Add CUTLASS fused moe kernels from TensorRT-LLM.#1113
Merged
yzh119 merged 1 commit intoflashinfer-ai:mainfrom Jun 4, 2025
Merged
Add CUTLASS fused moe kernels from TensorRT-LLM.#1113yzh119 merged 1 commit intoflashinfer-ai:mainfrom
yzh119 merged 1 commit intoflashinfer-ai:mainfrom
Conversation
wenscarl
force-pushed
the
trtllm_cutlass_fused_moe
branch
from
bddbbef to
9957425
Compare
wenscarl
force-pushed
the
trtllm_cutlass_fused_moe
branch
from
9957425 to
8d551aa
Compare
yzh119
approved these changes
Jun 4, 2025
Collaborator
yzh119
left a comment
yzh119
left a comment
There was a problem hiding this comment.
Unittests passed on my side, @wenscarl thanks for the huge effort, let's merge this in first!
Currently all of the c++/cude source code are within to csrc directory, where we store pytorch c++ interface and pybind code, we should move kernel definition and framework agnostic interface to include directory as part of the header only library and reuse infrastructure with other components, in future PRs.
Edenzzzz
pushed a commit
to Edenzzzz/flashinfer
that referenced
this pull request
Jun 6, 2025
…-ai#1113) <!-- .github/pull_request_template.md --> ## 📌 This PR added the CUTLASS implementation of fused Mixture of Expert from TensorRT-LLM. <!-- What does this PR do? Briefly describe the changes and why they’re needed. --> ## 🔍 Related Issues Supported data types are: fp32/bf16/fp16/float8_e4m3fn/float8_e2m1 The kernels also support expert/tensor parallellism. This PR also exposes quantization methods for nvfp4. <!-- Link any related issues here --> ## 🧪 Tests - [ ] tests/test_trtllm_cutlass_fused.py - [ ] tests/test_fp4_quantize.py ## Reviewer Notes <!-- Optional: anything you'd like reviewers to focus on, concerns, etc. -->
21 tasks
6 tasks
5 tasks
|
is this only support sm100 ？ |
Anerudhan
pushed a commit
to Anerudhan/flashinfer
that referenced
this pull request
Jun 28, 2025
…-ai#1113) <!-- .github/pull_request_template.md --> ## 📌 This PR added the CUTLASS implementation of fused Mixture of Expert from TensorRT-LLM. <!-- What does this PR do? Briefly describe the changes and why they’re needed. --> ## 🔍 Related Issues Supported data types are: fp32/bf16/fp16/float8_e4m3fn/float8_e2m1 The kernels also support expert/tensor parallellism. This PR also exposes quantization methods for nvfp4. <!-- Link any related issues here --> ## 🧪 Tests - [ ] tests/test_trtllm_cutlass_fused.py - [ ] tests/test_fp4_quantize.py ## Reviewer Notes <!-- Optional: anything you'd like reviewers to focus on, concerns, etc. -->
Collaborator
Hi, it seems nvfp4 is in the UT as well as in source code, thus I would appreciate it if I could know whether this supports nvfp4 as well |
5 tasks
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
4 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.
📌 This PR added the CUTLASS implementation of fused Mixture of Expert from TensorRT-LLM.
🔍 Related Issues
Supported data types are: fp32/bf16/fp16/float8_e4m3fn/float8_e2m1
The kernels also support expert/tensor parallellism.
This PR also exposes quantization methods for nvfp4.
🧪 Tests
Reviewer Notes