v21.08.02

v21.08.02

v21.08.01

v21.08.01

v21.08.00

🚨 Breaking Changes

• Remove deprecated target_weights in UMAP (#4081) @lowener
• Upgrade Treelite to 2.0.0 (#4072) @hcho3
• RF/DT cleanup (#4005) @venkywonka
• RF: memset and batch size optimization for computing splits (#4001) @venkywonka
• Remove old RF backend (#3868) @RAMitchell
• Enable warp-per-tree inference in FIL for regression and binary classification (#3760) @levsnv

🐛 Bug Fixes

• Disabling umap reproducibility tests for cuda 11.4 (#4128) @cjnolet
• Fix for crash in RF when max_leaves parameter is specified (#4126) @vinaydes
• Running umap mnmg test twice (#4112) @cjnolet
• Minimal fix for SparseRandomProjection (#4100) @viclafargue
• Creating copy of components in PCA transform and inverse transform (#4099) @divyegala
• Fix SVM model parameter handling in case n_support=0 (#4097) @tfeher
• Fix set_params for linear models (#4096) @lowener
• Fix train test split pytest comparison (#4062) @dantegd
• Fix fit_transform on KMeans (#4055) @lowener
• Fixing -1 key access in 1nn reduce op in HDBSCAN (#4052) @divyegala
• Disable installing gbench to avoid container permission issues (#4049) @dantegd
• Fix double fit crash in preprocessing models (#4040) @viclafargue
• Always add faiss library alias if it's missing (#4028) @trxcllnt
• Fixing intermittent HBDSCAN pytest failure in CI (#4025) @divyegala
• HDBSCAN bug on A100 (#4024) @divyegala
• Add treelite include paths to treelite targets (#4023) @trxcllnt
• Add Treelite_BINARY_DIR include to cuml++ build interface include paths (#4018) @trxcllnt
• Small ARIMA-related bug fixes in Hessenberg reduction and make_arima (#4017) @Nyrio
• Update setup.py (#4015) @ajschmidt8
• Update treelite version in get_treelite.cmake (#4014) @ajschmidt8
• Fix build with latest RAFT branch-21.08 (#4012) @trxcllnt
• Skipping hdbscan pytests when gpu is a100 (#4007) @cjnolet
• Using 64-bit array lengths to increase scale of pca & tsvd (#3983) @cjnolet
• Fix MNMG test in Dask RF (#3964) @hcho3
• Use nested include in destination of install headers to avoid docker permission issues (#3962) @dantegd
• Fix automerge #3939 (#3952) @dantegd
• Update UCX-Py version to 0.21 (#3950) @pentschev
• Fix kernel and line info in cmake (#3941) @dantegd
• Fix for multi GPU PCA compute failing bug after transform and added error handling when n_components is not passed (#3912) @akaanirban
• Tolerate QN linesearch failures when it's harmless (#3791) @achirkin

📖 Documentation

• Improve docstrings for silhouette score metrics. (#4026) @bdice
• Update CHANGELOG.md link (#3956) @Salonijain27
• Update documentation build examples to be generator agnostic (#3909) @robertmaynard
• Improve FIL code readability and documentation (#3056) @levsnv

🚀 New Features

• Add Multinomial and Bernoulli Naive Bayes variants (#4053) @lowener
• Add weighted K-Means sampling for SHAP (#4051) @Nanthini10
• Use chebyshev, canberra, hellinger and minkowski distance metrics (#3990) @mdoijade
• Implement vector leaf prediction for fil. (#3917) @RAMitchell
• change TargetEncoder's smooth argument from ratio to count (#3876) @daxiongshu
• Enable warp-per-tree inference in FIL for regression and binary classification (#3760) @levsnv

