[TRTLLM-11228][feat] Update quickstart for DFlash#13545
[TRTLLM-11228][feat] Update quickstart for DFlash#13545ziyixiong-nv wants to merge 1 commit intoNVIDIA:mainfrom
Conversation
|
/bot run |
📝 WalkthroughWalkthroughThe changes introduce support for DFlash speculative decoding algorithm through documentation and example code. Documentation describes how DFlash uses target model hidden states as cross-attention context in the draft model. Example code adds configuration support for initializing Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes 🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
examples/llm-api/quickstart_advanced.py (1)
1-1:⚠️ Potential issue | 🟠 MajorAdd the required NVIDIA copyright header to this modified Python file.
This file is modified but currently has no header at the top.
As per coding guidelines
**/*.{h,hpp,cpp,cc,cxx,cu,py}: “All TensorRT-LLM source files must contain an NVIDIA copyright header with the year of latest meaningful modification.”🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@examples/llm-api/quickstart_advanced.py` at line 1, Add the required NVIDIA copyright header at the very top of the modified Python file: insert the standard NVIDIA copyright block including the year of latest meaningful modification and the canonical NVIDIA header text required by project guidelines; ensure it precedes any imports (e.g., before the existing "import argparse") and uses the same formatting as other project source headers so linters and license checks will recognize it.docs/source/features/speculative-decoding.md (1)
223-233:⚠️ Potential issue | 🟡 MinorUpdate backend-support note to avoid contradiction with the new
DFlashoption.Line 230 adds
DFlashas an availabledecoding_type, but Line 233 still says PyTorch supports onlyEagle3. Please align this note with the newly documented option.✏️ Suggested doc fix
-> Note: The PyTorch backend supports only `Eagle3`. `decoding_type: Eagle` is accepted as a backward-compatible alias for `Eagle3`, but EAGLE (v1/v2) draft checkpoints are incompatible. +> Note: The PyTorch backend supports a subset of decoding types, including `Eagle3` and `DFlash`. `decoding_type: Eagle` is accepted as a backward-compatible alias for `Eagle3`, but EAGLE (v1/v2) draft checkpoints are incompatible.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/source/features/speculative-decoding.md` around lines 223 - 233, The backend-support note is now contradictory because `decoding_type: DFlash` was added but the note still claims the PyTorch backend supports only `Eagle3`; update the note to reflect actual PyTorch support by listing which decoding types PyTorch supports (e.g., `Eagle3` and `DFlash` if supported) and keep the backward-compatible alias (`decoding_type: Eagle`) remark; edit the sentence mentioning PyTorch to explicitly enumerate supported decoding types (referencing `decoding_type`, `Eagle3`, `DFlash`, and the `Eagle` alias) so the doc is consistent.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Outside diff comments:
In `@docs/source/features/speculative-decoding.md`:
- Around line 223-233: The backend-support note is now contradictory because
`decoding_type: DFlash` was added but the note still claims the PyTorch backend
supports only `Eagle3`; update the note to reflect actual PyTorch support by
listing which decoding types PyTorch supports (e.g., `Eagle3` and `DFlash` if
supported) and keep the backward-compatible alias (`decoding_type: Eagle`)
remark; edit the sentence mentioning PyTorch to explicitly enumerate supported
decoding types (referencing `decoding_type`, `Eagle3`, `DFlash`, and the `Eagle`
alias) so the doc is consistent.
In `@examples/llm-api/quickstart_advanced.py`:
- Line 1: Add the required NVIDIA copyright header at the very top of the
modified Python file: insert the standard NVIDIA copyright block including the
year of latest meaningful modification and the canonical NVIDIA header text
required by project guidelines; ensure it precedes any imports (e.g., before the
existing "import argparse") and uses the same formatting as other project source
headers so linters and license checks will recognize it.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 0c4d71e6-ff4d-42dd-91b3-b942e4d97076
📒 Files selected for processing (2)
docs/source/features/speculative-decoding.mdexamples/llm-api/quickstart_advanced.py
|
PR_Github #45864 [ run ] triggered by Bot. Commit: |
|
PR_Github #45864 [ run ] completed with state
|
ziyixiong-nv
force-pushed
the
dev-fxiong-dflash-task
branch
from
83ccef0 to
797531f
Compare
|
/bot run |
|
PR_Github #46011 [ run ] triggered by Bot. Commit: |
|
PR_Github #46011 [ run ] completed with state
|
ziyixiong-nv
force-pushed
the
dev-fxiong-dflash-task
branch
from
797531f to
cd9d65e
Compare
|
/bot run |
|
PR_Github #46089 [ run ] triggered by Bot. Commit: |
|
PR_Github #46089 [ run ] completed with state
|
Signed-off-by: ziyixiong-nv <219238287+ziyixiong-nv@users.noreply.github.com>
ziyixiong-nv
force-pushed
the
dev-fxiong-dflash-task
branch
from
cd9d65e to
d452a12
Compare
|
/bot run |
|
PR_Github #46161 [ run ] triggered by Bot. Commit: |
|
PR_Github #46161 [ run ] completed with state
|
|
/bot run |
|
PR_Github #46231 [ run ] triggered by Bot. Commit: |
|
PR_Github #46231 [ run ] completed with state |
2 participants
Summary by CodeRabbit
New Features
Documentation
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
To see a list of available CI bot commands, please comment
/bot help.