NVIDIA / NVIDIA/TensorRT-LLM

[Feature]: support for pipeline parallelism under speculative decoding in PyTorch workflow

Open
#9,009 2 comments 0 reactions 1 assignee View on GitHub

@laikhtewari is already working on this.

Since Nov 7, 2025.

feature request Pytorch Scale-out Speculative Decoding
Dominant language
Python
Stars
14.7k
Forks
2.8k
Avg merge
2d 23h
Merged PRs (30d)
489

Description

🚀 The feature, motivation and pitch

As of TRT-LLM 1.2.0rc2, attempting to use speculative decoding in the PyTorch workflow (in this case, N-gram speculative decoding) with a pipeline parallel size > 1 results in the following error at initialization time:

  File "/virtualenv/lib/python3.12/site-packages/tensorrt_llm/_torch/pyexecutor/py_executor.py", line 293, in __init__
    raise NotImplementedError(
    ^^^^^^^^^^^^^^^^^
NotImplementedError: Drafting is not supported for selected executor loop. Please disable disagg/pipeline parallelism scheduler.

Which implies that support for this configuration is simply not present. It would be great if this could be fixed, so as to enable use of speculative decoding with pipeline parallelism.

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

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.