[TRTLLM-8436][fix] restore list[list[list[int]]] in add_token #8502
Conversation
Signed-off-by: ixlmar <[email protected]>
|
/bot run |
|
PR_Github #21905 [ run ] triggered by Bot. Commit: |
📝 WalkthroughWalkthroughThe changes refactor token handling in the sampler system to use Python nested lists instead of PyTorch tensors for token parameters. Method signatures across add_token and draft token processing methods are updated, with conversion logic introduced to transition from tensor formats to list formats at ingestion points. Test files are updated to match the new calling conventions. Changes
Sequence DiagramsequenceDiagram
participant Host
participant Sampler as PyExecutor Sampler
participant Token as Token Processor
participant Draft as Draft Handler
Note over Host,Sampler: Old flow (Tensor-based)
Host->>Sampler: new_tokens: torch.Tensor
Sampler->>Token: add_token(Tensor)
Note over Host,Sampler: New flow (List-based)
Host->>Sampler: new_tokens: torch.Tensor
Sampler->>Sampler: new_tokens_list = new_tokens.tolist()
Sampler->>Token: add_token(list[list[list[int]]])
Note over Token: Extract tokens via direct indexing
Token->>Draft: Token reference
alt Drafting Strategy
Sampler->>Draft: process_draft_tokens(tensor + list)
rect rgb(200, 220, 240)
Note over Draft: Greedy/Tree/Rejection-Sampling
Draft->>Draft: Use list form for token access
Draft->>Draft: Use tensor form for batch operations
end
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Pre-merge checks and finishing touches❌ Failed checks (2 warnings)
✅ Passed checks (1 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tensorrt_llm/_torch/speculative/mtp.py (1)
243-250: Python 3.8 compatibility: builtin generics in type hints
list[list[int]]requires Python 3.9+ unlessfrom __future__ import annotationsis enabled. Our guidelines target Python 3.8+. Usetyping.List(or add the future import) to avoid runtime issues on 3.8.Apply one of these fixes:
+from __future__ import annotations from dataclasses import dataclassOr change the annotation:
-def _request_common_handling(self, request: LlmRequest, next_draft_tokens: list[list[int]]): +from typing import List +def _request_common_handling(self, request: LlmRequest, next_draft_tokens: List[List[int]]):As per coding guidelines.
🧹 Nitpick comments (1)
tensorrt_llm/_torch/pyexecutor/sampler.py (1)
942-951: Usepy_seq_slotconsistentlyThese assignments index
new_tokens_tensorwithrequest.seq_slot. Elsewhere we userequest.py_seq_slotfor Python-side bookkeeping. Recommend unifying topy_seq_slotto avoid surprises if the underlying binding’s field diverges.Apply:
- new_tokens_tensor[i, request.seq_slot, self.BEAM] = new_token + new_tokens_tensor[i, request.py_seq_slot, self.BEAM] = new_token ... - new_tokens_tensor[num_accepted, request.seq_slot, self.BEAM] = new_token + new_tokens_tensor[num_accepted, request.py_seq_slot, self.BEAM] = new_token
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
tensorrt_llm/_torch/pyexecutor/sampler.py(13 hunks)tensorrt_llm/_torch/speculative/mtp.py(1 hunks)tests/unittest/_torch/speculative/test_draft_token_tree_verification.py(1 hunks)tests/unittest/_torch/speculative/test_torch_rejection_sampling.py(2 hunks)
🧰 Additional context used
📓 Path-based instructions (3)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Use only spaces, no tabs; indent with 4 spaces.
Files:
tests/unittest/_torch/speculative/test_torch_rejection_sampling.pytensorrt_llm/_torch/speculative/mtp.pytests/unittest/_torch/speculative/test_draft_token_tree_verification.pytensorrt_llm/_torch/pyexecutor/sampler.py
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+.
Indent Python code with 4 spaces; do not use tabs.
Maintain module namespace when importing; prefer 'from package.subpackage import foo' then 'foo.SomeClass()' instead of importing the class directly.
Python filenames should be snake_case (e.g., some_file.py).
Python classes use PascalCase names.
Functions and methods use snake_case names.
Local variables use snake_case; prefix 'k' for variables that start with a number (e.g., k_99th_percentile).
Global variables use upper SNAKE_CASE prefixed with 'G' (e.g., G_MY_GLOBAL).
Constants use upper SNAKE_CASE (e.g., MY_CONSTANT).
Avoid shadowing variables from an outer scope.
Initialize all externally visible members of a class in the constructor.
Prefer docstrings for interfaces that may be used outside a file; comments for in-function or file-local interfaces.
Use Google-style docstrings for classes and functions (Sphinx-parsable).
Document attributes and variables inline so they render under the class/function docstring.
Avoid reflection when a simpler, explicit approach suffices (e.g., avoid dict(**locals()) patterns).
In try/except, catch the most specific exceptions possible.
For duck-typing try/except, keep the try body minimal and use else for the main logic.
