Merging

Signed-off-by: Adam Li <adam2392@gmail.com>

Author: adam2392

.circleci/config.yml

@@ -24,9 +24,6 @@ jobs:
       - OPENBLAS_NUM_THREADS: 2
       - CONDA_ENV_NAME: testenv
       - LOCK_FILE: build_tools/circle/doc_min_dependencies_linux-64_conda.lock
-      # Sphinx race condition in doc-min-dependencies is causing job to stall
-      # Here we run the job serially
-      - SPHINX_NUMJOBS: 1
     steps:
       - checkout
       - run: ./build_tools/circle/checkout_merge_commit.sh
@@ -61,8 +58,6 @@ jobs:
       - OPENBLAS_NUM_THREADS: 2
       - CONDA_ENV_NAME: testenv
       - LOCK_FILE: build_tools/circle/doc_linux-64_conda.lock
-      # Disable sphinx parallelism to avoid EOFError or job stalling in CircleCI
-      - SPHINX_NUMJOBS: 1
     steps:
       - checkout
       - run: ./build_tools/circle/checkout_merge_commit.sh

build_tools/azure/pylatest_conda_forge_mkl_linux-64_conda.lock

@@ -20,3 +20,6 @@ dependencies:
   - pytest-cov
   - coverage
   - ccache
+  - pytorch=1.13
+  - pytorch-cpu
+  - array-api-compat

build_tools/azure/pylatest_conda_forge_mkl_linux-64_environment.yml

@@ -88,9 +88,15 @@ def remove_from(alist, to_remove):
         "folder": "build_tools/azure",
         "platform": "linux-64",
         "channel": "conda-forge",
-        "conda_dependencies": common_dependencies + ["ccache"],
+        "conda_dependencies": common_dependencies + [
+            "ccache",
+            "pytorch",
+            "pytorch-cpu",
+            "array-api-compat",
+        ],
         "package_constraints": {
             "blas": "[build=mkl]",
+            "pytorch": "1.13",
         },
     },
     {

build_tools/update_environments_and_lock_files.py

@@ -7,18 +7,25 @@ SPHINXBUILD  ?= sphinx-build
 PAPER         =
 BUILDDIR      = _build
 
-# Run sequential by default, unless SPHINX_NUMJOBS is set.
-SPHINX_NUMJOBS ?= 1
-
 ifneq ($(EXAMPLES_PATTERN),)
     EXAMPLES_PATTERN_OPTS := -D sphinx_gallery_conf.filename_pattern="$(EXAMPLES_PATTERN)"
 endif
 
+ifeq ($(CI), true)
+    # On CircleCI using -j2 does not seem to speed up the html-noplot build
+    SPHINX_NUMJOBS_NOPLOT_DEFAULT=1
+else ($(shell uname), Darwin)
+    # Avoid stalling issues on MacOS
+    SPHINX_NUMJOBS_NOPLOT_DEFAULT=1
+else
+    SPHINX_NUMJOBS_NOPLOT_DEFAULT=auto
+endif
+
 # Internal variables.
 PAPEROPT_a4     = -D latex_paper_size=a4
 PAPEROPT_letter = -D latex_paper_size=letter
 ALLSPHINXOPTS   = -T -d $(BUILDDIR)/doctrees $(PAPEROPT_$(PAPER)) $(SPHINXOPTS)\
-    -j$(SPHINX_NUMJOBS) $(EXAMPLES_PATTERN_OPTS) .
+    $(EXAMPLES_PATTERN_OPTS) .
 
 
 .PHONY: help clean html dirhtml ziphtml pickle json latex latexpdf changes linkcheck doctest optipng
@@ -44,19 +51,27 @@ clean:
 	-rm -rf generated/*
 	-rm -rf modules/generated/
 
+# Default to SPHINX_NUMJOBS=1 for full documentation build. Using
+# SPHINX_NUMJOBS!=1 may actually slow down the build, or cause weird issues in
+# the CI (job stalling or EOFError), see
+# https://github.com/scikit-learn/scikit-learn/pull/25836 or
+# https://github.com/scikit-learn/scikit-learn/pull/25809
+html: SPHINX_NUMJOBS ?= 1
 html:
 	# These two lines make the build a bit more lengthy, and the
 	# the embedding of images more robust
 	rm -rf $(BUILDDIR)/html/_images
 	#rm -rf _build/doctrees/
-	$(SPHINXBUILD) -b html $(ALLSPHINXOPTS) $(BUILDDIR)/html/stable
+	$(SPHINXBUILD) -b html $(ALLSPHINXOPTS) -j$(SPHINX_NUMJOBS) $(BUILDDIR)/html/stable
 	@echo
 	@echo "Build finished. The HTML pages are in $(BUILDDIR)/html/stable"
 
-# rm $(BUILDDIR)/html/stable/index.html
-# mv $(BUILDDIR)/html/stable/fork_index.html $(BUILDDIR)/html/stable/index.html
+# Default to SPHINX_NUMJOBS=auto (except on MacOS and CI) since this makes
+# html-noplot build faster
+html-noplot: SPHINX_NUMJOBS ?= $(SPHINX_NUMJOBS_NOPLOT_DEFAULT)
 html-noplot:
-	$(SPHINXBUILD) -D plot_gallery=0 -b html $(ALLSPHINXOPTS) $(BUILDDIR)/html/stable
+	$(SPHINXBUILD) -D plot_gallery=0 -b html $(ALLSPHINXOPTS) -j$(SPHINX_NUMJOBS) \
+    $(BUILDDIR)/html/stable
 	@echo
 	@echo "Build finished. The HTML pages are in $(BUILDDIR)/html/stable."
 

