Fix checkpoint fallback 3595#3685
Open
SpiliosDimakopoulos wants to merge 4 commits into
Open
Conversation
added 4 commits
July 25, 2026 13:08
…oolkit#3677) temporal_importance_curves() raised 'ValueError: not enough values to unpack (expected 2, got 0)' when every tree in a fitted interval forest was a stump (no internal splits), because it tried to unpack an empty zip(sorted(...)). - temporal_importance_curves() now returns ([], []) instead of raising when the forest contains no splits. - plot_temporal_importance_curves() now raises a clear ValueError when given empty curves, instead of silently plotting NaN (previously np.mean([], axis=0) produced a RuntimeWarning and a NaN dashed line). - test_tic_curves is now deterministic (random_state=0), since it previously could hit this exact edge case flakily depending on the random split found in a tiny test forest. - Added a regression test (test_tic_curves_all_stumps) using a constant input series, which deterministically produces an all-stump forest. Fixes aeon-toolkit#3677
…-toolkit#3418) 'max_dims' predates aeon settling on 'channels' terminology for multivariate series dimensions, and is inconsistent with the rest of the codebase (e.g. n_channels_, capability:multivariate). This mirrors the identical rename already done for OrdinalTDE and IndividualOrdinalTDE (which already expose 'max_channels' with a deprecated 'max_dims' alias): adds 'max_channels' to TemporalDictionaryEnsemble and IndividualTDE, keeps 'max_dims' for one deprecation cycle with a FutureWarning mapping it to 'max_channels', and updates the internal IndividualTDE construction inside TemporalDictionaryEnsemble accordingly. Also updates/extends aeon/classification/dictionary_based/tests/test_tde.py, including a new test for the deprecation warning. Fixes aeon-toolkit#3418
…olkit#3595) 'try: tf.keras.models.load_model(...) except FileNotFoundError:' is unreliable: with save_best_only=True the ModelCheckpoint callback is not guaranteed to have written a file by the time fit() tries to load it back, and depending on the TensorFlow/Keras backend a missing/unwritten file can raise something other than FileNotFoundError, causing intermittent CI failures (see aeon-toolkit#3595, fixed for InceptionTimeRegressor in aeon-toolkit#3601). This applies the same fix already merged for InceptionTimeRegressor (explicit os.path.exists() check, falling back to the trained in-memory model if the checkpoint was never written) across every other deep learner with the same pattern: - aeon/classification/deep_learning: _inception_time, _lite_time, _fcn, _mlp, _resnet, _encoder, _cnn, _disjoint_cnn - aeon/regression/deep_learning: _fcn, _lite_time, _mlp, _resnet, _encoder, _cnn, _disjoint_cnn, _rnn - aeon/clustering/deep_learning: _ae_resnet, _ae_drnn, _ae_bgru, _ae_dcnn, _ae_fcn, _ae_abgru - aeon/transformations/collection/self_supervised: _trilite, _time_mcl Adds parametrized regression tests (mirroring the one added in aeon-toolkit#3601) across all fixed classification, regression, and self-supervised estimators, verifying fit() completes and falls back correctly when the checkpoint callback never writes a file. The existing parametrized aeon/clustering/deep_learning/tests/test_deep_clusterer_io.py already covers all 6 fixed clustering estimators and continues to pass. Fixes aeon-toolkit#3595
The old 'znormalized' parameter defaulted to True, meaning 'do not z-normalize because the input is already normalized' -- confusing, since True implied normalization would happen. This introduces 'znormalize' (default=True), which now always means 'z-normalize the windows'. The old 'znormalized' parameter is kept for one deprecation cycle: passing it raises a FutureWarning and maps to the equivalent 'znormalize' value (znormalize = not znormalized), per the project's deprecation policy. Also updates the SAX example notebook and REDCOMETS' internal SAX call to the new parameter name, and updates/extends the SAX test suite (including new tests for the deprecation warning and its behavioural equivalence). Fixes aeon-toolkit#3678
SpiliosDimakopoulos
requested review from
MatthewMiddlehurst,
TonyBagnall,
baraline,
chrisholder,
dguijo,
hadifawaz1999 and
patrickzib
as code owners
July 25, 2026 10:38
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
Contributor
Thank you for contributing to
|
Contributor
|
thanks for this, please start the PR title with either [ENH], [MNT], [DOC], [BUG], [REF], [DEP] or [GOV] as per template instructions and actually describe your PR |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
No description provided.