Nir Hutnik

Hi @tzoof ! Thanks for opening an issue! However, I'm not exactly sure what you meant here. I understand that you have an internal process in which with each commit...

### Tasks: * Remove "sample_size" from init: - 1 SP ** in ImageDatasetDrift (replace with n_samples from run) ** in train_test_validation suite (remove from init, will be passed in run)...

### Inputs: Add new drift method - "cramers_v_corrected" or something similar that is clear that corrects small categories. ### Logic: Try 2 methods: 1. all categories with less than X...

### Logic: In core/serialization/common.py, in create_failures_dataframe, we add the type of exception to the message we add to failed check. We should add a filter there, that for some Exceptions,...

Ok, scratch that: code before that just states which exceptions to include the name in, should just remove ModelValidationError and make sure that it makes sense for everywhere it's raised

Consider whether to add to this task to publish an imbalanced-data use-case, or as a separate task

### DoD: User can now ask us to pass custom weights, or automatically balance the label when searching for label or prediction drift. ### Inputs: class_weights: list[float] / 'weighted' ###...

1. Start from 2nd bullet - create infra 2. Continue with 4th bullet - Create Dice metric (prediction X true) 3. Plan next steps

Continued by #1888

Reducing to 3 SPs - * going over all checks and updating names, signatures and adding deprecation warnings - 2 SPS (using pycharm's refactor will do this for plots and...