Pytorch-frame

Added

- Added support for Python 3.12 and 3.13 ([496](https://github.com/pyg-team/pytorch-frame/pull/496))
- Added support for PyTorch 2.6 ([494](https://github.com/pyg-team/pytorch-frame/pull/494))

Added

- Added an example for training `Trompt` on multiple GPUs ([474](https://github.com/pyg-team/pytorch-frame/pull/474))
- Added support for materializing dataset for train and test dataframe separately([470](https://github.com/pyg-team/pytorch-frame/issues/470))
- Added support for PyTorch 2.5 ([464](https://github.com/pyg-team/pytorch-frame/pull/464))
- Added a benchmark script to compare PyTorch Frame with PyTorch Tabular ([398](https://github.com/pyg-team/pytorch-frame/pull/398), [#444](https://github.com/pyg-team/pytorch-frame/pull/444))
- Added `is_floating_point` method to `MultiNestedTensor` and `MultiEmbeddingTensor` ([445](https://github.com/pyg-team/pytorch-frame/pull/445))
- Added support for inferring `stype.categorical` from boolean columns in `utils.infer_series_stype` ([421](https://github.com/pyg-team/pytorch-frame/pull/421))
- Added `pin_memory()` to `TensorFrame`, `MultiEmbeddingTensor`, and `MultiNestedTensor` ([437](https://github.com/pyg-team/pytorch-frame/pull/437))

Changed

- Set `weights_only=True` in `torch_frame.load` from PyTorch 2.4 ([423](https://github.com/pyg-team/pytorch-frame/pull/423))

Deprecated

Removed

- Dropped support for Python 3.8 ([462](https://github.com/pyg-team/pytorch-frame/pull/462))

Fixed

- Fixed size mismatch `RuntimeError` in `transforms.CatToNumTransform` ([446](https://github.com/pyg-team/pytorch-frame/pull/446))
- Removed CUDA synchronizations from `nn.LinearEmbeddingEncoder` ([432](https://github.com/pyg-team/pytorch-frame/pull/432))
- Removed CUDA synchronizations from N/A imputation logic in `nn.StypeEncoder` ([433](https://github.com/pyg-team/pytorch-frame/pull/433), [#434](https://github.com/pyg-team/pytorch-frame/pull/434))

Added

- Added `MovieLens 1M` dataset ([397](https://github.com/pyg-team/pytorch-frame/pull/397))
- Added light-weight MLP ([372](https://github.com/pyg-team/pytorch-frame/pull/372))
- Added R^2 metric ([403](https://github.com/pyg-team/pytorch-frame/pull/403))

Changed

- Updated `ExcelFormer` implementation and related scripts ([391](https://github.com/pyg-team/pytorch-frame/pull/391))

Added

- Avoided for-loop in `EmbeddingEncoder` ([366](https://github.com/pyg-team/pytorch-frame/pull/366))
- Added `image_embedded` and one tabular image dataset ([344](https://github.com/pyg-team/pytorch-frame/pull/344))
- Added benchmarking suite for encoders ([360](https://github.com/pyg-team/pytorch-frame/pull/360))
- Added dataframe text benchmark script ([354](https://github.com/pyg-team/pytorch-frame/pull/354), [#367](https://github.com/pyg-team/pytorch-frame/pull/367))
- Added `DataFrameTextBenchmark` dataset ([349](https://github.com/pyg-team/pytorch-frame/pull/349))
- Added support for empty `TensorFrame` ([339](https://github.com/pyg-team/pytorch-frame/pull/339))

Changed

- Changed a workflow of Encoder's `na_forward` method resulting in performance boost ([364](https://github.com/pyg-team/pytorch-frame/pull/364))
- Removed ReLU applied in `FCResidualBlock` ([368](https://github.com/pyg-team/pytorch-frame/pull/368))

Deprecated

Removed

Fixed

- Fixed bug in empty `MultiNestedTensor` handling ([369](https://github.com/pyg-team/pytorch-frame/pull/369))
- Fixed the split of `DataFrameTextBenchmark` ([358](https://github.com/pyg-team/pytorch-frame/pull/358))
- Fixed empty `MultiNestedTensor` col indexing ([355](https://github.com/pyg-team/pytorch-frame/pull/355))

Added

- Support more stypes in `LinearModelEncoder` ([325](https://github.com/pyg-team/pytorch-frame/pull/325))
- Added `stype_encoder_dict` to some models ([319](https://github.com/pyg-team/pytorch-frame/pull/319))
- Added `HuggingFaceDatasetDict` ([287](https://github.com/pyg-team/pytorch-frame/pull/287))

Changed

- Supported decoder embedding model in `examples/transformers_text.py` ([333](https://github.com/pyg-team/pytorch-frame/pull/333))
- Removed implicit clones in `StypeEncoder` ([286](https://github.com/pyg-team/pytorch-frame/pull/286))

Deprecated

Removed

Fixed

- Fixed `TimestampEncoder` not applying `CyclicEncoder` to cyclic features ([311](https://github.com/pyg-team/pytorch-frame/pull/311))
- Fixed NaN masking in `multicateogrical` stype ([307](https://github.com/pyg-team/pytorch-frame/pull/307))