Files:
tests/unittest/_torch/speculative/test_torch_rejection_sampling.pytensorrt_llm/_torch/speculative/mtp.pytests/unittest/_torch/speculative/test_draft_token_tree_verification.pytensorrt_llm/_torch/pyexecutor/sampler.py
**/*.{cpp,cxx,cc,h,hpp,hh,hxx,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Prepend the NVIDIA Apache-2.0 copyright header with current year to the top of all source files (e.g., .cpp, .h, .cu, .py).
Files:
tests/unittest/_torch/speculative/test_torch_rejection_sampling.pytensorrt_llm/_torch/speculative/mtp.pytests/unittest/_torch/speculative/test_draft_token_tree_verification.pytensorrt_llm/_torch/pyexecutor/sampler.py
🧬 Code graph analysis (2)
tests/unittest/_torch/speculative/test_draft_token_tree_verification.py (1)
tensorrt_llm/_torch/pyexecutor/sampler.py (1)
_process_draft_tokens_tree(725-822)
tensorrt_llm/_torch/pyexecutor/sampler.py (2)
tensorrt_llm/_torch/pyexecutor/llm_request.py (1)
LlmRequest(422-626)tensorrt_llm/_torch/pyexecutor/sampling_utils.py (1)
sample_rejected(236-248)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Pre-commit Check
🔇 Additional comments (4)
tests/unittest/_torch/speculative/test_torch_rejection_sampling.py (1)
28-33: Typing-only cast is fine here
cast(int, torch.multinomial(...).item())is a no-op at runtime but makes the type explicit for linters. The surrounding logic remains unchanged. LGTM.tensorrt_llm/_torch/speculative/mtp.py (1)
259-261: Converting host tensor to nested Python listsSwitching to
tolist()aligns with the list-basedadd_tokenand avoids per-access tensor overhead. Just note this allocates a nested structure of size [steps][seq_slots][beam]; keep an eye on peak memory for large batches.Please confirm
state.host.new_tokenshas shape (max_draft_len+1, max_num_requests, 1) soadd_token(..., step=i)indexes correctly across both context and generation paths.tests/unittest/_torch/speculative/test_draft_token_tree_verification.py (1)
48-53: Test updated to new API and list pathPassing both
new_tokens_tensorandnew_tokens_listmatches the updated_process_draft_tokens_treesignature. Converting with.tolist()ensures the list index path is exercised. LGTM.tensorrt_llm/_torch/pyexecutor/sampler.py (1)
1939-1975: Verify decoder_state tensor shape in TRTLLM implementationThe indexing pattern in
add_token(line 299:new_tokens[step][seq_slot][beam]) expects shape[steps][sequences][beams], butstate.host.new_tokensoriginates fromself.store["decoder_state"].all_new_tokens, which is populated by the C++/CUDA backend. The shape of this tensor cannot be verified from the Python layer alone.Review concern is valid: if the decoder populates the tensor with sequences multiplexed by beam (shape
[steps][sequences*beam_width]) rather than separate dimensions, the indexing will misalign. Confirm the actual shape ofdecoder_state.all_new_tokensin the TRTLLM decoder implementation and verify it matches the indexing expectations inadd_token.
ixlmar
changed the title
yweng0828
left a comment
yweng0828
left a comment
There was a problem hiding this comment.
Just curious, is there any difference in performance between accessing data using lists and tensors?
|
/bot run --add-multi-gpu-test |
|
PR_Github #21915 [ run ] triggered by Bot. Commit: |
|
PR_Github #21905 [ run ] completed with state |
Yes, this had been noticed in #7730. I could imagine that it has to do with C++ binding overheads in |
|
PR_Github #21915 [ run ] completed with state |
Signed-off-by: ixlmar <[email protected]>
Signed-off-by: ixlmar <[email protected]> Signed-off-by: yufeiwu-nv <[email protected]>
Signed-off-by: ixlmar <[email protected]>
Signed-off-by: ixlmar <[email protected]>
Signed-off-by: ixlmar <[email protected]>
Signed-off-by: ixlmar <[email protected]>
4 participants
Description
Restore performance optimization introduced in #7730
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
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
/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...Provide a user friendly way for developers to interact with a Jenkins server.
Run
/bot [-h|--help]to print this help message.See details below for each supported subcommand.
Details
run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]Launch build/test pipelines. All previously running jobs will be killed.
--reuse-test (optional)pipeline-id(OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.--disable-reuse-test(OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.--disable-fail-fast(OPTIONAL) : Disable fail fast on build/tests/infra failures.--skip-test(OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.--stage-list "A10-PyTorch-1, xxx"(OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.--gpu-type "A30, H100_PCIe"(OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.--test-backend "pytorch, cpp"(OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.--only-multi-gpu-test(OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.--disable-multi-gpu-test(OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.--add-multi-gpu-test(OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.--post-merge(OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx"(OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".--detailed-log(OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.--debug(OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in thestage-listparameter to access the appropriate container environment. Note: Does NOT update GitHub check status.For guidance on mapping tests to stage names, see
docs/source/reference/ci-overview.mdand the
scripts/test_to_stage_mapping.pyhelper.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip testing for latest commit on pull request.
--comment "Reason for skipping build/test"is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.reuse-pipeline
reuse-pipelineReuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.
Summary by CodeRabbit