🛠️ Improvements

• Remove clang/clang-tools from conda recipe (#4109) @dantegd
• Pin dask version (#4108) @galipremsagar
• ANN warnings/tests updates (#4101) @viclafargue
• Removing local memory operations from computeSplitKernel and other optimizations (#4083) @vinaydes
• Fix libfaiss dependency to not expressly depend on conda-forge (#4082) @Ethyling
• Remove deprecated target_weights in UMAP (#4081) @lowener
• Upgrade Treelite to 2.0.0 (#4072) @hcho3
• Optimize dtype conversion for FIL (#4070) @dantegd
• Adding quick notes to HDBSCAN public API docs as to why discrepancies may occur between cpu and gpu impls. (#4061) @cjnolet
• Update conda environment name for CI (#4039) @ajschmidt8
• Rewrite random forest gtests (#4038) @RAMitchell
• Updating Clang Version to 11.0.0 (#4029) @codereport
• Raise ARIMA parameter limits from 4 to 8 (#4022) @Nyrio
• Testing extract clusters in HDBSCAN (#4009) @divyegala
• ARIMA - Kalman loop rewrite: single megakernel instead of host loop (#4006) @Nyrio
• RF/DT cleanup (#4005) @venkywonka
• Exposing condensed hierarchy through cython for easier unit-level testing (#4004) @cjnolet
• Use the 21.08 branch of rapids-cmake as rmm requires it (#4002) @robertmaynard
• RF: memset and batch size optimization for computing splits (#4001) @venkywonka
• Reducing cluster size to number of selected clusters. Returning stability scores (#3987) @cjnolet
• HDBSCAN: Lazy-loading (and caching) condensed & single-linkage tree objects (#3986) @cjnolet
• Fix 21.08 forward-merge conflicts (#3982) @ajschmidt8
• Update Dask/Distributed version (#3978) @pentschev
• Use clang-tools on x86 only (#3969) @jakirkham
• Promote trustworthiness_score to public header, add missing includes, update dependencies (#3968) @trxcllnt
• Moving FAISS ANN wrapper to raft (#3963) @cjnolet
• Add MG weighted k-means (#3959) @lowener
• Remove unused code in UMAP. (#3931) @trivialfis
• Fix automerge #3900 and correct package versions in meta packages (#3918) @dantegd
• Adaptive stress tests when GPU memory capacity is insufficient (#3916) @lowener
• Fix merge conflicts (#3892) @ajschmidt8
• Remove old RF backend (#3868) @RAMitchell
• Refactor to extract random forest objectives (#3854) @RAMitchell

v21.06.02

Release v21.06.02

v21.06.01

Release v21.06.01

v21.06.00

🚨 Breaking Changes

• Remove Base.enable_rmm_pool method as it is no longer needed (#3875) @teju85
• RF: Make experimental-backend default for regression tasks and deprecate old-backend. (#3872) @venkywonka
• Deterministic UMAP with floating point rounding. (#3848) @trivialfis
• Fix RF regression performance (#3845) @RAMitchell
• Add feature to print forest shape in FIL upon importing (#3763) @levsnv
• Remove 'seed' and 'output_type' deprecated features (#3739) @lowener

🐛 Bug Fixes

• Disable UMAP deterministic test on CTK11.2 (#3942) @trivialfis
• Revert #3869 (#3933) @hcho3
• RF: fix the bug in pdf_to_cdf device function that causes hang when n_bins > TPB && n_bins % TPB != 0 (#3921) @venkywonka
• Fix number of permutations in pytest and getting handle for cuml models (#3920) @dantegd
• Fix typo in umap target_weight parameter (#3914) @lowener
• correct compliation of cuml c library (#3908) @robertmaynard
• Correct install path for include folder to avoid double nesting (#3901) @dantegd
• Add type check for y in train_test_split (#3886) @Nanthini10
• Fix for MNMG test_rf_classification_dask_fil_predict_proba (#3831) @lowener
• Fix MNMG test test_rf_regression_dask_fil (#3830) @hcho3
• AgglomerativeClustering support single cluster and ignore only zero distances from self-loops (#3824) @cjnolet

📖 Documentation

• Small doc fixes for 21.06 release (#3936) @dantegd
• Document ability to export cuML RF to predict on other machines (#3890) @hcho3

🚀 New Features

• Deterministic UMAP with floating point rounding. (#3848) @trivialfis
• HDBSCAN (#3821) @cjnolet
• Add feature to print forest shape in FIL upon importing (#3763) @levsnv

