Jonas Mueller
Jonas Mueller
To apply tabular_prediction in settings where dataset does not fit into memory, each model should have specialized functionality to train without having to read in full dataset. In preprocessing of...
Current versions are there to: - Avoid this bug in the tensorflow-io library: https://github.com/tensorflow/io/issues/1687 - Avoid this bug in speechbrain library: https://github.com/speechbrain/speechbrain/issues/1459 Remaining TODOs (need to wait until next version...
Instead of setting the number of classes based on `len(unique(labels))`. This addresses issue: https://github.com/cleanlab/cleanlab/issues/85 Thanks @[vtsouval](https://github.com/vtsouval) for pointing out where some of the fixes had to be made and providing...
Currently `find_label_issues` docstring contains the following: Tip: If you encounter the error 'pred_probs is not defined', try setting `n_jobs = 1`. https://github.com/cleanlab/cleanlab/blob/e1d9ae4fcf95e802f4a8e112845929ee785705eb/cleanlab/filter.py#L257 Try to resolve this or provide clearer error...
Increment beyond latest released version so folks can more easily know when they are using developer version vs latest official release.
Also ensures audio tutorial always runs without manual user install (eg. on colab)
Beyond just enforcing `black` code style, we also want to run some extra CI checks to flag bad practices like: - `import *` - imports that are unused in the...
Run a version of the unit tests without all optional dependencies / lazy imports installed, in order to ensure these truly are not required to `import cleanlab`.
Specifically in the index.html, search.html files, I manually added v2.0.0 version link in sidebars when we released v2.1.0. But this should be there automatically in future releases. - [ ]...