Conversation
|
Can AI check variable initialization and add all doc strings that are missing in Aviary? |
I'm not sure what you mean by variable initialization in this case. AI could definitely write docstrings, but I don't trust it to correctly explain function arguments, so it would need to be heavily reviewed |
How about potential divide-by-zero errors and verbosity level? |
|
|
||
|
|
||
| def flops_validation_test( | ||
| test: unittest.TestCase, |
There was a problem hiding this comment.
case_name should probably be removed from here too, since it is not longer used due to the change in arguments in do_validation_test.
There was a problem hiding this comment.
This function actually does use case_name, it is only do_validation_test that doesn't ever use that argument
github-merge-queue
bot
removed this pull request from the merge queue due to a conflict with the base branch
3 participants
Summary
A subset of unittests were updated to use unittest's subTest feature, which allows for multiple failures to occur within a test, which prevents a developer from needing to update a test one assert at a time while debugging.
This work was done with AI (Claude 4.5 opus), and was a test of automating simple but tedious updates across multiple files. Only some files were updated due to limits/cost on repeatedly running an expensive LLM model and limitations on how many files in a row an agentic query could get through before stalling.
Updates to the FLOPS validation test helper functions was done manually.
Related Issues
Backwards incompatibilities
None
New Dependencies
None