Feature release
General DataModule (#253)
This DataModule supports loading datasets from Hugging Face Datasets format (e.g., "ylecun/mnist") or from local parquet files. It also allows specifying which columns to load and how to transform them using Column objects. It also handles splitting the dataset into training, validation, and test sets based on specified proportions.
Added Visual Transformer (#253)
It takes an image input and produces a fixed-size embedding vector. The CLS token from the ViT model is used as the output representation, and an optional linear projection can be applied on top of it to match a desired output dimension.
Other Changes
- Add validation loss (#252)
- Add GPU in-memory option for DataModule (#246)
- Update model training section in ReadTheDocs
- Clean-up experiment YAML files
Full Changelog: v0.4...v0.5.1