feat: add trtllm moe_allreduce_fusion#1108
Merged
yzh119 merged 52 commits intoflashinfer-ai:mainfrom Jun 17, 2025
Merged
Conversation
yzh119
reviewed
Jun 2, 2025
Collaborator
yzh119
left a comment
There was a problem hiding this comment.
Please remove all usage of packed/unpacked data type and use vec_t instead.
5 tasks
yyihuang
added a commit
that referenced
this pull request
Jun 9, 2025
<!-- .github/pull_request_template.md --> ## 📌 Description Update the create_ipc_buffer implementation. Add unit tests for create_ipc_buffer. ## 🔍 Related Issues To help debug #1108. ## 🚀 Pull Request Checklist Thank you for contributing to FlashInfer! Before we review your pull request, please make sure the following items are complete. ### ✅ Pre-commit Checks - [x] I have installed `pre-commit` by running `pip install pre-commit` (or used your preferred method). - [x] I have installed the hooks with `pre-commit install`. - [x] I have run the hooks manually with `pre-commit run --all-files` and fixed any reported issues. > If you are unsure about how to set up `pre-commit`, see [the pre-commit documentation](https://pre-commit.com/). ## 🧪 Tests - [x] Tests have been added or updated as needed. - [x] All tests are passing (`unittest`, etc.). ## Reviewer Notes <!-- Optional: anything you'd like reviewers to focus on, concerns, etc. -->
21 tasks
Collaborator
Author
|
Next step: uncomment and complete the fused quantization. Maybe dependent on #1142 |
yzh119
reviewed
Jun 16, 2025
| "hidden_dim * sizeof(T) must be a multiple of kBytesPerAccess"); | ||
| if (params.residual_out && not params.norm_out && params.quant_out) { | ||
| // pattern1: AR+Add_RMS+Quant | ||
| // [m, 7168] bf16 allreduce_in, [m, 7168] bf16 residual_in |
Collaborator
There was a problem hiding this comment.
Do we have shape check somewhere?
| torch.cuda.synchronize() | ||
|
|
||
| # 6. Check correctness | ||
| tolerance = 8e-2 if dtype == torch.float16 else 8e-1 |
Collaborator
There was a problem hiding this comment.
8e-1 seems too large for me, can you give an example about the distribution of all_reduce_out?
yzh119
reviewed
Jun 16, 2025
| // [m, d] bf16 allreduce_in, [m, d] bf16 residual_in | ||
| // [m, d] bf16 residual_out, [m, d] bf16 norm_out, [m, d] fp4 quant_out | ||
|
|
||
| if (params.allreduce_out && params.residual_out && !params.norm_out && params.quant_out) { |
Collaborator
There was a problem hiding this comment.
the remaining part can still be dispatched:
DISPATCH_MOEREDUCTION_KERNEL(T, params, launch_with_pdl, ar, res, rms, quant)
…into trtllm-moear
yzh119
approved these changes
Jun 17, 2025
Collaborator
yzh119
left a comment
There was a problem hiding this comment.
I'm good with the PR, thanks so much for your contribution!
Please refer to
9c229c9 on how to simplify the macro.
Some naming conventions (in flashinfer we usually write both runtime variable and constexpr in the macro definition, to make it easier to developer to track what are the new constexpr introduced in the macro):
#define DISPATCH_*(var, CONST_EXPR)
and we capitalize the CONST_EXPR.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
📌 Description
We try to add moe_all_reduce_fusion kernels from trt-llm.
🔍 Related Issues
We split this PR into multiple ones. #1061
And all_reduce_fusion will be the next.
🚀 Pull Request Checklist
Thank you for contributing to FlashInfer! Before we review your pull request, please make sure the following items are complete.
✅ Pre-commit Checks
pre-commitby runningpip install pre-commit(or used your preferred method).pre-commit install.pre-commit run --all-filesand fixed any reported issues.🧪 Tests
unittest, etc.).Reviewer Notes