[None][doc] Update doc for multimodal #7347
Conversation
Signed-off-by: Chang Liu (Enterprise Products) <[email protected]>
📝 WalkthroughWalkthroughRestructures docs/source/reference/multimodal-feature-support-matrix.md from a per-model table to a feature-centric matrix keyed by architecture. Renames model identifiers to architecture-based names, adds eight feature columns (including Overlap Scheduler, CUDA Graph, Chunked Prefill, Torch Sampler, TLLM C++ Sampler, KV Cache Reuse, Logits Post Processor, EPD Disaggregated Serving), and updates per-row values (some N/A). Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Possibly related PRs
Suggested labels
Suggested reviewers
✨ Finishing Touches🧪 Generate unit tests
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. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR/Issue comments)Type Other keywords and placeholders
Status, Documentation and Community
|
![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
🧹 Nitpick comments (4)
docs/source/reference/multimodal-feature-support-matrix.md (4)
3-4: Clarify column heading and expand acronym(s).
- "Model Architecture/Feature" reads ambiguous; use "Model Architecture".
- Expand "EPD" at first mention to avoid confusion.
Apply:
-| Model Architecture/Feature | Overlap Scheduler | CUDA Graph | Chunked Prefill | Torch Sampler | TLLM C++ Sampler | KV Cache Reuse | Logits Post Processor | EPD Disaggregated Serving | +| Model Architecture | Overlap Scheduler | CUDA Graph | Chunked Prefill | Torch Sampler | TLLM C++ Sampler | KV Cache Reuse | Logits Post Processor | Enterprise Platform Deployment (EPD) — Disaggregated Serving |
3-13: Add a legend and scope note to make the matrix self-explanatory.Define Yes/No/N/A and clarify scope (PyTorch backend, TRT-LLM version/date). This prevents misinterpretation downstream.
Proposed addition (place right below the table):
Legend: Yes = supported and tested in CI; No = not supported; N/A = not applicable for the model/architecture. Scope: PyTorch backend only. Last updated: 2025-08-28. TensorRT-LLM version: current main. Notes: - “Torch Sampler” vs “TLLM C++ Sampler” indicate two independent sampling paths; “Yes” in both means both are available. - “Logits Post Processor” refers to post-decoding processors (e.g., repetition penalty) implemented in backend kernels.
3-13: Link features to docs for discoverability.Add links from column headers to their respective docs/sections (Overlap Scheduler, CUDA Graph, Chunked Prefill, KV Cache Reuse, Samplers, Logits Post Processor, EPD).
If you share the preferred anchors, I can generate the exact markdown-link diff.
5-13: Consider sorting/grouping rows for quicker scanning.Alphabetical by architecture or grouped by vendor/family (Llama/LLaVA/Qwen/Gemma/Mistral/Phi) improves usability.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
docs/source/reference/multimodal-feature-support-matrix.md(1 hunks)
⏰ 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)
docs/source/reference/multimodal-feature-support-matrix.md (4)
5-5: Revisit N/A entries for Gemma3.Gemma3ForConditionalGeneration shows N/A for Chunked Prefill and KV Cache Reuse while similar rows are No/Yes elsewhere. Ensure N/A is intentional (truly inapplicable) rather than “No”.
3-13: Nice restructure—feature-centric matrix improves clarity.The expanded feature columns and consistent Yes/No cells make comparisons faster.
5-13: Model identifiers verified; match code registry. No changes required.
5-13: All matrix entries are correct as-is.
- Overlap Scheduler is enabled by default in the PyTorch backend (docs/source/torch/features/overlap_scheduler.md:18) ⇒ “Yes” for every model is expected.
- Disaggregated serving is currently limited to decoder-only engines (advanced/disaggregated-service.md:45) ⇒ “No” for all multimodal (encoder-decoder/VLM) models is correct.
- “N/A” denotes a feature that isn’t applicable to that architecture, while “No” marks an applicable feature that hasn’t been implemented. The Chunked Prefill and KV Cache Reuse columns follow this convention for encoder-decoder vs. causal-LM variants in the matrix.
