Skip to content

Run whole test suite on numba backend #811

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Draft
wants to merge 12 commits into
base: main
Choose a base branch
from
Draft

Run whole test suite on numba backend #811

wants to merge 12 commits into from

Conversation

ricardoV94
Copy link
Member

@ricardoV94 ricardoV94 commented Jun 7, 2024 •
edited
Loading

Description

What is broken / not supported by Numba:

Related Issue

  • Closes #
  • Related to #

Checklist

Type of change

  • New feature / enhancement
  • Bug fix
  • Documentation
  • Maintenance
  • Other (please specify):

@ricardoV94 ricardoV94 force-pushed the test_on_numba branch 2 times, most recently from 2eb7fe7 to 1f093cd Compare June 7, 2024 17:24
@ricardoV94
Copy link
Member Author

Locally I ran all the tests in tests/tensor/rewriting/test_basic.py twice (to allow caching) before and after making Numba default.
Before: 6s
After: 1m34s

Numba caching is baaaaaaaaaaaaaaaaaaaaaaaaaaaaaaad

@ricardoV94
Copy link
Member Author

ricardoV94 commented Jun 7, 2024 •
edited
Loading

Some fun looking kernel crashes, such as in tests/compile/function/test_types.py::TestFunction::test_copy_share_memory or tests/tensor/rewriting/test_elemwise.py::TestFusion::test_elemwise_fusion[case72]

Both seem to be related to Elemwise (CC @aseyboldt)

@ricardoV94 ricardoV94 changed the title Test on numba Run whole test suite on numba backend Jun 7, 2024
@aseyboldt
Copy link
Member

Found a small reproducer for one of the aborts:

x = dvector("x")
out = 1 / x**2
graph = FunctionGraph(outputs=[out])
func = numba_funcify(graph)

It works if I run it through the fast_run rewrites, but fails without them.
I'll try to figure out more tomorrow.

@ricardoV94
Copy link
Member Author

Didn't we find an issue with power specifically recently?

@ricardoV94
Copy link
Member Author

numba/numba#9554

@ricardoV94
Copy link
Member Author

Okay, I added a patch for the power crash, disabling fastmath if the power is an integer for now

@ricardoV94 ricardoV94 force-pushed the test_on_numba branch 4 times, most recently from 6f6e8bb to 53d9a26 Compare June 13, 2024 14:43
@ricardoV94 ricardoV94 added this to the 3.0 release milestone Feb 14, 2025
@ricardoV94
Copy link
Member Author

After caching the test file now runs in 50s after caching vs 6s before the PR, so only 8x slower now :(

@ricardoV94 ricardoV94 mentioned this pull request Mar 28, 2025
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
PyTensor hackathon issues
Status: No status
Milestone
3.0 release
Development

Successfully merging this pull request may close these issues.
2 participants
@ricardoV94 @aseyboldt