Librecommender

💥 Breaking Changes
+ Explicit negative sampling(`build_negative_samples`) has been deprecated. It will be removed in the future (263).
+ Negative sampling is controlled by the `neg_sampling` parameter (254). See [Negative sampling](https://librecommender.readthedocs.io/en/latest/user_guide/data_processing.html#negative-sampling).
+ [merge_trainset](https://librecommender.readthedocs.io/en/latest/api/data/dataset.html#libreco.data.dataset.DatasetFeat.merge_trainset) will return a new `data_info` (250).

🚀 Features
+ Add dynamic sequence recommendation (257).
+ Add more sampling strategies (239).
+ Multiprocess data loading (237).

🐛 Bug Fixes
+ Fix dense layer activation order (249).

📚 Documentation
+ Add Colab notebook (230).

🧻 Miscellaneous
+ Add Windows & macOs CI (246).

**Full Changelog**: https://github.com/massquantity/LibRecommender/compare/v1.0.1...v1.1.0

🐛 Bug Fixes
* Fix PyTorch device issue (222).

**Full Changelog**: https://github.com/massquantity/LibRecommender/compare/v1.0.0...v1.0.1

💥 Breaking Changes
+ The parameter type of `hidden_units` and `device` has been changed (205).
+ Data construction in model retrain use new functions (212).

🐛 Bug Fixes
+ Fix incompatible batch_normalization and initializer layer from TensorFlow2 (206).

📚 Documentation
+ Add formal [documentation](https://librecommender.readthedocs.io/en/stable/) (#208).
+ Add docstrings (207).

🧻 Miscellaneous
+ Use [`pyproject.toml`](https://github.com/massquantity/LibRecommender/blob/master/pyproject.toml) for building project according to [PEP 621](https://peps.python.org/pep-0621/), which is also officially recommended in [PyPA](https://packaging.python.org/en/latest/specifications/declaring-project-metadata/#declaring-project-metadata), [pip](https://pip.pypa.io/en/stable/reference/build-system/pyproject-toml/) and [Setuptools](https://setuptools.pypa.io/en/latest/userguide/pyproject_config.html) (#209).
+ Since currently Python3.6 can't be fully built from `pyproject.toml` in Setuptools, we read related metadata from `pyproject.toml` into `setup.py` (210).

**Full Changelog**: https://github.com/massquantity/LibRecommender/compare/v0.12.6...v1.0.0

🐛 Bug Fixes
+ Fix sequence data sampling bug (195)

📚 Documentation
+ Add some caveats on duplicate samples in data, see [User Guide](https://github.com/massquantity/LibRecommender/blob/master/doc/user_guide.md#data-format) (196)

**Full Changelog**: https://github.com/massquantity/LibRecommender/compare/v0.12.4...v0.12.6

🐛 Bug Fixes
+ Fix dense feature bugs in `DIN` & `GraphSage` & `PinSage` (184)

🧻 Miscellaneous
+ Replace `pytest-cov` with [`coveragepy`](https://github.com/nedbat/coveragepy) in CI (#185)
+ Add Sanic serving test (187)
+ Lint with [`ruff`](https://github.com/charliermarsh/ruff) and [`isort`](https://github.com/PyCQA/isort) (#188)
+ Limit num of requests with semaphore in tf model serving (190)

💥 Breaking Changes
+ The recommendation result will return a dict, with users as keys and `numpy.array` of items as values.

🚀 Features
+ Add recommend strategy (170). See [User Guide](https://github.com/massquantity/LibRecommender/blob/master/doc/user_guide.md#recommend-strategy).
+ Add batch recommend (171).

🐛 Bug Fixes
+ Fix `UserCF` & `ItemCF` inner_id bugs (172)

📚 Documentation
+ Add [Implementation Details](https://github.com/massquantity/LibRecommender/blob/master/doc/implementation_details.md) (#174)
+ Update User Guide (174)