Solve forward merger conflicts of #6313

CHANGELOG.md

@@ -1,3 +1,77 @@
+# cuml 25.02.00 (13 Feb 2025)
+
+## 🚨 Breaking Changes
+
+- update pip devcontainers to UCX 1.18 ([#6249](https://github.com/rapidsai/cuml/pull/6249)) [@jameslamb](https://github.com/jameslamb)
+
+## 🐛 Bug Fixes
+
+- Fix dask logistic regression segfault/hang due to wrongly typed variable ([#6281](https://github.com/rapidsai/cuml/pull/6281)) [@dantegd](https://github.com/dantegd)
+- Log UMAP arrays at trace verbosity level. ([#6274](https://github.com/rapidsai/cuml/pull/6274)) [@csadorf](https://github.com/csadorf)
+- Ensure all method signatures are sklearn compatible ([#6260](https://github.com/rapidsai/cuml/pull/6260)) [@jcrist](https://github.com/jcrist)
+- Fix UMAP transform illegal memory access error when data_on_host=True ([#6259](https://github.com/rapidsai/cuml/pull/6259)) [@csadorf](https://github.com/csadorf)
+- Ignore cudf's __dataframe__ deprecation. ([#6229](https://github.com/rapidsai/cuml/pull/6229)) [@bdice](https://github.com/bdice)
+- Fix for porter stemmer for cuDF change and ARIMA pytest adjustments ([#6227](https://github.com/rapidsai/cuml/pull/6227)) [@dantegd](https://github.com/dantegd)
+- Avoid duplicate log entries ([#6222](https://github.com/rapidsai/cuml/pull/6222)) [@jcrist](https://github.com/jcrist)
+- Further fixes for Scipy 1.15 update for PR and nightly CI ([#6213](https://github.com/rapidsai/cuml/pull/6213)) [@dantegd](https://github.com/dantegd)
+- Update setulb call from Scipy for updated signature of 1.15 ([#6207](https://github.com/rapidsai/cuml/pull/6207)) [@dantegd](https://github.com/dantegd)
+- Adjust test_kmeans to avoid false positive failures ([#6193](https://github.com/rapidsai/cuml/pull/6193)) [@dantegd](https://github.com/dantegd)
+- Adjust margin of logistic regression `log_proba` pytest to avoid false positive failures ([#6188](https://github.com/rapidsai/cuml/pull/6188)) [@dantegd](https://github.com/dantegd)
+- Skip flaky test of kernel_density in nightly job of CUDA 12.0.1 ([#6184](https://github.com/rapidsai/cuml/pull/6184)) [@dantegd](https://github.com/dantegd)
+- Try to reduce network usage in cuML tests. ([#6174](https://github.com/rapidsai/cuml/pull/6174)) [@bdice](https://github.com/bdice)
+- cuML dask fixes to unblock CI ([#6170](https://github.com/rapidsai/cuml/pull/6170)) [@dantegd](https://github.com/dantegd)
+- Remap BATCH_TREE_REORG to TREE_REORG in FIL ([#6161](https://github.com/rapidsai/cuml/pull/6161)) [@wphicks](https://github.com/wphicks)
+
+## 📖 Documentation
+
+- Fixes github link from pyx files ([#6202](https://github.com/rapidsai/cuml/pull/6202)) [@thomasjpfan](https://github.com/thomasjpfan)
+
+## 🚀 New Features
+
+- Allow CUDA ODR violations in 25.02 ([#6264](https://github.com/rapidsai/cuml/pull/6264)) [@robertmaynard](https://github.com/robertmaynard)
+- Define block size for sm_120 ([#6250](https://github.com/rapidsai/cuml/pull/6250)) [@robertmaynard](https://github.com/robertmaynard)
+- Support `alpha=0` in `Ridge` ([#6236](https://github.com/rapidsai/cuml/pull/6236)) [@jcrist](https://github.com/jcrist)
+- Add `as_sklearn` and `from_sklearn` APIs to serialize to CPU sklearn-estimators for supported models ([#6102](https://github.com/rapidsai/cuml/pull/6102)) [@dantegd](https://github.com/dantegd)
+
+## 🛠️ Improvements
+
+- Use `rapids-pip-retry` in CI jobs that might need retries ([#6293](https://github.com/rapidsai/cuml/pull/6293)) [@gforsyth](https://github.com/gforsyth)
+- Avoid large device allocation in UMAP with nndescent ([#6292](https://github.com/rapidsai/cuml/pull/6292)) [@jcrist](https://github.com/jcrist)
+- Revert CUDA 12.8 shared workflow branch changes ([#6282](https://github.com/rapidsai/cuml/pull/6282)) [@vyasr](https://github.com/vyasr)
+- Build and test with CUDA 12.8.0 ([#6266](https://github.com/rapidsai/cuml/pull/6266)) [@bdice](https://github.com/bdice)
+- update pip devcontainers to UCX 1.18 ([#6249](https://github.com/rapidsai/cuml/pull/6249)) [@jameslamb](https://github.com/jameslamb)
+- Drop deprecated thrust features and replace with libcu++ ones ([#6248](https://github.com/rapidsai/cuml/pull/6248)) [@miscco](https://github.com/miscco)
