Forward-merge branch-21.08 to branch-21.10 [skip gpuci]

CHANGELOG.md

@@ -2,9 +2,101 @@
 
 Please see https://github.com/rapidsai/cuml/releases/tag/v21.10.00a for the latest changes to this development branch.
 
-# cuML 21.08.00 (Date TBD)
+# cuML 21.08.00 (4 Aug 2021)
 
-Please see https://github.com/rapidsai/cuml/releases/tag/v21.08.00a for the latest changes to this development branch.
+## 🚨 Breaking Changes
+
+- Remove deprecated target_weights in UMAP ([#4081](https://github.com/rapidsai/cuml/pull/4081)) [@lowener](https://github.com/lowener)
+- Upgrade Treelite to 2.0.0 ([#4072](https://github.com/rapidsai/cuml/pull/4072)) [@hcho3](https://github.com/hcho3)
+- RF/DT cleanup ([#4005](https://github.com/rapidsai/cuml/pull/4005)) [@venkywonka](https://github.com/venkywonka)
+- RF: memset and batch size optimization for computing splits ([#4001](https://github.com/rapidsai/cuml/pull/4001)) [@venkywonka](https://github.com/venkywonka)
+- Remove old RF backend ([#3868](https://github.com/rapidsai/cuml/pull/3868)) [@RAMitchell](https://github.com/RAMitchell)
+- Enable warp-per-tree inference in FIL for regression and binary classification ([#3760](https://github.com/rapidsai/cuml/pull/3760)) [@levsnv](https://github.com/levsnv)
+
+## 🐛 Bug Fixes
+
+- Disabling umap reproducibility tests for cuda 11.4 ([#4128](https://github.com/rapidsai/cuml/pull/4128)) [@cjnolet](https://github.com/cjnolet)
+- Fix for crash in RF when `max_leaves` parameter is specified ([#4126](https://github.com/rapidsai/cuml/pull/4126)) [@vinaydes](https://github.com/vinaydes)
+- Running umap mnmg test twice ([#4112](https://github.com/rapidsai/cuml/pull/4112)) [@cjnolet](https://github.com/cjnolet)
+- Minimal fix for `SparseRandomProjection` ([#4100](https://github.com/rapidsai/cuml/pull/4100)) [@viclafargue](https://github.com/viclafargue)
+- Creating copy of `components` in PCA transform and inverse transform ([#4099](https://github.com/rapidsai/cuml/pull/4099)) [@divyegala](https://github.com/divyegala)
+- Fix SVM model parameter handling in case n_support=0 ([#4097](https://github.com/rapidsai/cuml/pull/4097)) [@tfeher](https://github.com/tfeher)
+- Fix set_params for linear models ([#4096](https://github.com/rapidsai/cuml/pull/4096)) [@lowener](https://github.com/lowener)
+- Fix train test split pytest comparison ([#4062](https://github.com/rapidsai/cuml/pull/4062)) [@dantegd](https://github.com/dantegd)
+- Fix fit_transform on KMeans ([#4055](https://github.com/rapidsai/cuml/pull/4055)) [@lowener](https://github.com/lowener)
+- Fixing -1 key access in 1nn reduce op in HDBSCAN ([#4052](https://github.com/rapidsai/cuml/pull/4052)) [@divyegala](https://github.com/divyegala)
+- Disable installing gbench to avoid container permission issues ([#4049](https://github.com/rapidsai/cuml/pull/4049)) [@dantegd](https://github.com/dantegd)
+- Fix double fit crash in preprocessing models ([#4040](https://github.com/rapidsai/cuml/pull/4040)) [@viclafargue](https://github.com/viclafargue)
+- Always add `faiss` library alias if it's missing ([#4028](https://github.com/rapidsai/cuml/pull/4028)) [@trxcllnt](https://github.com/trxcllnt)
+- Fixing intermittent HBDSCAN pytest failure in CI ([#4025](https://github.com/rapidsai/cuml/pull/4025)) [@divyegala](https://github.com/divyegala)
+- HDBSCAN bug on A100 ([#4024](https://github.com/rapidsai/cuml/pull/4024)) [@divyegala](https://github.com/divyegala)
+- Add treelite include paths to treelite targets ([#4023](https://github.com/rapidsai/cuml/pull/4023)) [@trxcllnt](https://github.com/trxcllnt)
+- Add Treelite_BINARY_DIR include to `cuml++` build interface include paths ([#4018](https://github.com/rapidsai/cuml/pull/4018)) [@trxcllnt](https://github.com/trxcllnt)
+- Small ARIMA-related bug fixes in Hessenberg reduction and make_arima ([#4017](https://github.com/rapidsai/cuml/pull/4017)) [@Nyrio](https://github.com/Nyrio)
+- Update setup.py ([#4015](https://github.com/rapidsai/cuml/pull/4015)) [@ajschmidt8](https://github.com/ajschmidt8)
+- Update `treelite` version in `get_treelite.cmake` ([#4014](https://github.com/rapidsai/cuml/pull/4014)) [@ajschmidt8](https://github.com/ajschmidt8)
