Skip to content

Conversation

@guan404ming
Copy link
Member

@guan404ming guan404ming commented Nov 12, 2025

Related Issue

#18401

Why

  • When run_ep_decomposition=True is enabled, PyTorch decomposes high-level operators like conv2d.default into
    lower-level convolution.default, which is not supported yet.

How

  • Added _convolution() method in base_fx_graph_translator.py to handle the generic aten.convolution.default operator
  • Updated test_keep_params to use run_ep_decomposition=True

@guan404ming guan404ming force-pushed the add-decomposed-support-for-conv branch from 6324760 to 83c1772 Compare November 12, 2025 03:52
@guan404ming guan404ming marked this pull request as ready for review November 12, 2025 05:40
@guan404ming
Copy link
Member Author

guan404ming commented Nov 12, 2025

cc @tlopex @mshr-h

Copy link
Member

@tlopex tlopex left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Thanks!

@tlopex tlopex merged commit ce5f287 into apache:main Nov 12, 2025
15 checks passed
@guan404ming guan404ming changed the title [Relax][PyTorch] Add support for decomposed operators and fix IR of ops tests [Relax][PyTorch] Add decomposed operator support for Conv Nov 14, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants