create new MoELayer
  • Add test for MoELayer

Checklist:

  • I have read and followed the contributing guidelines
  • The functionality is complete
  • I have commented my code, particularly in hard-to-understand areas
  • I have made corresponding changes to the documentation
  • My changes generate no new warnings
  • I have added tests that prove my fix is effective or that my feature works
  • New and existing unit tests pass locally with my changes

…e path

Signed-off-by: tdophung <tdophung@nvidia.com>
Signed-off-by: tdophung <tdophung@nvidia.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant