[Feature]: [AutoDeploy] Accuracy issue w/ piecewise cudagraph for DeepSeek R1
Open
@laikhtewari is already working on this.
Since Apr 11, 2026.
CUDA Graph
feature request
- Dominant language
- Python
- Stars
- 14.7k
- Forks
- 2.8k
- Avg merge
- 2d 23h
- Merged PRs (30d)
- 489
Description
🚀 The feature, motivation and pitch
How to reproduce :
- Clone https://github.com/NVIDIA/TensorRT-LLM/pull/12946
- In examples/auto_deploy/model_registry/configs/deepseek-r1.yaml, turn on piecewise_enabled: true
pytest tests/integration/defs/accuracy/test_llm_api_autodeploy.py::TestModelRegistryAccuracy::test_autodeploy_from_registry -k "DeepSeek-R1-0528" -vss
W/o piecewise cudagraph (as default) : MMLU 84.60 / Gsm8K 93.17
W/ piecewise cudagraph : MMLU 75.12
Alternatives
No response
Additional context
No response
Before submitting a new issue...
- Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.