You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Added
Add AutoHurdleRegressor for zero-inflated regression with configurable estimators and quantile grids, mean, median, and quantile predictions, and a Facebook Comment Volume example. (#395)
Add an example that fits the Amazon Bestseller task of MulTaBench with TabPFNWithImages, on the numeric fields, on the picture and on both. (#398)
Add the image extra: pip install "tabpfn-extensions[image]" installs transformers, torchvision and pillow, the dependencies of the image extension. (#399)
Add the image extension: TabPFNWithImages and ImageTransformer turn declared image columns, file paths, base64 strings, bytes or PIL images, into frozen DINOv3 features before TabPFN sees them. (#400)
Add the internals of the image extension: PIL decoding of image cells, given as base64, bytes, a file path or a PIL image, image_to_bytes, and the frozen DINOv3 encoder with its image processor. (#401)