Final updates for the 0.5.0 release (#532)

* Fix lastfm.py example for recommend
* Disable building GPU extension on windows
* Update changelog

CHANGELOG.md

@@ -1,3 +1,33 @@
+## 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](https://github.com/benfred/implicit/pull/484)
+* Return numpy arrays from recommend methods [#482](https://github.com/benfred/implicit/pull/482)
+* Don't require empty rows in user_items and item_users parameters [#526](https://github.com/benfred/implicit/pull/526)
+* Unify API for rank_items/recommend/recommend_all [#489](https://github.com/benfred/implicit/issues/489)
+
+### Performance Improvements
+
+* Speedup evaluation by using batch recommend functions [#520](https://github.com/benfred/implicit/pull/520)
+* Use FAISS for GPU inference [#506](https://github.com/benfred/implicit/pull/506)
+* Multithreaded speedups for CPU models [#517](https://github.com/benfred/implicit/pull/517)
+* Use thrust::binary_search to verify negative samples on GPU [#524](https://github.com/benfred/implicit/pull/524)
+* Release GIL on GPU calls [#528](https://github.com/benfred/implicit/pull/528)
+
+### New Features
+
+* Add incremental retraining support for ALS models [#527](https://github.com/benfred/implicit/pull/527)
+* Add filtering for similar_items and similar_users [#488](https://github.com/benfred/implicit/issues/488)
+* Add support for recalculate_users/items on the GPU [#515](https://github.com/benfred/implicit/pull/515)
+* Add methods for converting MF models to/from gpu [#521](https://github.com/benfred/implicit/pull/521)
+* Add a tutorial notebook for the lastfm example [#529](https://github.com/benfred/implicit/pull/529)
+* Approximate nearest neighbour for BPR/LMF and GPU models [#487](https://github.com/benfred/implicit/issues/487)
+* Dynamically detect CUDA availability [#174](https://github.com/benfred/implicit/issues/174)
+
 ## v0.4.5
 
 * GPU Inference [#406](https://github.com/benfred/implicit/pull/406)

examples/lastfm.py

@@ -151,7 +151,9 @@ def calculate_recommendations(output_filename, model_name="als"):
             to_generate = np.arange(len(users))
             for startidx in range(0, len(to_generate), batch_size):
                 batch = to_generate[startidx : startidx + batch_size]
-                ids, scores = model.recommend(batch, user_plays, filter_already_liked_items=True)
+                ids, scores = model.recommend(
+                    batch, user_plays[batch], filter_already_liked_items=True
+                )
                 for i, userid in enumerate(batch):
                     username = users[userid]
                     for other, score in zip(ids[i], scores[i]):

implicit/gpu/CMakeLists.txt

@@ -1,10 +1,12 @@
-add_cython_target(_cuda CXX)
-
+if (UNIX)
 find_package(CUDAToolkit)
 
-
 if(CUDAToolkit_FOUND)
+if (${CUDAToolkit_VERSION} VERSION_LESS "11.0.0")
+    message("implicit requires CUDA 11.0 or greater for GPU acceleration - found CUDA ${CUDAToolkit_VERSION}")
+else()
     enable_language(CUDA)
+    add_cython_target(_cuda CXX)
 
     # use rapids-cmake to install dependencies
     file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-22.02/RAPIDS.cmake
@@ -60,6 +62,8 @@ if(CUDAToolkit_FOUND)
 
     install(TARGETS _cuda LIBRARY DESTINATION implicit/gpu)
 endif()
+endif()
+endif()
 
 FILE(GLOB gpu_python_files *.py)
 install(FILES ${gpu_python_files} DESTINATION implicit/gpu)