deepspeedai / deepspeedai/DeepSpeed

where is tensor parallel implemented during inference?

Open
#3,374 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
43.1k
Forks
5k
Avg merge
4d 15h
Merged PRs (30d)
112

Description

Hi, just curious to know where the tensor parallel is implemented during inference.
In the inference/engine.py file, when replace_with_kernel_inject is set to False, does that mean TP will not be used during inference? And from my understanding, all_reduce is frequently called for TP, can you show me which part does this?
Thanks

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.

Research direction

Start with inference/engine.py and trace the inference path when replace_with_kernel_inject is false. Search the inference implementation for tensor-parallel and all_reduce references, then document the relevant files and call sites. Done means a concise answer explains whether tensor parallel is used and where its collectives occur.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
distributed-systems, machine-learning
Issue type
Documentation
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.