🛠️ Improvements

• Pin dask ot 2021.5.1 for 21.06 release (#3937) @dantegd
• Upgrade xgboost to 1.4.2 (#3925) @dantegd
• Use UCX-Py 0.20 (#3911) @jakirkham
• Upgrade NCCL to 2.9.9 (#3902) @dantegd
• Update conda developer environments (#3898) @viclafargue
• ARIMA: pre-allocation of temporary memory to reduce latencies (#3895) @Nyrio
• Condense TSNE parameters into a struct (#3884) @lowener
• Update CHANGELOG.md links for calver (#3883) @ajschmidt8
• Make sure __init__ is called in graph callback. (#3881) @trivialfis
• Update docs build script (#3877) @ajschmidt8
• Remove Base.enable_rmm_pool method as it is no longer needed (#3875) @teju85
• RF: Make experimental-backend default for regression tasks and deprecate old-backend. (#3872) @venkywonka
• Enable probability output from RF binary classifier (alternative implementaton) (#3869) @hcho3
• CI test speed improvement (#3851) @lowener
• Fix RF regression performance (#3845) @RAMitchell
• Update to CMake 3.20 features, rapids-cmake and CPM (#3844) @dantegd
• Support sparse input features in QN solvers and Logistic Regression (#3827) @achirkin
• Trustworthiness score improvements (#3826) @viclafargue
• Performance optimization of RF split kernels by removing empty cycles (#3818) @vinaydes
• Correct deprecate positional args decorator for CalVer (#3784) @lowener
• ColumnTransformer & FunctionTransformer (#3745) @viclafargue
• Remove 'seed' and 'output_type' deprecated features (#3739) @lowener

v0.19.0

🚨 Breaking Changes

• Use the new RF backend by default for classification (#3686) @hcho3
• Deprecating quantile-per-tree and removing three previously deprecated Random Forest parameters (#3667) @vinaydes
• Update predict() / predict_proba() of RF to match sklearn (#3609) @hcho3
• Upgrade FAISS to 1.7.x (#3509) @viclafargue
• cuML's estimator Base class for preprocessing models (#3270) @viclafargue

🐛 Bug Fixes

• Fix brute force KNN distance metric issue (#3755) @viclafargue
• Fix min_max_axis (#3735) @viclafargue
• Fix NaN errors observed with ARIMA in CUDA 11.2 builds (#3730) @Nyrio
• Fix random state generator (#3716) @viclafargue
• Fixes the out of memory access issue for computeSplit kernels (#3715) @vinaydes
• Fixing umap gtest failure under cuda 11.2. (#3696) @cjnolet
• Fix irreproducibility issue in RF classification (#3693) @vinaydes
• BUG fix BatchedLevelAlgo DtClsTest & DtRegTest failing tests (#3690) @venkywonka
• Restore the functionality of RF score() (#3685) @hcho3
• Use main build.sh to build docs in docs CI (#3681) @dantegd
• Revert "Update conda recipes pinning of repo dependencies" (#3680) @raydouglass
• Skip tests that fail on CUDA 11.2 (#3679) @dantegd
• Dask KNN Cl&Re 1D labels (#3668) @viclafargue
• Update conda recipes pinning of repo dependencies (#3666) @mike-wendt
• OOB access in GLM SoftMax (#3642) @divyegala
• SilhouetteScore C++ tests seed (#3640) @divyegala
• SimpleImputer fix (#3624) @viclafargue
• Silhouette Score make_monotonic for non-monotonic label set (#3619) @divyegala
• Fixing support for empty rows in sparse Jaccard / Cosine (#3612) @cjnolet
• Fix train_test_split with stratify option (#3611) @Nanthini10
• Update predict() / predict_proba() of RF to match sklearn (#3609) @hcho3
• Change dask and distributed branch to main (#3593) @dantegd
• Fixes memory allocation for experimental backend and improves quantile computations (#3586) @vinaydes
• Add ucx-proc package back that got lost during an auto merge conflict (#3550) @dantegd
• Fix failing Hellinger gtest (#3549) @cjnolet
• Directly invoke make for non-CMake docs target (#3534) @wphicks
• Fix Codecov.io Coverage Upload for Branch Builds (#3524) @mdemoret-nv
• Ensure global_output_type is thread-safe (#3497) @wphicks
• List as input for SimpleImputer (#3489) @viclafargue

