+ [Fix] Fix customized preprossors based on polars and update demos + [Doc] Add copyrights + [Feature] Add `embedding_dim` setting to numeric features
2.3
[Doing] Add support for saving pb file, exporting embeddings [Doing] Add support of multi-gpu training
2.3.0
+ [Refactor] Support reading CSV and Parquet files as inputs + [Feature] Add dataloader for parquet + [Feature] Add the `rebuild_dataset=False` setting to skip rebuiding when the input dataset has already been preprocessed with ID feature mapping. This enables customized feature mapping instead of using FuxiCTR Preprocessor (which is slow for large dataset).
-------------------------------
2.2.3
+ [Fix] Quick fix to v2.2.2 that miss one line when committing
2.2.2
+ [Feature] Update to use polars instead of pandas for faster feature processing + [Fix] When num_workers > 1, NpzBlockDataLoader cannot keep the reading order of samples ([86](https://github.com/xue-pai/FuxiCTR/issues/86))
2.2.1
+ [Fix] Fix issue of evaluation not performed at epoch end when streaming=True ([85](https://github.com/xue-pai/FuxiCTR/issues/85)) + [Fix] Fix issue when loading pretrain_emb in npz format ([84](https://github.com/xue-pai/FuxiCTR/issues/84))