Don't fail when gpytorch import fails #913
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
This is mostly relevant for CI, as most users probably don't have gpytorch installed. The problem is that some gpytorch versions raise an AttributeError with some torch versions. Since we try to import gpytorch in utils.py but only catch ImportError, the AttributeError will be raised. However, the presence of an incompatible gpytorch version should not preclude the usage of skorch. In the gpytorch tests themselves, we already check for compatible versions and skip if they are not compatible.
It is in fact possible to import it only locally with minimal performance overhead and thus avoid exposing skorch to possible gpytorch import issues. On top of that, the import can be commented out for now because it is only needed in case multiclass classification with gaussian processes will be supported.
It's been there long enough without changes.
|
@ottonemo I made the changes as discussed. Please take another look. |
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Outdated suggestions cannot be applied.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 is mostly relevant for CI, as most users probably don't have gpytorch installed.
The problem is that some gpytorch versions now raise an
AttributeErrorwith some torch versions. Since we try to import gpytorch inutils.pyand intest_probabilistic.pybut only catchImportError, theAttributeErrorwill be raised. However, the presence of an incompatible gpytorch version should not preclude the usage of skorch.In the gpytorch tests themselves, we already check for compatible versions and skip if they are not compatible. However, to check versions, we first need to successfully import gpytorch, which fails.
Here is an example of a job that currently fails because of this:
https://github.com/skorch-dev/skorch/actions/runs/3421070263/jobs/5696711066
The solution now is to catch
AttributeErrors during gpytorch imports. Forutils.py, we then treat it as if gpytorch were not installed, for the tests we skip. A disadvantage of this indiscriminate skipping is that it could happen that all versions are skipped and we won't notice since CI is green. I don't have a good solution for this.