update trtllm-gen fp4 autotuner and routing #1573
Merged
Conversation
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
yzh119
reviewed
Aug 26, 2025
| # TODO(siyuan): support fp8 | ||
| moe_op.trtllm_fp4_block_scale_moe( | ||
| routing_logits.to(torch.bfloat16), | ||
| routing_logits, |
Collaborator
There was a problem hiding this comment.
remove .to(torch.bfloat16) which causes illegal memory access when using deepseek v3 routing.
could you explain why?
Collaborator
Author
There was a problem hiding this comment.
@yzh119 Currently DeepSeek V3 requires fp32 routing logits: https://github.com/flashinfer-ai/flashinfer/blob/main/tests/test_trtllm_gen_fused_moe.py#L1987-L1991
IwakuraRein
changed the title
yzh119
approved these changes
Aug 27, 2025
5 tasks
nvpohanh
added a commit
to nvpohanh/vllm
that referenced
this pull request
Sep 4, 2025
Mainly to get the GPT-OSS MXFP4 trtllm-gen MoE autotuning and the bug fix in: flashinfer-ai/flashinfer#1573 Signed-off-by: Po-Han Huang <pohanh@nvidia.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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
.to(torch.bfloat16)which causes illegal memory access when using deepseek v3 routing.routing_logitsisNoneto supporttrtllm_fp4_block_scale_routed_moeinvalid argumenterror in llama and renormalize routing🔍 Related Issues
🚀 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