📖 Documentation

• Add sparse docstring comments (#3712) @JohnZed
• FIL and Dask demo (#3698) @miroenev
• Deprecating quantile-per-tree and removing three previously deprecated Random Forest parameters (#3667) @vinaydes
• Fixing Indentation for Docstring Generators (#3650) @mdemoret-nv
• Update doc to indicate ExtraTree support (#3635) @hcho3
• Update doc, now that FIL supports multi-class classification (#3634) @hcho3
• Document model_type='xgboost_json' in FIL (#3633) @hcho3
• Including log loss metric to the documentation website (#3617) @lowener
• Update the build doc regarding the use of GCC 7.5 (#3605) @hcho3
• Update One-Hot Encoder doc (#3600) @lowener
• Fix documentation of KMeans (#3595) @lowener

🚀 New Features

• Reduce the size of the cuml libraries (#3702) @robertmaynard
• Use ninja as default CMake generator (#3664) @wphicks
• Single-Linkage Hierarchical Clustering Python Wrapper (#3631) @cjnolet
• Support for precomputed distance matrix in DBSCAN (#3585) @Nyrio
• Adding haversine to brute force knn (#3579) @cjnolet
• Support for sample_weight parameter in LogisticRegression (#3572) @viclafargue
• Provide "--ccache" flag for build.sh (#3566) @wphicks
• Eliminate unnecessary includes discovered by cppclean (#3564) @wphicks
• Single-linkage Hierarchical Clustering C++ (#3545) @cjnolet
• Expose sparse distances via semiring to Python API (#3516) @lowener
• Use cmake --build in build.sh to facilitate switching build tools (#3487) @wphicks
• Add cython hinge_loss (#3409) @Nanthini10
• Adding CodeCov Info for Dask Tests (#3338) @mdemoret-nv
• Add predict_proba() to XGBoost-style models in FIL C++ (#2894) @levsnv

🛠️ Improvements

• Updating docs, readme, and umap param tests for 0.19 (#3731) @cjnolet
• Locking RAFT hash for 0.19 (#3721) @cjnolet
• Upgrade to Treelite 1.1.0 (#3708) @hcho3
• Update to XGBoost 1.4.0rc1 (#3699) @hcho3
• Use the new RF backend by default for classification (#3686) @hcho3
• Update LogisticRegression documentation (#3677) @viclafargue
• Preprocessing out of experimental (#3676) @viclafargue
• ENH Decision Tree new backend computeSplit*Kernel histogram calculation optimization (#3674) @venkywonka
• Remove check_cupy8 (#3669) @viclafargue
• Use custom conda build directory for ccache integration (#3658) @dillon-cullinan
• Disable three flaky tests (#3657) @hcho3
• CUDA 11.2 developer environment (#3648) @dantegd
• Store data frequencies in tree nodes of RF (#3647) @hcho3
• Row major Gram matrices (#3639) @tfeher
• Converting all Estimator Constructors to Keyword Arguments (#3636) @mdemoret-nv
• Adding make_pipeline + test score with pipeline (#3632) @viclafargue
• ENH Decision Tree new backend computeSplitClassificationKernel histogram calculation and occupancy optimization (#3616) @venkywonka
• Revert "ENH Fix stale GHA and prevent duplicates " (#3614) @mike-wendt
• ENH Fix stale GHA and prevent duplicates (#3613) @mike-wendt
• KNN from RAFT (#3603) @viclafargue
• Update Changelog Link (#3601) @ajschmidt8
• Move SHAP explainers out of experimental (#3596) @dantegd
• Fixing compatibility issue with CUDA array interface (#3594) @lowener
• Remove cutlass usage in row major input for euclidean exp/unexp, cosine and L1 distance matrix (#3589) @mdoijade
• Test FIL probabilities with absolute error thresholds in python (#3582) @levsnv
• Removing sparse prims and fused l2 nn prim from cuml (#3578) @cjnolet
• Prepare Changelog for Automation (#3570) @ajschmidt8
• Print debug message if SVM convergence is poor (#3562) @tfeher
• Fix merge conflicts in 3552 (#3557) @ajschmidt8
• Additional distance metrics for ANN (#3533) @viclafargue
• Improve warning message when QN solver reaches max_iter (#3515) @tfeher
• Fix merge conflicts in 3502 (#3513) @ajschmidt8
• Upgrade FAISS to 1.7.x (#3509) @viclafargue
• ENH Pass ccache variables to conda recipe & use Ninja in CI (#3508) @Ethyling
• Fix forward-merger conflicts in #3502 (#3506) @dantegd
• Sklearn meta-estimators into namespace (#3493) @viclafargue
• Add flexibility to copyright checker (#3466) @lowener
• Update sparse KNN to use rmm device buffer (#3460) @lowener
• Fix forward-merger conflicts in #3444 (#3455) @ajschmidt8
• Replace ML::MetricType with raft::distance::DistanceType (#3389) @lowener
• RF param initialization cython and C++ layer cleanup (#3358) @venkywonka
• MNMG RF broadcast feature (#3349) @viclafargue
• cuML's estimator Base class for preprocessing models (#3270) @viclafargue
• Make _get_tags a class/static method (#3257) @dantegd
• NVTX Markers for RF and RF-backend (#3014) @venkywonka

