pytorch / pytorch/executorch

Move the aten:: namespaced prim ops to portable.

Open
#14,018 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
5k
Forks
1.2k
Avg merge
2d 10h
Merged PRs (30d)
581

Description

🚀 The feature, motivation and pitch

Originally ET didn't have a way to represent non out variant kernels and things like sym_size cannot be expressed as out variant since the return is a prim and aten codegen wont pass in a mutable reference to a prim (probably because you cant do this in python). ET now has support for non out variant kernels (inplace ops and all_close) so we could probably handle these scalar return ops natively.

Alternatives

No response

Additional context

No response

RFC (Optional)

No response

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 by tracing ExecuTorch's existing support for non-out variant kernels and how aten codegen handles scalar-return prim ops such as sym_size. Identify which aten:: namespaced prim ops should move to portable, then verify that they are represented natively without relying on out variants.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
embedded-iot, machine-learning
Issue type
Feature
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.