- KV Cache Reuse support aligns with the paged_context_fmha requirement and release notes (docs/source/release-notes.md:108, 330) and matches the “Yes” models in the table.
Signed-off-by: Chang Liu (Enterprise Products) <[email protected]>
chang-l
changed the title
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (4)
docs/source/reference/multimodal-feature-support-matrix.md (4)
3-4: Clarify the first column header; it reads ambiguous.“Model Architecture/Feature” conflates two concepts. Use a concrete label and keep “Feature” out of that cell.
Apply:
-| Model Architecture/Feature | Overlap Scheduler | CUDA Graph | Chunked Prefill | Torch Sampler | TLLM C++ Sampler | KV Cache Reuse | Logits Post Processor | EPD Disaggregated Serving | +| Model (HF class) | Overlap Scheduler | CUDA Graph | Chunked Prefill | Torch Sampler | TLLM C++ Sampler | KV Cache Reuse | Logits Post Processor | EPD Disaggregated Serving |Add a short legend under the table (after Line 14):
Legend: Yes = supported/validated; No = not supported; N/A = not applicable to this architecture. Scope: PyTorch backend only. Last updated: 2025-08-28.
7-9: Normalize LLaVA/VILA naming for consistency.Mixing “LlavaLlamaModel (VILA)” with “LlavaNextForConditionalGeneration” is inconsistent. Prefer a single convention (either product name with class in parentheses, or class only) across both rows.
Example:
-| LlavaLlamaModel (VILA) | Yes | Yes | No | Yes | Yes | No | Yes | No | +| LLaVA (LlavaLlamaModel) | Yes | Yes | No | Yes | Yes | No | Yes | No | -| LlavaNextForConditionalGeneration | Yes | Yes | No | Yes | Yes | No | Yes | No | +| LLaVA-NeXT (LlavaNextForConditionalGeneration) | Yes | Yes | No | Yes | Yes | No | Yes | No |
3-3: Link feature headers to reference docs (discoverability).Turn feature column headers into links to their reference sections (Overlap Scheduler, CUDA Graph, Chunked Prefill, Torch Sampler, TLLM C++ Sampler, KV Cache Reuse, Logits Post Processor, EPD Disaggregated Serving).
If paths exist (e.g., docs under docs/source/reference/), convert:
-| Overlap Scheduler | CUDA Graph | Chunked Prefill | Torch Sampler | TLLM C++ Sampler | KV Cache Reuse | Logits Post Processor | EPD Disaggregated Serving | +| [Overlap Scheduler](../<path>) | [CUDA Graph](../<path>) | [Chunked Prefill](../<path>) | [Torch Sampler](../<path>) | [TLLM C++ Sampler](../<path>) | [KV Cache Reuse](../<path>) | [Logits Post Processor](../<path>) | [EPD Disaggregated Serving](../<path>) |Replace with the correct anchors.
8-8: Minor whitespace alignment nit.There’s uneven spacing around the “KV Cache Reuse” cell in this row; not harmful, but easy to tidy for consistency.
-| LlavaNextForConditionalGeneration | Yes | Yes | No | Yes | Yes | No | Yes | No | +| LlavaNextForConditionalGeneration | Yes | Yes | No | Yes | Yes | No | Yes | No |
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
docs/source/reference/multimodal-feature-support-matrix.md(1 hunks)
⏰ 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 (2)
docs/source/reference/multimodal-feature-support-matrix.md (2)
5-5: N/A vs No: confirm criteria for Gemma3 entries.Gemma3 shows N/A for Chunked Prefill and KV Cache Reuse while peers show No. If “not applicable” means architecturally impossible vs “not implemented,” add a legend (see prior comment) and verify N/A is correct here.
5-13: Class names verified. All HF-style identifiers in multimodal-feature-support-matrix.md exactly match the canonical class names in the codebase; no updates needed.
chang-l
changed the title
chang-l
changed the title
chang-l
changed the title
Summary by CodeRabbit
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
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.
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.