+- Add upper bound to prevent usage of numba 0.61.0 ([#6244](https://github.com/rapidsai/cuml/pull/6244)) [@galipremsagar](https://github.com/galipremsagar)
+- Normalize whitespace ([#6238](https://github.com/rapidsai/cuml/pull/6238)) [@bdice](https://github.com/bdice)
+- Rename cpp/test to cpp/tests. ([#6237](https://github.com/rapidsai/cuml/pull/6237)) [@bdice](https://github.com/bdice)
+- Use cuda.bindings layout. ([#6233](https://github.com/rapidsai/cuml/pull/6233)) [@bdice](https://github.com/bdice)
+- Skip dispatching to GPU for unimplemented metrics in UMAP ([#6224](https://github.com/rapidsai/cuml/pull/6224)) [@betatim](https://github.com/betatim)
+- Use GCC 13 in CUDA 12 conda builds. ([#6221](https://github.com/rapidsai/cuml/pull/6221)) [@bdice](https://github.com/bdice)
+- declare cuda-python dependency for wheels, other small packaging changes ([#6217](https://github.com/rapidsai/cuml/pull/6217)) [@jameslamb](https://github.com/jameslamb)
+- Bump Treelite to 4.4.1 ([#6212](https://github.com/rapidsai/cuml/pull/6212)) [@hcho3](https://github.com/hcho3)
+- Support raft's logger targets ([#6208](https://github.com/rapidsai/cuml/pull/6208)) [@vyasr](https://github.com/vyasr)
+- Use rapids-cmake for the logger ([#6205](https://github.com/rapidsai/cuml/pull/6205)) [@vyasr](https://github.com/vyasr)
+- consolidate pytest config in pyproject.toml ([#6201](https://github.com/rapidsai/cuml/pull/6201)) [@jameslamb](https://github.com/jameslamb)
+- introduce libcuml wheels ([#6199](https://github.com/rapidsai/cuml/pull/6199)) [@jameslamb](https://github.com/jameslamb)
+- Check if nightlies have succeeded recently enough ([#6196](https://github.com/rapidsai/cuml/pull/6196)) [@vyasr](https://github.com/vyasr)
+- Remove sphinx pinning ([#6195](https://github.com/rapidsai/cuml/pull/6195)) [@vyasr](https://github.com/vyasr)
+- simplify wheel CI scripts, other small packaging changes ([#6190](https://github.com/rapidsai/cuml/pull/6190)) [@jameslamb](https://github.com/jameslamb)
+- Update for raft logger changes ([#6187](https://github.com/rapidsai/cuml/pull/6187)) [@vyasr](https://github.com/vyasr)
+- update telemetry actions to fluent-bit friendly style ([#6186](https://github.com/rapidsai/cuml/pull/6186)) [@msarahan](https://github.com/msarahan)
+- HDBSCAN and KMeans API improvements for improving CPU interoperability ([#6181](https://github.com/rapidsai/cuml/pull/6181)) [@dantegd](https://github.com/dantegd)
+- CPU/GPU interop with RandomForest ([#6175](https://github.com/rapidsai/cuml/pull/6175)) [@hcho3](https://github.com/hcho3)
+- Update version references in workflow ([#6172](https://github.com/rapidsai/cuml/pull/6172)) [@AyodeAwe](https://github.com/AyodeAwe)
+- gate telemetry dispatch calls on TELEMETRY_ENABLED env var ([#6171](https://github.com/rapidsai/cuml/pull/6171)) [@msarahan](https://github.com/msarahan)
+- Use rapids-logger to generate the cuml logger ([#6162](https://github.com/rapidsai/cuml/pull/6162)) [@vyasr](https://github.com/vyasr)
+- Use estimator tags to improve sparse error handling ([#6151](https://github.com/rapidsai/cuml/pull/6151)) [@dantegd](https://github.com/dantegd)
+- FEA Add support for accepting a Numpy RandomState ([#6150](https://github.com/rapidsai/cuml/pull/6150)) [@betatim](https://github.com/betatim)
+- prefer system install of UCX in pip devcontainers, update outdated RAPIDS references ([#6149](https://github.com/rapidsai/cuml/pull/6149)) [@jameslamb](https://github.com/jameslamb)
+- Improve infrastructure for experimental dispatching of non existing methods in cuML ([#6148](https://github.com/rapidsai/cuml/pull/6148)) [@dantegd](https://github.com/dantegd)
+- Adapt to rmm logger changes ([#6147](https://github.com/rapidsai/cuml/pull/6147)) [@vyasr](https://github.com/vyasr)
+- Require approval to run CI on draft PRs ([#6145](https://github.com/rapidsai/cuml/pull/6145)) [@bdice](https://github.com/bdice)
+- Add breaking change workflow trigger ([#6130](https://github.com/rapidsai/cuml/pull/6130)) [@AyodeAwe](https://github.com/AyodeAwe)
+- Switch to native traceback in `cuml` ([#6078](https://github.com/rapidsai/cuml/pull/6078)) [@galipremsagar](https://github.com/galipremsagar)
+
 # cuml 24.12.00 (11 Dec 2024)
 