doc/Makefile

@@ -0,0 +1,10 @@
+{
+  "jupyter-lite-schema-version": 0,
+  "jupyter-config-data": {
+    "litePluginSettings": {
+      "@jupyterlite/pyodide-kernel-extension:kernel": {
+        "pyodideUrl": "https://cdn.jsdelivr.net/pyodide/v0.23.1/full/pyodide.js"
+      }
+    }
+  }
+} 

doc/jupyter-lite.json

@@ -0,0 +1,5 @@
+{
+  "LiteBuildConfig": {
+    "no_sourcemaps": true
+  }
+}

doc/jupyter_lite_config.json

@@ -12,6 +12,8 @@ Array API support (experimental)
 
 The `Array API <https://data-apis.org/array-api/latest/>`_ specification defines
 a standard API for all array manipulation libraries with a NumPy-like API.
+Scikit-learn's Array API support requires
+`array-api-compat <https://github.com/data-apis/array-api-compat>`__ to be installed.
 
 Some scikit-learn estimators that primarily rely on NumPy (as opposed to using
 Cython) to implement the algorithmic logic of their `fit`, `predict` or
@@ -23,8 +25,8 @@ At this stage, this support is **considered experimental** and must be enabled
 explicitly as explained in the following.
 
 .. note::
-    Currently, only `cupy.array_api` and `numpy.array_api` are known to work
-    with scikit-learn's estimators.
+    Currently, only `cupy.array_api`, `numpy.array_api`, `cupy`, and `PyTorch`
+    are known to work with scikit-learn's estimators.
 
 Example usage
 =============
@@ -36,11 +38,11 @@ Here is an example code snippet to demonstrate how to use `CuPy
     >>> from sklearn.datasets import make_classification
     >>> from sklearn import config_context
     >>> from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
-    >>> import cupy.array_api as xp
+    >>> import cupy
 
     >>> X_np, y_np = make_classification(random_state=0)
-    >>> X_cu = xp.asarray(X_np)
-    >>> y_cu = xp.asarray(y_np)
+    >>> X_cu = cupy.asarray(X_np)
+    >>> y_cu = cupy.asarray(y_np)
     >>> X_cu.device
     <CUDA Device 0>
 
@@ -57,12 +59,30 @@ GPU. We provide a experimental `_estimator_with_converted_arrays` utility that
 transfers an estimator attributes from Array API to a ndarray::
 
     >>> from sklearn.utils._array_api import _estimator_with_converted_arrays
-    >>> cupy_to_ndarray = lambda array : array._array.get()
+    >>> cupy_to_ndarray = lambda array : array.get()
     >>> lda_np = _estimator_with_converted_arrays(lda, cupy_to_ndarray)
     >>> X_trans = lda_np.transform(X_np)
     >>> type(X_trans)
     <class 'numpy.ndarray'>
 
+PyTorch Support
+---------------
+
+PyTorch Tensors are supported by setting `array_api_dispatch=True` and passing in
+the tensors directly::
+
+    >>> import torch
+    >>> X_torch = torch.asarray(X_np, device="cuda", dtype=torch.float32)
+    >>> y_torch = torch.asarray(y_np, device="cuda", dtype=torch.float32)
+
+    >>> with config_context(array_api_dispatch=True):
+    ...     lda = LinearDiscriminantAnalysis()
+    ...     X_trans = lda.fit_transform(X_torch, y_torch)
+    >>> type(X_trans)
+    <class 'torch.Tensor'>
+    >>> X_trans.device.type
+    'cuda'
+
 .. _array_api_estimators:
 
 Estimators with support for `Array API`-compatible inputs

doc/modules/array_api.rst

@@ -230,6 +230,13 @@ Changelog
   :class:`decomposition.MiniBatchNMF` which can produce different results than previous
   versions. :pr:`25438` by :user:`Yotam Avidar-Constantini <yotamcons>`.
 
+:mod:`sklearn.discriminant_analysis`
+....................................
+
+- |Enhancement| :class:`discriminant_analysis.LinearDiscriminantAnalysis` now
+  supports the `PyTorch <https://pytorch.org/>`__. See
+  :ref:`array_api` for more details. :pr:`25956` by `Thomas Fan`_.
+
 :mod:`sklearn.ensemble`
 .......................
 

doc/whats_new/v1.3.rst

@@ -95,7 +95,7 @@
 score_label = f"Score on original\ndata: {score_iris:.2f}\n(p-value: {pvalue_iris:.3f})"
 ax.text(0.7, 10, score_label, fontsize=12)
 ax.set_xlabel("Accuracy score")
-_ = ax.set_ylabel("Probability")
+_ = ax.set_ylabel("Probability density")
 
 # %%
 # Random data
@@ -116,7 +116,7 @@
 score_label = f"Score on original\ndata: {score_rand:.2f}\n(p-value: {pvalue_rand:.3f})"
 ax.text(0.14, 7.5, score_label, fontsize=12)
 ax.set_xlabel("Accuracy score")
-ax.set_ylabel("Probability")
+ax.set_ylabel("Probability density")
 plt.show()
 
 # %%