v0.7.2

Changes

🚀 Features

• Use np.random.Generator for rng @benfred (#694)
• use threadpoolctl for detecting BLAS threadpools @benfred (#692)

🧰 Maintenance

• bump cibuildwheel version in gha @benfred (#693)

v0.7.1

Changes

🚀 Features

• Add support for H100 gpu's @benfred (#669)

🐛 Bug Fixes

• Clear cached properties on partial_fit_* @benfred (#685)
• Add check_csr for partial_fit methods @benfred (#683)
• Add noexcept declarations for Cython 3.0 @benfred (#680)
• Fix cuda build on conda-forge @benfred (#668)

🧰 Maintenance

• use raft 23.08 @benfred (#681)
• Add cmake/ninja to pyproject.toml @benfred (#677)
• Remove setup_requires @benfred (#674)

v0.7.0

Changes

🚀 Features

• use RAFT for select_k on the gpu @benfred (#656)
• float16 support for GPU als model @benfred (#661)

🐛 Bug Fixes

• Fix ALS loss calculation on GPU @benfred (#663)
• Fix for ALS calculate_loss on the cpu @benfred (#662)

🧰 Maintenance

• Build / publish sdist package to pypi as well @benfred (#665)
• Set cython directives globally @benfred (#666)
• Test h5py in GHA for python 3.11 @benfred (#664)
• Use the batch functionality in the movielens example @benfred (#659)
• update isort/pylint/black versions @benfred (#657)
• Remove import of numpy.distutils @benfred (#660)

v0.6.2

Changes

🚀 Features

• Build arm64 wheels for linux @benfred (#630)
• Build python 3.11 wheels @benfred (#634)

🐛 Bug Fixes

• Fix more issues with pickling GPU models @benfred (#636)
• Fix pickling GPU models @benfred (#632)
• Fix ordering in .recommend for KNN models @benfred (#629)

📄 Documentation

• Update README / github actions versions @benfred (#635)
• Update evaluation.pyx @mrticker (#617)
• Fix CosineRecommender documentation @benfred (#619)

🧰 Maintenance

• fix for newer packaging versions @benfred (#633)
• Remove unnecessary argument @Soonmok (#621)

v0.6.1

Changes

🚀 Features

• Tweak callback API @benfred (#612)
• Allow GPU models to train on sparse matrices that exceed the size of available GPU memory @benfred (#605)
• Fit callback in BPR model @apat1n (#590)

🐛 Bug Fixes

• Handle more than 2**31 parameters on GPU models @benfred (#599)
• Fix recalculate_item / recalculate_user on the GPU @benfred (#598)

📄 Documentation

• Minor documentation fixes @benfred (#609)
• docs: Fix a few typos @timgates42 (#596)

v0.6.0

Changes

🚀 Features

• Add an alpha parameter for the ALS models @benfred (#581)
• Use num_threads for batch recommendations in MF models @benfred (#562)
• Add save/load functions for models @benfred (#552)
• Publish binary wheels to PyPi using cibuildwheel @benfred (#546)

🐛 Bug Fixes

• Fix default argument typo in benchmark_als.py @atakanfilgoz (#585)
• Fix save/load for models that haven't been fit @benfred (#577)
• Warn on non-csr input passed to model.fit @benfred (#578)
• Fix partial_fit_items bug with only growing by a single item @benfred (#563)

🏎💨 Performance Improvements

• Only call sort_indices when verifying negative samples @benfred (#583)

📄 Documentation

• Update link to dataset in lastfm example script @benfred (#573)

🧰 Maintenance

• Add release-drafter workflow for generating release notes @benfred (#579)
• Disable building GPU extension if IMPLICIT_DISABLE_CUDA env var is set @benfred (#561)
• Build binary wheels for apple silicon @benfred (#553)
• Use scikit-build>=0.13.1 @benfred (#549)
• Disable compute_86 compilation for cuda 11.0 @benfred (#536)
• Require cmake v3.17+ @benfred (#534)

v0.5.0

v0.5.0

Breaking API Changes

The API for implicit has substantially changed in v0.5.0 - and any code written for the previous
API will need to be rewritten:

• Change model.fit to take a user_items sparse matrix #484
• Return numpy arrays from recommend methods #482
• Don't require empty rows in user_items and item_users parameters #526
• Unify API for rank_items/recommend/recommend_all #489

Performance Improvements

• Speedup evaluation by using batch recommend functions #520
• Use FAISS for GPU inference #506
• Multithreaded speedups for CPU models #517
• Use thrust::binary_search to verify negative samples on GPU #524
• Release GIL on GPU calls #528

New Features

• Add incremental retraining support for ALS models #527
• Add filtering for similar_items and similar_users #488
• Add support for recalculate_users/items on the GPU #515
• Add methods for converting MF models to/from gpu #521
• Add a tutorial notebook for the lastfm example #529
• Approximate nearest neighbour for BPR/LMF and GPU models #487
• Dynamically detect CUDA availability #174