Pearl#13551
Draft
zhaoyangwang-nvidia wants to merge 2 commits intoNVIDIA:mainfrom
Draft
Conversation
Signed-off-by: ZhaoyangWang <zhaoyangw@nvidia.com>
Implement CPU-initiated libibverbs RDMA transport to offload the draft
model to a separate GPU, replacing local draft inference with a remote
RDMA peer. The target model (TRT-LLM) writes accumulated output tokens
to the draft server via RDMA Write; the draft server returns speculative
tokens via RDMA Write back.
Key changes:
- rdma_draft_offload.py: ibverbs RC QP client with GPUDirect MR registration,
QP state machine (RESET->INIT->RTR->RTS), and per-round request/response
- rdma_draft_protocol.py: fixed-size binary protocol (32B header + 256B tokens,
MAGIC-checked, 4096B total) for target<->draft RDMA buffers
- draft_target.py: RDMA offload path in DraftTargetOneModelWorker.forward(),
output token history accumulation, warmup pre-connection
- llm_args.py: DraftTargetDecodingConfig RDMA fields; allow speculative_model=None
when draft_offload_enabled=True
- model_loader.py: skip draft weight loading when draft_offload_enabled
- modeling_speculative.py: skip draft model instantiation; thread is_warmup
- _util.py: skip separate draft KV cache when draft_offload_enabled
- model_engine.py: pass is_warmup flag through to model forward inputs
- .gitignore: ignore cmake-created symlinks deep_ep/deep_gemm/flash_mla
Signed-off-by: ZhaoyangWang <zhaoyangw@nvidia.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
@coderabbitai summary
Description
Test Coverage
PR Checklist
Please review the following before submitting your PR:
PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
Update tava architecture diagram if there is a significant design change in PR.
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.
GitHub Bot Help
To see a list of available CI bot commands, please comment
/bot help.