Merge pull request #19 from rapidsai/branch-0.17

sync with upstream

CHANGELOG.md

@@ -3,13 +3,15 @@
 ## New Features
 
 ## Improvements
+- PR #3077: Improve runtime for test_kmeans
 - PR #3070: Speed up dask/test_datasets tests
 - PR #3075: Speed up test_linear_model tests
 - PR #3078: Speed up test_incremental_pca tests
 - PR #2902: `matrix/matrix.cuh` in RAFT namespacing
 - PR #2903: Moving linalg's gemm, gemv, transpose to RAFT namespaces
 - PR #2905: `stats` prims `mean_center`, `sum` to RAFT namespaces
 - PR #2904: Moving `linalg` basic math ops to RAFT namespaces
+- PR #2956: Follow cuML array conventions in ARIMA and remove redundancy
 - PR #3000: Pin cmake policies to cmake 3.17 version, bump project version to 0.17
 - PR #3083: Improving test_make_blobs testing time
 - PR #2906: Moving `linalg` decomp to RAFT namespaces
@@ -18,8 +20,20 @@
 - PR #3044: Move leftover `linalg` and `stats` to RAFT namespaces
 - PR #3067: Deleting prims moved to RAFT and updating header paths
 - PR #3074: Reducing dask coordinate descent test runtime
+- PR #3096: Avoid memory transfers in CSR WeakCC for DBSCAN
+- PR #3088: More readable and robust FIL C++ test management
+- PR #3052: Speeding up MNMG KNN Cl&Re testing
+- PR #3115: Speeding up MNMG UMAP testing
+- PR #3112: Speed test_array
+- PR #3111: Adding Cython to Code Coverage
+- PR #3129:  Update notebooks README
+- PR #3040: Improved Array Conversion with CumlArrayDescriptor and Decorators
+- PR #3134: Improving the Deprecation Message Formatting in Documentation
 
 ## Bug Fixes
+- PR #3069: Prevent conversion of DataFrames to Series in preprocessing
+- PR #3065: Refactoring prims metrics function names from camelcase to underscore format
+- PR #3033: Splitting ml metrics to individual files
 - PR #3072: Fusing metrics and score directories in src_prims
 - PR #3037: Avoid logging deadlock in multi-threaded C code
 - PR #2983: Fix seeding of KISS99 RNG
@@ -28,10 +42,17 @@
 - PR #3012: Increasing learning rate for SGD log loss and invscaling pytests
 - PR #3021: Fix a hang in cuML RF experimental backend
 - PR #3039: Update RF and decision tree parameter initializations in benchmark codes
+- PR #3060: Speed up test suite `test_fil`
 - PR #3061: Handle C++ exception thrown from FIL predict
 - PR #3073: Update mathjax CDN URL for documentation
 - PR #3062: Bumping xgboost version to match cuml version
+- PR #3084: Fix artifacts in t-SNE results
 - PR #3086: Reverting FIL Notebook Testing
+- PR #3114: Fixed a typo in SVC's predict_proba AttributeError
+- PR #3117: Fix two crashes in experimental RF backend
+- PR #3119: Fix memset args for benchmark 
+- PR #3130: Return Python string from `dump_as_json()` of RF
+- PR #3136: Fix stochastic gradient descent example
 
 # cuML 0.16.0 (Date TBD)
 
@@ -90,6 +111,7 @@
 - PR #2928: Updating Estimators Derived from Base for Consistency
 - PR #2942: Adding `cuml.experimental` to the Docs
 - PR #3010: Improve gpuCI Scripts
+- PR #3141: Move DistanceType enum to RAFT
 
 ## Bug Fixes
 - PR #2973: Allow data imputation for nan values

build.sh

@@ -19,7 +19,7 @@ ARGS=$*
 REPODIR=$(cd $(dirname $0); pwd)
 
 VALIDTARGETS="clean libcuml cuml cpp-mgtests prims bench prims-bench cppdocs pydocs"
-VALIDFLAGS="-v -g -n --allgpuarch --buildfaiss --buildgtest --singlegpu --nvtx --show_depr_warn -h --help "
+VALIDFLAGS="-v -g -n --allgpuarch --buildfaiss --buildgtest --singlegpu --nvtx --show_depr_warn --codecov -h --help "
 VALIDARGS="${VALIDTARGETS} ${VALIDFLAGS}"
 HELP="$0 [<target> ...] [<flag> ...]
  where <target> is:
@@ -43,6 +43,8 @@ HELP="$0 [<target> ...] [<flag> ...]
    --singlegpu      - Build libcuml and cuml without multigpu components
    --nvtx           - Enable nvtx for profiling support
    --show_depr_warn - show cmake deprecation warnings
+   --codecov        - Enable code coverage support by compiling with Cython linetracing
+                      and profiling enabled (WARNING: Impacts performance)
    -h               - print this text
 
