Merging 0.10.0rc1 back to master #550
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
Changes to detector kwargs and addition of get_config methods.
Addition of pydantic validation of detector configs
Main config driven save/load functionality, testing, and docs.
This reverts commit bd0fa00.
Utility functions to get and set seeds for the Python, Numpy, TensorFlow and PyTorch random number generators.
This PR implements a number of final fixes and refactorings for the config driven detectors: 1. The `preprocess_at_init` and `preprocess_at_pred` logic implemented in #381 and #458 has been reworked. This turned out to have a problem in how it dealt with `update_x_ref`, since regardless of `x_ref_preprocessed`, we still need to update reference data within `.predict()` when `update_x_ref` is set. All offline drift detectors have been reworked to use the old logic (but with `preprocess_x_ref` renamed `preprocess_at_init`), with the addition that `self.x_ref_preprocessed` is also checked internally. 2. The previous `get_config` methods involved a lot of boilerplate to try to recover the original args/kwargs from detector attributes. The new approach calls a generic `_set_config()` with `__init__`, and then `self.config` is returned by `get_config`. This should significantly reduce the workload to add save/load to new detectors. To avoid memory overheads, large artefacts such as `x_ref` are not set at `__init__`, and instead are added within `get_config`. 3. Owing to the ease of implementation with the new `get_config` approach, save/load has been added for the model uncertainty and online detectors! 4. Kernels and `preprocess_fn`'s were previously resolved in `_load_detector_config`, which wasn't consistent with how other artefacts were resolved (it also caused added extra challenges). These are now resolved in `resolve_config` instead. Following this the `KernelConfigResolved` and `PreprocessConfigResolved` pydantic models have been removed (they could be added back but it would complicate `resolve_config`). 5. Fixing determinism in #496 has allowed us to compare original and loaded detector predictions in `test_saving.py`. This uncovered bugs with how kernels were saved and loaded. These have been fixed. 6. The readthedocs.yml has been fully updated to the V2 schema so that we can use Python 3.9 for building the docs. This is required as the `class NDArray(Generic[T], np.ndarray[Any, T])` in `utils._typing` causes an error with `autodoc` on older Python versions.
* Fix for sklearn classifier * Fix issue with orig_files in saving.py * Fix NDArray on Python<3.9
Add MANIFEST.in, update CHANGELOG.md and version info.
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
All tests passed except |
ascillitoe
added a commit
that referenced
this pull request
Jul 7, 2022
This reverts commit d05ec7d.
ascillitoe
added a commit
that referenced
this pull request
Jul 7, 2022
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.
Merging
0.10.0rc1back tomasterso that we can include the optional dependency work (#537) in the final0.10.0release.