[TRTLLM-10382][feat] Support rejection sampling in one-model spec dec#11001
[TRTLLM-10382][feat] Support rejection sampling in one-model spec dec#11001ziyixiong-nv wants to merge 1 commit intoNVIDIA:mainfrom
Conversation
ziyixiong-nv
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
e19ea8b to
41a04a5
Compare
|
/bot run |
📝 WalkthroughWalkthroughThis PR introduces rejection sampling support for draft token selection in speculative decoding. It adds infrastructure for capturing draft logits, computing draft probabilities, and routing sampling through a rejection-sampling-aware path when enabled across Eagle3 and MTP models. Changes
Sequence Diagram(s)sequenceDiagram
actor Forward as Forward Pass (Eagle3/MTP)
participant Draft as Draft Model
participant LogitCapture as Logit Capture
participant ProbCompute as Probability Computation
participant Sampling as Rejection Sampling
participant Output as Output Tokens
Forward->>Draft: Generate draft tokens per step
Draft-->>LogitCapture: Emit draft logits
LogitCapture->>LogitCapture: Accumulate in draft_logits_list<br/>(when use_rejection_sampling=True)
Forward->>ProbCompute: Call _compute_and_store_draft_probs<br/>(after draft loop)
ProbCompute->>ProbCompute: Apply temperature & top-k/top-p<br/>Compute softmax → draft_probs
ProbCompute->>ProbCompute: Store in spec_metadata.draft_probs
Forward->>Sampling: sample_and_accept_draft_tokens
alt use_rejection_sampling enabled
Sampling->>Sampling: Reshape draft_probs to<br/>[batch_size, max_draft_len, vocab_size]
Sampling->>Sampling: _sample_and_accept_draft_tokens_rejection<br/>(target_probs vs draft_probs)
Sampling->>Sampling: Perform probabilistic acceptance
else fallback to base path
Sampling->>Sampling: Existing sampling logic
end
Sampling-->>Output: Accepted tokens & counts
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Suggested reviewers
🚥 Pre-merge checks | ✅ 1 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (1 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Tip Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs). 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: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tensorrt_llm/llmapi/llm_args.py (1)
689-696: Enforceallow_advanced_samplingwhen rejection sampling is enabled.The rejection-sampling path expects per-token temperatures/top-k/top-p tensors; without
allow_advanced_sampling=True, those stay unset and later slicing will fail. Add a validator to fail fast when the config is invalid.✅ Proposed fix
class DecodingBaseConfig(StrictBaseModel): @@ # Prototype. If true, allows non-greedy sampling when speculation is used. Only applicable # to 1-model code paths; non-greedy sampling is always enabled on 2-model paths. allow_advanced_sampling: bool = False # If true, uses rejection sampling for draft token acceptance instead of strict token equality. # Rejection sampling provides lossless acceleration that exactly matches the target model's # distribution. Requires allow_advanced_sampling=True. Only applicable to 1-model code paths. use_rejection_sampling: bool = False + + `@model_validator`(mode="after") + def _validate_rejection_sampling(self): + if self.use_rejection_sampling and not self.allow_advanced_sampling: + raise ValueError( + "use_rejection_sampling requires allow_advanced_sampling=True") + return self
|
PR_Github #33574 [ run ] triggered by Bot. Commit: |
|
PR_Github #33574 [ run ] completed with state
|
|
/bot run |
|
PR_Github #33584 [ run ] triggered by Bot. Commit: |
|
PR_Github #33584 [ run ] completed with state
|
|
/bot run |
|
PR_Github #33608 [ run ] triggered by Bot. Commit: |
|
PR_Github #33608 [ run ] completed with state
|
|
/bot run |
ziyixiong-nv
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
86e7b19 to
5264e8a
Compare
|
/bot run |
|
PR_Github #33646 [ run ] triggered by Bot. Commit: |
|
PR_Github #33647 [ run ] triggered by Bot. Commit: |
|
PR_Github #33647 [ run ] completed with state
|
ziyixiong-nv
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
5264e8a to
e351950
Compare
|
/bot run |
|
PR_Github #33679 [ run ] triggered by Bot. Commit: |
|
PR_Github #33679 [ run ] completed with state
|
ziyixiong-nv
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
e351950 to
37bd7cb
Compare
|
/bot run |
|
PR_Github #33703 [ run ] triggered by Bot. Commit: |
ziyixiong-nv
force-pushed
the
dev-fxiong-rejection-sampling
branch
2 times, most recently
from
b590ec3 to
03682c2
Compare
|
/bot run |
|
PR_Github #37931 [ run ] triggered by Bot. Commit: |
|
PR_Github #37931 [ run ] completed with state
|
|
/bot run |
|
PR_Github #37973 [ run ] triggered by Bot. Commit: |
|
PR_Github #37973 [ run ] completed with state
|
|
/bot run |
|
PR_Github #37983 [ run ] triggered by Bot. Commit: |
|
PR_Github #37983 [ run ] completed with state
|
|
/bot run |
ziyixiong-nv
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
03682c2 to
ca9b7f0
Compare
|
PR_Github #38113 [ run ] triggered by Bot. Commit: |
|
PR_Github #38113 [ run ] completed with state
|
|
/bot run |
ziyixiong-nv
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
ca9b7f0 to
fa740bf
Compare
|
PR_Github #38320 [ run ] triggered by Bot. Commit: |
|
PR_Github #38320 [ run ] completed with state
|
mikeiovine
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
fa740bf to
9aad2c7
Compare
|
/bot run |
|
PR_Github #39305 [ run ] triggered by Bot. Commit: |
|
PR_Github #39305 [ run ] completed with state
|
Signed-off-by: ziyixiong-nv <219238287+ziyixiong-nv@users.noreply.github.com>
mikeiovine
force-pushed
the
dev-fxiong-rejection-sampling
branch
from
9aad2c7 to
0a55552
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #39473 [ run ] triggered by Bot. Commit: |
|
PR_Github #39473 [ run ] completed with state
|
5 participants
Summary by CodeRabbit
New Features
Configuration
use_rejection_samplingoption enables rejection sampling-based draft token acceptance in speculative decoding (requiresallow_advanced_sampling).✏️ Tip: You can customize this high-level summary in your review settings.
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
/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.