Add combinatorial tests to the CI #2991
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
abidwael
changed the title
|
The remaining piece is to add the marker |
justinxzhao
reviewed
Mar 21, 2023
| @@ -188,6 +188,43 @@ jobs: | |||
| name: Unit Test Results (Python ${{ matrix.python-version }} ${{ matrix.test-markers }}) | |||
There was a problem hiding this comment.
In the Unit Tests, Integration Tests, and Regression Tests above, do we need to extend the not slow or benchmark clause to include the combinatorial marker?
There was a problem hiding this comment.
The combinatorial tests will run in their own separate runner and the other jobs can stay intact. I opted for this option as to not add to the current time it takes to run these tests and speed up development.
justinxzhao
reviewed
Mar 22, 2023
There was a problem hiding this comment.
@abidwael, my comments are mainly around code structure, variable naming, nits, and questions. Solid work overall.
Higher-level comments:
- The code in
explore_schemais pretty juicy. It would be great to add unit tests for this code. - In terms of directory structure, I might suggest putting most of the code in a new directory
ludwig/config_samplingand putting tests intests/ludwig/config_sampling. - Tests that actually do config sampling and run training for 1 step can live in
tests/robustness(ortraining_successas you currently have it, thoughrobustnessseems more clear to me) - Move
explore_<atomic type>()functions to their own file.explore_schema.pyis really big already. - Throughout your code, you use
keyanditem, which aren't particularly descriptive. In one of my comments, I suggest usingjsonschema_propertyinstead ofitemandparameter_name/pathinstead ofkey, which I would take back ifkeyanditemare formal concepts in jsonschema. - Prefer to use f"{var}.{var}" strings instead of
var + "." + var - nit: Consider defining a dataclass for the
dqinstead of passing around tuple objects.
abidwael
force-pushed
the
training-success-tests
branch
from
4f1cc0b
to
e518eab
Compare
justinxzhao
reviewed
Apr 3, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds combinatorial Ludwig config testing to the CI to automatically generate valid configs and train models using them to test model training success.
TODO: add a README to explain how it works and how to add more combinatorial tests.