Lightning-AI / Lightning-AI/lightning-thunder

enable using python_callable without mapping symbols to their impls

Open
#1,168 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

Currently, we can't use python_callable if we want to retrace because it does half of transform for execution.
We should absolutely get a flag on python_callable for that because we have a workaround in Thunder quite often now (in thunder/core/transforms.py but also in the autograd.Function handling in jit_ext now).
(the ["output"] might not make this example trivial, but still).

https://github.com/Lightning-AI/lightning-thunder/blob/ea96657689ab3c60a41897fcd0be5d00e685a449/thunder/core/transforms.py#L1438-L1445

cc @apaz-cli

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 with the referenced section of thunder/core/transforms.py and inspect the autograd.Function handling in jit_ext. Determine how python_callable currently performs transformation and what flag would preserve retracing without mapping symbols to implementations; done means the existing Thunder workaround is no longer needed and the behavior is covered by the relevant tests.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
compilers
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.