v0.18.0

Breaking Changes 🚨

• cuml.experimental SHAP improvements (#3433) @dantegd
• Enable feature sampling for the experimental backend of Random Forest (#3364) @vinaydes
• re-enable cuML's copyright checker script (#3363) @teju85
• Batched Silhouette Score (#3362) @divyegala
• Update failing MNMG tests (#3348) @viclafargue
• Rename print_summary() of Dask RF to get_summary_text(); it now returns string to the client (#3341) @hcho3
• Rename dump_as_json() -> get_json(); expose it from Dask RF (#3340) @hcho3
• MNMG KNN consolidation (#3307) @viclafargue
• Return confusion matrix as int unless float weights are used (#3275) @lowener
• Approximate Nearest Neighbors (#2780) @viclafargue

Bug Fixes 🐛

• HOTFIX Add ucx-proc package back that got lost during an auto merge conflict (#3551) @dantegd
• Non project-flash CI ml test 18.04 issue debugging and bugfixing (#3495) @dantegd
• Temporarily xfail KBinsDiscretizer uniform tests (#3494) @wphicks
• Fix illegal memory accesses when NITEMS > 1, and nrows % NITEMS != 0. (#3480) @canonizer
• Update call to dask client persist (#3474) @dantegd
• Adding warning for IVFPQ (#3472) @viclafargue
• Fix failing sparse NN test in CI by allowing small number of index discrepancies (#3454) @cjnolet
• Exempting thirdparty code from copyright checks (#3453) @lowener
• Relaxing Batched SilhouetteScore Test Constraint (#3452) @divyegala
• Mark kbinsdiscretizer quantile tests as xfail (#3450) @wphicks
• Fixing documentation on SimpleImputer (#3447) @lowener
• Skipping IVFPQ (#3429) @viclafargue
• Adding tol to dask test_kmeans (#3426) @lowener
• Fix memory bug for SVM with large n_rows (#3420) @tfeher
• Allow linear regression for with CUDA >=11.0 (#3417) @wphicks
• Fix vectorizer tests by restoring sort behavior in groupby (#3416) @JohnZed
• Ensure make_classification respects output type (#3415) @wphicks
• Clean Up #include Dependencies (#3402) @mdemoret-nv
• Fix Nearest Neighbor Stress Test (#3401) @lowener
• Fix array_equal in tests (#3400) @viclafargue
• Improving Copyright Check When Not Running in CI (#3398) @mdemoret-nv
• Also xfail zlib errors when downloading newsgroups data (#3393) @JohnZed
• Fix for ANN memory release bug (#3391) @viclafargue
• XFail Holt Winters test where statsmodels has known issues with gcc 9.3.0 (#3385) @JohnZed
• FIX Update cupy to >= 7.8 and remove unused build.sh script (#3378) @dantegd
• re-enable cuML's copyright checker script (#3363) @teju85
• Update failing MNMG tests (#3348) @viclafargue
• Rename print_summary() of Dask RF to get_summary_text(); it now returns string to the client (#3341) @hcho3
• Fixing make_blobs to Respect the Global Output Type (#3339) @mdemoret-nv
• Fix permutation explainer (#3332) @RAMitchell
• k-means bug fix in debug build (#3321) @akkamesh
• Fix for default arguments of PCA (#3320) @lowener
• Provide workaround for cupy.percentile bug (#3315) @wphicks
• Fix SVR unit test parameter (#3294) @tfeher
• Add xfail on fetching 20newsgroup dataset (test_naive_bayes) (#3291) @lowener
• Remove unused keyword in PorterStemmer code (#3289) @wphicks
• Remove static specifier in DecisionTree unit test for C++14 compliance (#3281) @wphicks
• Correct pure virtual declaration in manifold_inputs_t (#3279) @wphicks

