Spec Dec: Fix Overlap Scheduler Garbage Output#168
Draft
govind-ramnarayan wants to merge 9 commits into
Draft
Conversation
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-load-one-model
branch
from
February 2, 2026 18:07
3a48be2 to
1223bd2
Compare
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-enable-overlap
branch
from
February 2, 2026 18:24
02cc308 to
b8a1e4b
Compare
lucaslie
reviewed
Feb 2, 2026
There was a problem hiding this comment.
please make sure that any potential changes from NVIDIA#11199 are reflected in this PR before you leave
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-load-one-model
branch
3 times, most recently
from
February 6, 2026 04:58
faf218a to
bb76da8
Compare
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-enable-overlap
branch
from
February 6, 2026 11:30
b8a1e4b to
6db75be
Compare
| task.evaluate(llm, sampling_params=sampling_params) | ||
| task.evaluate(llm, | ||
| sampling_params=sampling_params, | ||
| extra_evaluator_kwargs={'dataset_path': None}) |
Author
There was a problem hiding this comment.
TODO: Get rid of this when ready to merge to CI, I believe this downloads from HF.
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-load-one-model
branch
from
February 6, 2026 12:05
bb76da8 to
816b53d
Compare
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-enable-overlap
branch
from
February 6, 2026 12:05
6db75be to
0bee833
Compare
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-load-one-model
branch
from
February 7, 2026 00:54
816b53d to
36b543c
Compare
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
Signed-off-by: Govind Ramnarayan <105831528+govind-ramnarayan@users.noreply.github.com>
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-enable-overlap
branch
from
February 7, 2026 01:01
0bee833 to
9605a12
Compare
govind-ramnarayan
force-pushed
the
gramnarayan/eagle-load-one-model
branch
3 times, most recently
from
February 10, 2026 08:11
7ffb25b to
cf407df
Compare
1 task
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
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.
fixes: NVIDIA#11175
_prepare_inputs()to readnum_acceptedfrom GPU and modify KV cache and position pointers accordingly.However, it uses a GPU-CPU sync to do this currently, with a TODO describing what needs to happen to make it GPU-only.
Fixes existing bug with one-model code that does not properly set cached pages before running EagleWrapper.forward() - cached pages included pages that were allocated but not yet used. This is fixed by separating "allocated pages" from "used pages" during
_prepare_inputs().Fixes bug where
cache_locwas not properly modified as we increased pages used, leading to sequences overwriting each other's pages while speculating. This is fixed by introducing an "allocated pages" tensor, and reading in slices of it into the "used pages" tensor whenever a sequence grows (or shrinks) to need a different number of pages.Refactors one model execution code to use the same function for metadata update that we use on the
cache_seq_infoto fix the overlap scheduler. Now there is just one utility with metadata updates that we need to make overlap scheduler / CUDA graph compatible.Testing Plan:
increment_position_ids(), both with a positive increment (fast-forward) and negative increment (rewind).disable_overlap_scheduler=Truetodisable_overlap_scheduler=False. After fixing bugs to make this pass, spec dec acceptance rate greatly improved. E.g., before it was ~10% even with the overlap scheduler off, and now with overlap scheduler on:Example of fix on simple prompt:
Prompts:
prompts = [ "What is the capital of France?", "Please explain the concept of gravity in simple words and a single sentence.", ]Before:
Request 1 output:
Request 2 output:
After:
Request 1 output:
Request 2 output:
@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
/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.