 ## 🚨 Breaking Changes

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -43,6 +43,7 @@ dependencies:
 - libcuvs==25.4.*,>=0.0.0a0
 - libraft==25.4.*,>=0.0.0a0
 - librmm==25.4.*,>=0.0.0a0
+- matplotlib
 - nbsphinx
 - ninja
 - nltk

conda/environments/all_cuda-128_arch-x86_64.yaml

@@ -40,6 +40,7 @@ dependencies:
 - libcuvs==25.4.*,>=0.0.0a0
 - libraft==25.4.*,>=0.0.0a0
 - librmm==25.4.*,>=0.0.0a0
+- matplotlib
 - nbsphinx
 - ninja
 - nltk

cpp/src/umap/fuzzy_simpl_set/naive.cuh

@@ -320,9 +320,9 @@ void launcher(nnz_t n,
   if (ML::default_logger().should_log(rapids_logger::level_enum::debug)) {
     CUML_LOG_DEBUG("Smooth kNN Distances");
     auto str = raft::arr2Str(sigmas.data(), 25, "sigmas", stream);
-    CUML_LOG_DEBUG("%s", str.c_str());
+    CUML_LOG_TRACE("%s", str.c_str());
     str = raft::arr2Str(rhos.data(), 25, "rhos", stream);
-    CUML_LOG_DEBUG("%s", str.c_str());
+    CUML_LOG_TRACE("%s", str.c_str());
   }
   */
 
@@ -353,7 +353,7 @@ void launcher(nnz_t n,
     CUML_LOG_DEBUG("Compute Membership Strength");
     std::stringstream ss;
     ss << in;
-    CUML_LOG_DEBUG(ss.str().c_str());
+    CUML_LOG_TRACE(ss.str().c_str());
   }
   */
 

cpp/src/umap/simpl_set_embed/algo.cuh

@@ -343,7 +343,7 @@ void launcher(
   if (ML::default_logger().should_log(rapids_logger::level_enum::debug)) {
     std::stringstream ss;
     ss << raft::arr2Str(epochs_per_sample.data(), out.nnz, "epochs_per_sample", stream);
-    CUML_LOG_DEBUG(ss.str().c_str());
+    CUML_LOG_TRACE(ss.str().c_str());
   }
   */
 

cpp/src/umap/supervised.cuh

@@ -307,10 +307,10 @@ void perform_general_intersection(const raft::handle_t& handle,
     std::stringstream ss1, ss2;
     ss1 << raft::arr2Str(
       y_knn_indices.data(), rgraph_coo->n_rows * params->target_n_neighbors, "knn_indices", stream);
-    CUML_LOG_DEBUG("%s", ss1.str().c_str());
+    CUML_LOG_TRACE("%s", ss1.str().c_str());
     ss2 << raft::arr2Str(
       y_knn_dists.data(), rgraph_coo->n_rows * params->target_n_neighbors, "knn_dists", stream);
-    CUML_LOG_DEBUG("%s", ss2.str().c_str());
+    CUML_LOG_TRACE("%s", ss2.str().c_str());
   }
   */
 