Added

- Added support for Boolean masks in `index_select` of `_MultiTensor` [334](https://github.com/pyg-team/pytorch-frame/pull/334)
- Added more text documentation ([291](https://github.com/pyg-team/pytorch-frame/pull/291))
- Added `col_to_model_cfg` ([270](https://github.com/pyg-team/pytorch-frame/pull/270))
- Support saving/loading of GBDT models ([269](https://github.com/pyg-team/pytorch-frame/pull/269))
- Added documentation on handling different stypes ([271](https://github.com/pyg-team/pytorch-frame/pull/271))
- Added `TimestampEncoder` ([225](https://github.com/pyg-team/pytorch-frame/pull/225))
- Added `LightGBM` ([248](https://github.com/pyg-team/pytorch-frame/pull/248))
- Added time columns to the `MultimodalTextBenchmark` ([253](https://github.com/pyg-team/pytorch-frame/pull/253))
- Added `CyclicEncoding` ([251](https://github.com/pyg-team/pytorch-frame/pull/251))
- Added `PositionalEncoding` ([249](https://github.com/pyg-team/pytorch-frame/pull/249))
- Added optional `col_names` argument in `StypeEncoder` ([247](https://github.com/pyg-team/pytorch-frame/pull/247))
- Added `col_to_text_embedder_cfg` and use `MultiEmbeddingTensor` for `text_embedded` ([246](https://github.com/pyg-team/pytorch-frame/pull/246))
- Added `col_encoder_dict` in `StypeWiseFeatureEncoder` ([244](https://github.com/pyg-team/pytorch-frame/pull/244))
- Added `LinearEmbeddingEncoder` for `embedding` stype ([243](https://github.com/pyg-team/pytorch-frame/pull/243))
- Added support for `torch_frame.text_embedded` in `GBDT` ([239](https://github.com/pyg-team/pytorch-frame/pull/239))
- Support `Metric` in `GBDT` ([236](https://github.com/pyg-team/pytorch-frame/pull/236))
- Added auto-inference of `stype` ([221](https://github.com/pyg-team/pytorch-frame/pull/221))
- Enabled `list` input in `multicategorical` stype ([224](https://github.com/pyg-team/pytorch-frame/pull/224))
- Added `Timestamp` stype ([212](https://github.com/pyg-team/pytorch-frame/pull/212))
- Added `multicategorical` to `MultimodalTextBenchmark` ([208](https://github.com/pyg-team/pytorch-frame/pull/208))
- Added support for saving and loading of `TensorFrame` with complex `stypes`. ([197](https://github.com/pyg-team/pytorch-frame/pull/197))
- Added `stype.embedding` ([194](https://github.com/pyg-team/pytorch-frame/pull/194))
- Added `TensorFrame` concatenation of complex stypes. ([190](https://github.com/pyg-team/pytorch-frame/pull/190))
- Added `text_tokenized` example ([174](https://github.com/pyg-team/pytorch-frame/pull/174))
- Added Cohere embedding example ([186](https://github.com/pyg-team/pytorch-frame/pull/186))
- Added `AmazonFineFoodReviews` dataset and OpenAI embedding example ([182](https://github.com/pyg-team/pytorch-frame/pull/182))
- Added save and load logic for `FittableBaseTransform` ([178](https://github.com/pyg-team/pytorch-frame/pull/178))
- Added `MultiEmbeddingTensor` ([181](https://github.com/pyg-team/pytorch-frame/pull/181), [#193](https://github.com/pyg-team/pytorch-frame/pull/193), [#198](https://github.com/pyg-team/pytorch-frame/pull/198), [#199](https://github.com/pyg-team/pytorch-frame/pull/199), [#217](https://github.com/pyg-team/pytorch-frame/pull/217))
- Added `to_dense()` for `MultiNestedTensor` ([170](https://github.com/pyg-team/pytorch-frame/pull/170))
- Added example for `multicategorical` stype ([162](https://github.com/pyg-team/pytorch-frame/pull/162))
- Added `sequence_numerical` stype ([159](https://github.com/pyg-team/pytorch-frame/pull/159))
- Added `MultiCategoricalEmbeddingEncoder` ([155](https://github.com/pyg-team/pytorch-frame/pull/155))
- Added advanced indexing for `MultiNestedTensor` ([150](https://github.com/pyg-team/pytorch-frame/pull/150), [#161](https://github.com/pyg-team/pytorch-frame/pull/161), [#163](https://github.com/pyg-team/pytorch-frame/pull/163), [#165](https://github.com/pyg-team/pytorch-frame/pull/165))
- Added `multicategorical` stype ([128](https://github.com/pyg-team/pytorch-frame/pull/128), [#151](https://github.com/pyg-team/pytorch-frame/pull/151))
- Added `MultiNestedTensor` ([149](https://github.com/pyg-team/pytorch-frame/pull/149))

Changed

- Set `stype.embedding` as the parent of `stype.text_embedded` and unified `stype.text_embedded` with its parent in :obj:`tensor_frame` ([277](https://github.com/pyg-team/pytorch-frame/pull/277))
- Renamed `torch_frame.stype` module to `torch_frame._stype` ([275](https://github.com/pyg-team/pytorch-frame/pull/275))
- Renamed `text_tokenized_cfg` into `col_to_text_tokenized_cfg` ([257](https://github.com/pyg-team/pytorch-frame/pull/257))
- Made `Trompt` output 2-dim embeddings in `forward`
- Renamed `text_embedder_cfg` into `col_to_text_embedder_cfg`

Removed

- No manual passing of `in_channels` to `LinearEmbeddingEncoder` for `stype.text_embedded` ([222](https://github.com/pyg-team/pytorch-frame/pull/222))