+- Fix build with latest RAFT branch-21.08 ([#4012](https://github.com/rapidsai/cuml/pull/4012)) [@trxcllnt](https://github.com/trxcllnt)
+- Skipping hdbscan pytests when gpu is a100 ([#4007](https://github.com/rapidsai/cuml/pull/4007)) [@cjnolet](https://github.com/cjnolet)
+- Using 64-bit array lengths to increase scale of pca & tsvd ([#3983](https://github.com/rapidsai/cuml/pull/3983)) [@cjnolet](https://github.com/cjnolet)
+- Fix MNMG test in Dask RF ([#3964](https://github.com/rapidsai/cuml/pull/3964)) [@hcho3](https://github.com/hcho3)
+- Use nested include in destination of install headers to avoid docker permission issues ([#3962](https://github.com/rapidsai/cuml/pull/3962)) [@dantegd](https://github.com/dantegd)
+- Fix automerge #3939 ([#3952](https://github.com/rapidsai/cuml/pull/3952)) [@dantegd](https://github.com/dantegd)
+- Update UCX-Py version to 0.21 ([#3950](https://github.com/rapidsai/cuml/pull/3950)) [@pentschev](https://github.com/pentschev)
+- Fix kernel and line info in cmake ([#3941](https://github.com/rapidsai/cuml/pull/3941)) [@dantegd](https://github.com/dantegd)
+- Fix for multi GPU PCA compute failing bug after transform and added error handling when n_components is not passed ([#3912](https://github.com/rapidsai/cuml/pull/3912)) [@akaanirban](https://github.com/akaanirban)
+- Tolerate QN linesearch failures when it's harmless ([#3791](https://github.com/rapidsai/cuml/pull/3791)) [@achirkin](https://github.com/achirkin)
+
+## 📖 Documentation
+
+- Improve docstrings for silhouette score metrics. ([#4026](https://github.com/rapidsai/cuml/pull/4026)) [@bdice](https://github.com/bdice)
+- Update CHANGELOG.md link ([#3956](https://github.com/rapidsai/cuml/pull/3956)) [@Salonijain27](https://github.com/Salonijain27)
+- Update documentation build examples to be generator agnostic ([#3909](https://github.com/rapidsai/cuml/pull/3909)) [@robertmaynard](https://github.com/robertmaynard)
+- Improve FIL code readability and documentation ([#3056](https://github.com/rapidsai/cuml/pull/3056)) [@levsnv](https://github.com/levsnv)
+
+## 🚀 New Features
+
+- Add Multinomial and Bernoulli Naive Bayes variants ([#4053](https://github.com/rapidsai/cuml/pull/4053)) [@lowener](https://github.com/lowener)
+- Add weighted K-Means sampling for SHAP ([#4051](https://github.com/rapidsai/cuml/pull/4051)) [@Nanthini10](https://github.com/Nanthini10)
+- Use chebyshev, canberra, hellinger and minkowski distance metrics ([#3990](https://github.com/rapidsai/cuml/pull/3990)) [@mdoijade](https://github.com/mdoijade)
+- Implement vector leaf prediction for fil. ([#3917](https://github.com/rapidsai/cuml/pull/3917)) [@RAMitchell](https://github.com/RAMitchell)
+- change TargetEncoder's smooth argument from ratio to count ([#3876](https://github.com/rapidsai/cuml/pull/3876)) [@daxiongshu](https://github.com/daxiongshu)
+- Enable warp-per-tree inference in FIL for regression and binary classification ([#3760](https://github.com/rapidsai/cuml/pull/3760)) [@levsnv](https://github.com/levsnv)
+
+## 🛠️ Improvements
+
+- Remove clang/clang-tools from conda recipe ([#4109](https://github.com/rapidsai/cuml/pull/4109)) [@dantegd](https://github.com/dantegd)
+- Pin dask version ([#4108](https://github.com/rapidsai/cuml/pull/4108)) [@galipremsagar](https://github.com/galipremsagar)
+- ANN warnings/tests updates ([#4101](https://github.com/rapidsai/cuml/pull/4101)) [@viclafargue](https://github.com/viclafargue)
+- Removing local memory operations from computeSplitKernel and other optimizations ([#4083](https://github.com/rapidsai/cuml/pull/4083)) [@vinaydes](https://github.com/vinaydes)
+- Fix libfaiss dependency to not expressly depend on conda-forge ([#4082](https://github.com/rapidsai/cuml/pull/4082)) [@Ethyling](https://github.com/Ethyling)
+- Remove deprecated target_weights in UMAP ([#4081](https://github.com/rapidsai/cuml/pull/4081)) [@lowener](https://github.com/lowener)
+- Upgrade Treelite to 2.0.0 ([#4072](https://github.com/rapidsai/cuml/pull/4072)) [@hcho3](https://github.com/hcho3)