dependencies.yaml

@@ -435,6 +435,7 @@ dependencies:
           - ipykernel
           - nbsphinx
           - numpydoc
+          - matplotlib
           # https://github.com/pydata/pydata-sphinx-theme/issues/1539
           - pydata-sphinx-theme!=0.14.2
           - recommonmark

python/cuml/cuml/cluster/hdbscan/hdbscan.pyx

@@ -386,8 +386,6 @@ class HDBSCAN(UniversalBase, ClusterMixin, CMajorInputTagMixin):
         feature array. If metric is a string or callable, it must be one of
         the options allowed by metrics.pairwise.pairwise_distances for its
         metric parameter.
-        If metric is "precomputed", X is assumed to be a distance matrix and
-        must be square.
 
     p : int, optional (default=2)
         p value to use if using the minkowski metric.

python/cuml/cuml/ensemble/randomforest_common.pyx

@@ -110,8 +110,7 @@ class BaseRandomForestModel(UniversalBase):
                  criterion=None,
                  max_batch_size=4096, **kwargs):
 
-        sklearn_params = {"criterion": criterion,
-                          "min_weight_fraction_leaf": min_weight_fraction_leaf,
+        sklearn_params = {"min_weight_fraction_leaf": min_weight_fraction_leaf,
                           "max_leaf_nodes": max_leaf_nodes,
                           "min_impurity_split": min_impurity_split,
                           "oob_score": oob_score, "n_jobs": n_jobs,
@@ -169,6 +168,12 @@ class BaseRandomForestModel(UniversalBase):
                 " RandomForest split criterion"
             )
 
+        if self.split_criterion == MAE:
+            raise NotImplementedError(
+                "cuML does not currently support mean average error as a"
+                " RandomForest split criterion"
+            )
+
         self.min_samples_leaf = min_samples_leaf
         self.min_samples_split = min_samples_split
         self.min_impurity_decrease = min_impurity_decrease
@@ -369,10 +374,6 @@ class BaseRandomForestModel(UniversalBase):
                                       else None),
                     check_rows=self.n_rows, check_cols=1)
 
-        if self.dtype == np.float64:
-            warnings.warn("To use pickling first train using float32 data "
-                          "to fit the estimator")
-
         max_feature_val = self._get_max_feat_val()
         if isinstance(self.min_samples_leaf, float):
             self.min_samples_leaf = \

python/cuml/cuml/ensemble/randomforestclassifier.pyx

@@ -262,7 +262,7 @@ class RandomForestClassifier(BaseRandomForestModel,
 
     _hyperparam_interop_translator = {
         "criterion": {
-            "log_loss": "NotImplemented"
+            "log_loss": "NotImplemented",
         },
         "oob_score": {
             True: "NotImplemented",

python/cuml/cuml/ensemble/randomforestregressor.pyx

@@ -265,7 +265,7 @@ class RandomForestRegressor(BaseRandomForestModel,
     _hyperparam_interop_translator = {
         "criterion": {
             "friedman_mse": "NotImplemented",
-            "absolute_error": "NotImplemented"
+            "absolute_error": "NotImplemented",
         },
         "oob_score": {
             True: "NotImplemented",

python/cuml/cuml/experimental/accel/__main__.py

@@ -22,10 +22,6 @@
 import runpy
 import sys
 
-from cuml.internals import logger
-
-from . import install
-
 
 @click.command()
 @click.option("-m", "module", required=False, help="Module to run")
@@ -60,10 +56,32 @@
     count=True,
     help="Increase output verbosity (can be used multiple times, e.g. -vv). Default shows warnings only.",
 )
+@click.option(
+    "--cudf-pandas",
+    is_flag=True,
+    default=False,
+    help="Turn on cudf.pandas alongside cuml.accel.",
+)
 @click.argument("args", nargs=-1)
 def main(
-    module, convert_to_sklearn, format, output, disable_uvm, verbose, args
+    module,
+    convert_to_sklearn,
+    format,
+    output,
+    disable_uvm,
+    verbose,
+    cudf_pandas,
+    args,
 ):
+    if cudf_pandas:
+        import cudf.pandas
+
+        cudf.pandas.install()
+
+    # avoid importing cuML before cuDF.pandas
+    from cuml.internals import logger
+    from . import install
+
     default_logger_level_index = list(logger.level_enum).index(
         logger.level_enum.warn
     )