  default action (no args) is to build and install 'libcuml', 'cuml', and 'prims' targets only for the detected GPU arch
@@ -58,7 +60,7 @@ BUILD_TYPE=Release
 INSTALL_TARGET=install
 BUILD_ALL_GPU_ARCH=0
 SINGLEGPU_CPP_FLAG=""
-SINGLEGPU_PYTHON_FLAG=""
+BUILD_PYTHON_ARGS=${BUILD_PYTHON_ARGS:=""}
 NVTX=OFF
 CLEAN=0
 BUILD_DISABLE_DEPRECATION_WARNING=ON
@@ -115,7 +117,7 @@ if hasArg --allgpuarch; then
     BUILD_ALL_GPU_ARCH=1
 fi
 if hasArg --singlegpu; then
-    SINGLEGPU_PYTHON_FLAG="--singlegpu"
+    BUILD_PYTHON_ARGS="${BUILD_PYTHON_ARGS} --singlegpu"
     SINGLEGPU_CPP_FLAG=ON
 fi
 if hasArg cpp-mgtests; then
@@ -133,6 +135,9 @@ fi
 if hasArg --show_depr_warn; then
     BUILD_DISABLE_DEPRECATION_WARNING=OFF
 fi
+if hasArg --codecov; then
+    BUILD_PYTHON_ARGS="${BUILD_PYTHON_ARGS} --linetrace=1 --profile"
+fi
 if hasArg clean; then
     CLEAN=1
 fi
@@ -224,9 +229,9 @@ fi
 if completeBuild || hasArg cuml || hasArg pydocs; then
     cd ${REPODIR}/python
     if [[ ${INSTALL_TARGET} != "" ]]; then
-        python setup.py build_ext -j${PARALLEL_LEVEL:-1} ${SINGLEGPU_PYTHON_FLAG} --library-dir=${LIBCUML_BUILD_DIR} install --single-version-externally-managed --record=record.txt
+        python setup.py build_ext -j${PARALLEL_LEVEL:-1} ${BUILD_PYTHON_ARGS} --library-dir=${LIBCUML_BUILD_DIR} install --single-version-externally-managed --record=record.txt
     else
-        python setup.py build_ext -j${PARALLEL_LEVEL:-1} --library-dir=${LIBCUML_BUILD_DIR} ${SINGLEGPU_PYTHON_FLAG}
+        python setup.py build_ext -j${PARALLEL_LEVEL:-1} ${BUILD_PYTHON_ARGS} --library-dir=${LIBCUML_BUILD_DIR}
     fi
 
     if hasArg pydocs; then

ci/gpu/build.sh

@@ -98,7 +98,7 @@ if [[ -z "$PROJECT_FLASH" || "$PROJECT_FLASH" == "0" ]]; then
     ################################################################################
 
     gpuci_logger "Build from source"
-    $WORKSPACE/build.sh clean libcuml cuml prims bench -v
+    $WORKSPACE/build.sh clean libcuml cuml prims bench -v --codecov
 
     gpuci_logger "Resetting LD_LIBRARY_PATH"
 
@@ -190,7 +190,7 @@ else
     conda install -c $WORKSPACE/ci/artifacts/cuml/cpu/conda-bld/ libcuml
 
     gpuci_logger "Building cuml"
-    "$WORKSPACE/build.sh" -v cuml
+    "$WORKSPACE/build.sh" -v cuml --codecov
 
     gpuci_logger "Python pytest for cuml"
     cd $WORKSPACE/python

cpp/CMakeLists.txt

@@ -401,8 +401,19 @@ if(BUILD_CUML_CPP_LIBRARY)
     src/holtwinters/holtwinters.cu
     src/kmeans/kmeans.cu
     src/knn/knn.cu
-    src/metrics/metrics.cu
+    src/metrics/accuracy_score.cu
+    src/metrics/adjusted_rand_index.cu
+    src/metrics/completeness_score.cu
+    src/metrics/entropy.cu
+    src/metrics/homogeneity_score.cu
+    src/metrics/kl_divergence.cu
+    src/metrics/mutual_info_score.cu
+    src/metrics/pairwise_distance.cu
+    src/metrics/r2_score.cu
+    src/metrics/rand_index.cu
+    src/metrics/silhouette_score.cu
     src/metrics/trustworthiness.cu
+    src/metrics/v_measure.cu
     src/pca/pca.cu
     src/randomforest/randomforest.cu
     src/random_projection/rproj.cu

cpp/bench/common/ml_benchmark.hpp

@@ -55,14 +55,14 @@ struct CudaEventTimer {
     CUDA_CHECK(cudaEventCreate(&stop));
     // flush L2?
     if (ptr != nullptr && l2CacheSize > 0) {
-      CUDA_CHECK(cudaMemsetAsync(ptr, sizeof(char) * l2CacheSize, 0, s));
+      CUDA_CHECK(cudaMemsetAsync(ptr, 0, sizeof(char) * l2CacheSize, s));
       CUDA_CHECK(cudaStreamSynchronize(stream));
     }
     CUDA_CHECK(cudaEventRecord(start, stream));
   }
   CudaEventTimer() = delete;
 