+- Optimize dtype conversion for FIL ([#4070](https://github.com/rapidsai/cuml/pull/4070)) [@dantegd](https://github.com/dantegd)
+- Adding quick notes to HDBSCAN public API docs as to why discrepancies may occur between cpu and gpu impls. ([#4061](https://github.com/rapidsai/cuml/pull/4061)) [@cjnolet](https://github.com/cjnolet)
+- Update `conda` environment name for CI ([#4039](https://github.com/rapidsai/cuml/pull/4039)) [@ajschmidt8](https://github.com/ajschmidt8)
+- Rewrite random forest gtests ([#4038](https://github.com/rapidsai/cuml/pull/4038)) [@RAMitchell](https://github.com/RAMitchell)
+- Updating Clang Version to 11.0.0 ([#4029](https://github.com/rapidsai/cuml/pull/4029)) [@codereport](https://github.com/codereport)
+- Raise ARIMA parameter limits from 4 to 8 ([#4022](https://github.com/rapidsai/cuml/pull/4022)) [@Nyrio](https://github.com/Nyrio)
+- Testing extract clusters in HDBSCAN ([#4009](https://github.com/rapidsai/cuml/pull/4009)) [@divyegala](https://github.com/divyegala)
+- ARIMA - Kalman loop rewrite: single megakernel instead of host loop ([#4006](https://github.com/rapidsai/cuml/pull/4006)) [@Nyrio](https://github.com/Nyrio)
+- RF/DT cleanup ([#4005](https://github.com/rapidsai/cuml/pull/4005)) [@venkywonka](https://github.com/venkywonka)
+- Exposing condensed hierarchy through cython for easier unit-level testing ([#4004](https://github.com/rapidsai/cuml/pull/4004)) [@cjnolet](https://github.com/cjnolet)
+- Use the 21.08 branch of rapids-cmake as rmm requires it ([#4002](https://github.com/rapidsai/cuml/pull/4002)) [@robertmaynard](https://github.com/robertmaynard)
+- RF: memset and batch size optimization for computing splits ([#4001](https://github.com/rapidsai/cuml/pull/4001)) [@venkywonka](https://github.com/venkywonka)
+- Reducing cluster size to number of selected clusters. Returning stability scores ([#3987](https://github.com/rapidsai/cuml/pull/3987)) [@cjnolet](https://github.com/cjnolet)
+- HDBSCAN: Lazy-loading (and caching) condensed & single-linkage tree objects ([#3986](https://github.com/rapidsai/cuml/pull/3986)) [@cjnolet](https://github.com/cjnolet)
+- Fix `21.08` forward-merge conflicts ([#3982](https://github.com/rapidsai/cuml/pull/3982)) [@ajschmidt8](https://github.com/ajschmidt8)
+- Update Dask/Distributed version ([#3978](https://github.com/rapidsai/cuml/pull/3978)) [@pentschev](https://github.com/pentschev)
+- Use clang-tools on x86 only ([#3969](https://github.com/rapidsai/cuml/pull/3969)) [@jakirkham](https://github.com/jakirkham)
+- Promote `trustworthiness_score` to public header, add missing includes, update dependencies ([#3968](https://github.com/rapidsai/cuml/pull/3968)) [@trxcllnt](https://github.com/trxcllnt)
+- Moving FAISS ANN wrapper to raft ([#3963](https://github.com/rapidsai/cuml/pull/3963)) [@cjnolet](https://github.com/cjnolet)
+- Add MG weighted k-means ([#3959](https://github.com/rapidsai/cuml/pull/3959)) [@lowener](https://github.com/lowener)
+- Remove unused code in UMAP. ([#3931](https://github.com/rapidsai/cuml/pull/3931)) [@trivialfis](https://github.com/trivialfis)
+- Fix automerge #3900 and correct package versions in meta packages ([#3918](https://github.com/rapidsai/cuml/pull/3918)) [@dantegd](https://github.com/dantegd)
+- Adaptive stress tests when GPU memory capacity is insufficient ([#3916](https://github.com/rapidsai/cuml/pull/3916)) [@lowener](https://github.com/lowener)
+- Fix merge conflicts ([#3892](https://github.com/rapidsai/cuml/pull/3892)) [@ajschmidt8](https://github.com/ajschmidt8)
+- Remove old RF backend ([#3868](https://github.com/rapidsai/cuml/pull/3868)) [@RAMitchell](https://github.com/RAMitchell)
+- Refactor to extract random forest objectives ([#3854](https://github.com/rapidsai/cuml/pull/3854)) [@RAMitchell](https://github.com/RAMitchell)
 
 # cuML 21.06.00 (9 Jun 2021)
 

cpp/CMakeLists.txt

@@ -421,6 +421,8 @@ endif(BUILD_CUML_EXAMPLES)
 # ###################################################################################################
 # # - install targets -------------------------------------------------------------------------------
 
+include(CPack)
+
 set(CUML_TARGETS ${CUML_CPP_TARGET})
 
 if(BUILD_CUML_C_LIBRARY)