refactor: Enhanced handling of decoder requests and logits within the batch manager

[Funatiq]

Description

• Filter requests and store logits in HandleContextLogits and HandleGenerationLogits.
• Refactored DecoderInputBuffers to hold decoder requests and logits in request order.
• Updated GuidedDecoder and LogitsPostProcessor to utilize DecoderInputBuffers directly.
• Adjusted related tests and bindings to reflect changes in the DecoderInputBuffers interface.

This change streamlines the interaction between components and improves clarity in managing decoder inputs and outputs.

Test Coverage

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.

run [--disable-fail-fast --skip-test --stage-list "A10-1, xxx" --gpu-type "A30, H100_PCIe" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-[Post-Merge]-1, xxx"]

Launch build/test pipelines. All previously running jobs will be killed.

--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 "A10-1, xxx" (OPTIONAL) : Only run the specified test stages. Examples: "A10-1, 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.

--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. Will also run 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-[Post-Merge]-1, xxx" (OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-[Post-Merge]-1, xxx".

For guidance on mapping tests to stage names, see docs/source/reference/ci-overview.md.

kill

kill

Kill all running builds associated with pull request.

skip

skip --comment COMMENT

Skip 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-pipeline

Reuse 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.

[Funatiq]

/bot run

[tensorrt-cicd]

PR_Github #11929 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #11929 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #8847 completed with status: 'FAILURE'

[Funatiq]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #11953 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #11953 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #8870 completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.

[Copilot]

Pull Request Overview

This PR refactors the decoder request handling and logits management within the batch manager to improve component interaction clarity. The changes centralize logits storage and request filtering within DecoderInputBuffers, moving away from a slot-based approach to a direct request-order approach.

• Filter requests and store logits in HandleContextLogits and HandleGenerationLogits functions
• Refactor DecoderInputBuffers to hold decoder requests and logits in request order rather than by sequence slots
• Update GuidedDecoder and LogitsPostProcessor to utilize DecoderInputBuffers directly instead of separate request vectors

Reviewed Changes

Copilot reviewed 16 out of 16 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
tensorrt_llm/_torch/pyexecutor/sampler.py	Remove maxNumSequences parameter from DecoderInputBuffers constructor
cpp/tests/runtime/gptDecoderBatchedTest.cpp	Update DecoderInputBuffers constructor calls in tests
cpp/tests/batch_manager/guidedDecoderTest.cpp	Update test to use new DecoderInputBuffers interface and fix batch index references
cpp/tensorrt_llm/pybind/batch_manager/bindings.cpp	Update Python bindings for DecoderInputBuffers constructor and add decoderRequests field
cpp/tensorrt_llm/pybind/batch_manager/algorithms.cpp	Update algorithm function signatures to use DecoderInputBuffers instead of separate request vectors
cpp/tensorrt_llm/batch_manager/trtGptModelInflightBatching.cpp	Update to use new DecoderInputBuffers interface and pass buffers directly to components
cpp/tensorrt_llm/batch_manager/makeDecodingBatchInputOutput.cpp	Refactor to process requests from DecoderInputBuffers and improve logits handling logic
cpp/tensorrt_llm/batch_manager/logitsPostProcessor.cpp	Update to work with DecoderInputBuffers and use stream instead of runtime
cpp/tensorrt_llm/batch_manager/handleGenerationLogits.cpp	Add logic to populate decoderRequests and logits vectors in DecoderInputBuffers
cpp/tensorrt_llm/batch_manager/handleContextLogits.cpp	Add logic to populate decoderRequests and logits vectors for context requests
cpp/tensorrt_llm/batch_manager/guidedDecoder.cpp	Update to use DecoderInputBuffers interface and process requests in order
cpp/tensorrt_llm/batch_manager/decoderBuffers.cpp	Remove maxNumSequences parameter and logits vector initialization
cpp/include/tensorrt_llm/batch_manager/makeDecodingBatchInputOutput.h	Update function signature to use DecoderInputBuffers
cpp/include/tensorrt_llm/batch_manager/logitsPostProcessor.h	Update interface to use DecoderInputBuffers and stream pointer
cpp/include/tensorrt_llm/batch_manager/guidedDecoder.h	Update execute method to use DecoderInputBuffers
cpp/include/tensorrt_llm/batch_manager/decoderBuffers.h	Add decoderRequests field and update constructor signature

[DomBrown]

LGTM but you probably want to talk to @Linda-Stadter about the bindings so we can coordinate with her PR #6149 and #5961

[Linda-Stadter]

LGTM but you probably want to talk to @Linda-Stadter about the bindings so we can coordinate with her PR #6149 and #5961

You can merge it now and I can include the changes