-  /** 
+  /**
    * @brief The dtor stops the timer and performs a synchroniazation. Time of
    *       the benchmark::State object provided to the ctor will be set to the
    *       value given by `cudaEventElapsedTime()`.

cpp/bench/prims/distance_common.cuh

@@ -26,7 +26,7 @@ struct Params {
   int m, n, k;
 };  // struct Params
 
-template <typename T, ML::Distance::DistanceType DType>
+template <typename T, raft::distance::DistanceType DType>
 struct Distance : public Fixture {
   Distance(const std::string& name, const Params& p)
     : Fixture(name, std::shared_ptr<deviceAllocator>(

cpp/bench/prims/distance_cosine.cu

@@ -21,7 +21,7 @@ namespace Bench {
 namespace Distance {
 
 DIST_BENCH_REGISTER(DistanceCosine,
-                    ML::Distance::DistanceType::EucExpandedCosine);
+                    raft::distance::DistanceType::EucExpandedCosine);
 
 }  // namespace Distance
 }  // namespace Bench

cpp/bench/prims/distance_exp_l2.cu

@@ -20,9 +20,9 @@ namespace MLCommon {
 namespace Bench {
 namespace Distance {
 
-DIST_BENCH_REGISTER(DistanceL2Sq, ML::Distance::DistanceType::EucExpandedL2);
+DIST_BENCH_REGISTER(DistanceL2Sq, raft::distance::DistanceType::EucExpandedL2);
 DIST_BENCH_REGISTER(DistanceL2Sqrt,
-                    ML::Distance::DistanceType::EucExpandedL2Sqrt);
+                    raft::distance::DistanceType::EucExpandedL2Sqrt);
 
 }  // namespace Distance
 }  // namespace Bench

cpp/bench/prims/distance_l1.cu

@@ -20,7 +20,7 @@ namespace MLCommon {
 namespace Bench {
 namespace Distance {
 
-DIST_BENCH_REGISTER(DistanceL1, ML::Distance::DistanceType::EucUnexpandedL1);
+DIST_BENCH_REGISTER(DistanceL1, raft::distance::DistanceType::EucUnexpandedL1);
 
 }  // namespace Distance
 }  // namespace Bench

cpp/bench/prims/distance_unexp_l2.cu

@@ -21,9 +21,9 @@ namespace Bench {
 namespace Distance {
 
 DIST_BENCH_REGISTER(DistanceUnexpL2Sq,
-                    ML::Distance::DistanceType::EucUnexpandedL2);
+                    raft::distance::DistanceType::EucUnexpandedL2);
 DIST_BENCH_REGISTER(DistanceUnexpL2Sqrt,
-                    ML::Distance::DistanceType::EucUnexpandedL2Sqrt);
+                    raft::distance::DistanceType::EucUnexpandedL2Sqrt);
 
 }  // namespace Distance
 }  // namespace Bench

cpp/include/cuml/cluster/kmeans.hpp

@@ -53,7 +53,7 @@ struct KMeansParams {
   int seed = 0;
 
   // Metric to use for distance computation. Any metric from
-  // ML::Distance::DistanceType can be used
+  // raft::distance::DistanceType can be used
   int metric = 0;
 
   // Number of instance k-means algorithm will be run with different seeds.
@@ -184,7 +184,7 @@ void predict(const raft::handle_t &handle, const KMeansParams &params,
  * sample in 'X' (it should be same as the dimension for each cluster centers in
  * 'centroids').
  * @param[in]     metric        Metric to use for distance computation. Any
- * metric from ML::Distance::DistanceType can be used
+ * metric from raft::distance::DistanceType can be used
  * @param[out]    X_new         X transformed in the new space..
  */
 void transform(const raft::handle_t &handle, const KMeansParams &params,

cpp/include/cuml/fil/fil.h

@@ -71,6 +71,9 @@ enum output_t {
   /** output class label: either apply threshold to the output of the previous stage (for binary classification),
       or select the class with the most votes to get the class label (for multi-class classification).  */
   CLASS = 0x100,
+  SIGMOID_CLASS = SIGMOID | CLASS,
+  AVG_CLASS = AVG | CLASS,
+  AVG_SIGMOID_CLASS = AVG | SIGMOID | CLASS,
 };
 
 /** storage_type_t defines whether to import the forests as dense or sparse */