Documentation 📖

• Correct import path in docs for experimental preprocessing features (#3488) @wphicks
• Minor doc updates for 0.18 (#3475) @JohnZed
• Improve Python Docs with Default Role (#3445) @mdemoret-nv
• Fixing Python Documentation Errors and Warnings (#3428) @mdemoret-nv
• Remove outdated references to changelog in CONTRIBUTING.md (#3328) @wphicks
• Adding highlighting to bibtex in readme (#3296) @cjnolet

New Features 🚀

• Improve runtime performance of RF to Treelite conversion (#3410) @wphicks
• Parallelize Treelite to FIL conversion over trees (#3396) @wphicks
• Parallelize RF to Treelite conversion over trees (#3395) @wphicks
• Allow saving Dask RandomForest models immediately after training (fixes #3331) (#3388) @jameslamb
• genetic programming initial structures (#3387) @teju85
• MNMG DBSCAN (#3382) @Nyrio
• FIL to use L1 cache when input columns don't fit into shared memory (#3370) @levsnv
• Enable feature sampling for the experimental backend of Random Forest (#3364) @vinaydes
• Batched Silhouette Score (#3362) @divyegala
• Rename dump_as_json() -> get_json(); expose it from Dask RF (#3340) @hcho3
• Exposing model_selection in a similar way to scikit-learn (#3329) @ptartan21
• Promote IncrementalPCA from experimental in 0.18 release (#3327) @lowener
• Create labeler.yml (#3324) @jolorunyomi
• Add slow high-precision mode to KNN (#3304) @wphicks
• Sparse TSNE (#3293) @divyegala
• Sparse Generalized SPMV (semiring) Primitive (#3146) @cjnolet
• Multiclass meta estimator wrappers and multiclass SVC (#3092) @tfeher
• Approximate Nearest Neighbors (#2780) @viclafargue
• Add KNN parameter to t-SNE (#2592) @aleksficek

Improvements 🛠️

• Update stale GHA with exemptions & new labels (#3507) @mike-wendt
• Add GHA to mark issues/prs as stale/rotten (#3500) @Ethyling
• Fix naive bayes inputs (#3448) @cjnolet
• Prepare Changelog for Automation (#3442) @ajschmidt8
• cuml.experimental SHAP improvements (#3433) @dantegd
• Speed up knn tests (#3411) @JohnZed
• Replacing sklearn functions with cuml in RF MNMG notebook (#3408) @lowener
• Auto-label PRs based on their content (#3407) @jolorunyomi
• Use stable 1.0.0 version of Treelite (#3394) @hcho3
• API update to match RAFT PR #120 (#3386) @drobison00
• Update linear models to use RMM memory allocation (#3365) @lowener
• Updating dense pairwise distance enum names (#3352) @cjnolet
• Upgrade Treelite module (#3316) @hcho3
• Removed FIL node types with _t suffix (#3314) @canonizer
• MNMG KNN consolidation (#3307) @viclafargue
• Updating PyTests to Stay Below 4 Gb Limit (#3306) @mdemoret-nv
• Refactoring: move internal FIL interface to a separate file (#3292) @canonizer
• Return confusion matrix as int unless float weights are used (#3275) @lowener
• 018 add unfitted error pca & tests on IPCA (#3272) @lowener
• Linear models predict function consolidation (#3256) @dantegd
• Preparing sparse primitives for movement to RAFT (#3157) @cjnolet

v0.17.0

v0.17.0 Release

v0.16.0

v0.16.0 Release