Lightning-AI / Lightning-AI/lightning-thunder

[DTensor] Refactor to re-use prims

Open
#2,722 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

DTensor
Dominant language
Python
Stars
1.5k
Forks
121
PR merge metrics
No merged PRs in 30d

Description

Investigate: Try to decrease duplication of prims - reuse the existing prims one by one to support propagating DTensor layout and providing the executors information about the supported types (Tensor, DTensor) which the executor opt in or out of. (See [comment](https://github.com/Lightning-AI/lightning-thunder/pull/2382#discussion_r2272971488))

Contributor guide

No contributing guide indexed for this repository

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 by reading the existing prims and the discussion linked from the issue, especially the referenced Lightning-Thunder pull request. Investigate how DTensor layout propagation and executor support for Tensor and DTensor are currently handled. Done means reducing prim duplication while preserving layout propagation and allowing executors to opt in or out of supported types.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
compilers, distributed-systems
Issue type
Refactor
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.