add notebook tests, build.sh args (#62)

Closes #53 This pulls over the remaining changes from that PR... thought that was easier than fixing the merge conflicts between that and recent PRs. * adds notebook testing * adds all the remaining `build.sh` changes from #53 (mostly for interactive development) * adds `cugraph-pyg` and `cugraph-dgl` tests in `conda-python-tests` (missed in #59) ## Notes for Reviewers I left other comments explaining these changes inline. Authors: - James Lamb (https://github.com/jameslamb) - Tingyu Wang (https://github.com/tingyu66) Approvers: - Tingyu Wang (https://github.com/tingyu66) - Ray Douglass (https://github.com/raydouglass) URL: #62
rapidsai · Nov 8, 2024 · d260ccb · d260ccb
1 parent f7ab898
commit d260ccb
Show file tree

Hide file tree

Showing 17 changed files with 323 additions and 28 deletions.
diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml
@@ -18,6 +18,7 @@ jobs:
       - conda-cpp-tests
       - conda-python-build
       - conda-python-tests
+      - conda-notebook-tests
       - wheel-build-pylibwholegraph
       - wheel-tests-pylibwholegraph
       - wheel-build-cugraph-dgl
@@ -51,6 +52,7 @@ jobs:
           - '!CONTRIBUTING.md'
           - '!README.md'
           - '!docs/**'
+          - '!readme_pages/**'
         test_python:
           - '**'
           - '!.devcontainers/**'
@@ -59,6 +61,7 @@ jobs:
           - '!docs/**'
           - '!img/**'
           - '!notebooks/**'
+          - '!readme_pages/**'
   checks:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]
@@ -83,6 +86,17 @@ jobs:
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]
     with:
       build_type: pull-request
+  conda-notebook-tests:
+    needs: [conda-python-build, changed-files]
+    secrets: inherit
+    uses: rapidsai/shared-workflows/.github/workflows/[email protected]
+    if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_notebooks
+    with:
+      build_type: pull-request
+      node_type: "gpu-v100-latest-1"
+      arch: "amd64"
+      container_image: "rapidsai/ci-conda:cuda11.8.0-ubuntu22.04-py3.12"
+      run_script: "ci/test_notebooks.sh"
   conda-python-tests:
     needs: [conda-python-build, changed-files]
     secrets: inherit

diff --git a/.github/workflows/test.yaml b/.github/workflows/test.yaml
@@ -22,6 +22,16 @@ jobs:
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
+  conda-notebook-tests:
+    secrets: inherit
+    uses: rapidsai/shared-workflows/.github/workflows/[email protected]
+    if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_notebooks
+    with:
+      build_type: pull-request
+      node_type: "gpu-v100-latest-1"
+      arch: "amd64"
+      container_image: "rapidsai/ci-conda:cuda11.8.0-ubuntu22.04-py3.12"
+      run_script: "ci/test_notebooks.sh"
   conda-python-tests:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]

diff --git a/.gitignore b/.gitignore
@@ -15,7 +15,7 @@ __pycache__
 DartConfiguration.tcl
 .DS_Store
 rmm_log.txt
-.ruff_cache
+.ruff_cache/
 
 # Unit test / coverage reports
 htmlcov/

diff --git a/build.sh b/build.sh
@@ -29,12 +29,14 @@ VALIDARGS="
    pylibwholegraph
    libwholegraph
    tests
+   benchmarks
    all
    -v
    -g
    -n
    --pydevelop
    --allgpuarch
+   --compile-cmd
    --clean
    -h
    --help
@@ -49,13 +51,16 @@ HELP="$0 [<target> ...] [<flag> ...]
    pylibwholegraph            - build the pylibwholegraph Python package
    libwholegraph              - build the libwholegraph library
    tests                      - build the C++ tests
+   benchmarks                 - build benchmarks
    all                        - build everything
  and <flag> is:
    -v                         - verbose build mode
    -g                         - build for debug
    -n                         - do not install after a successful build (does not affect Python packages)
    --pydevelop                - install the Python packages in editable mode
    --allgpuarch               - build for all supported GPU architectures
+   --enable-nvshmem            - build with nvshmem support (beta).
+   --compile-cmd               - only output compile commands (invoke CMake without build)
    --clean                    - clean an individual target (note: to do a complete rebuild, use the clean target described above)
    -h                         - print this text
 
@@ -140,11 +145,22 @@ if hasArg --pydevelop; then
     PYTHON_ARGS_FOR_INSTALL="${PYTHON_ARGS_FOR_INSTALL} -e"
 fi
 
+if hasArg --enable-nvshmem; then
+    BUILD_WITH_NVSHMEM=ON
+else
+    BUILD_WITH_NVSHMEM=OFF
+fi
 if hasArg tests; then
     BUILD_TESTS=ON
 else
     BUILD_TESTS=OFF
 fi
+if hasArg benchmarks; then
+    BUILD_BENCHMARKS=ON
+else
+    BUILD_BENCHMARKS=OFF
+fi
+
 
 # If clean or uninstall targets given, run them prior to any other steps
 if hasArg uninstall; then
@@ -250,7 +266,7 @@ if hasArg cugraph-pyg || buildDefault || hasArg all; then
     fi
 fi
 
-# Install the cugraph-dgl extensions for DGL
+# Build and install the cugraph-dgl Python package
 if hasArg cugraph-dgl || buildDefault ||hasArg all; then
     if hasArg --clean; then
         cleanPythonDir ${REPODIR}/python/cugraph-dgl

diff --git a/ci/build_cpp.sh b/ci/build_cpp.sh
@@ -13,7 +13,11 @@ export CMAKE_GENERATOR=Ninja
 
 rapids-print-env
 
-version=$(rapids-generate-version)
+# TODO: revert this once we start publishing nightly packages
+#       from the 'cugraph-gnn' repo and stop publishing them from
+#       the 'cugraph' / 'wholegraph' repos
+#version=$(rapids-generate-version)
+version="24.12.00a1000"
 
 rapids-logger "Begin cpp build"
 

diff --git a/ci/build_python.sh b/ci/build_python.sh
@@ -15,7 +15,11 @@ rapids-print-env
 
 CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
 
-rapids-generate-version > ./VERSION
+# TODO: revert this once we start publishing nightly packages
+#       from the 'cugraph-gnn' repo and stop publishing them from
+#       the 'cugraph' / 'wholegraph' repos
+# rapids-generate-version > ./VERSION
+echo "24.12.00a1000" > ./VERSION
 
 sccache --zero-stats
 

diff --git a/ci/test_notebooks.sh b/ci/test_notebooks.sh
@@ -5,6 +5,8 @@ set -Eeuo pipefail
 
 . /opt/conda/etc/profile.d/conda.sh
 
+RAPIDS_VERSION="$(rapids-version)"
+
 rapids-logger "Generate notebook testing dependencies"
 rapids-dependency-file-generator \
   --output conda \
@@ -24,18 +26,20 @@ rapids-logger "Downloading artifacts from previous jobs"
 CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
 PYTHON_CHANNEL=$(rapids-download-conda-from-s3 python)
 
+# TODO: remove the '>=24.12.00a1000' once we start publishing nightly packages
+#       from the 'cugraph-gnn' repo and stop publishing them from
+#       the 'cugraph' / 'wholegraph' repos
 rapids-mamba-retry install \
   --channel "${CPP_CHANNEL}" \
   --channel "${PYTHON_CHANNEL}" \
-  libcugraph pylibcugraph cugraph
+  --channel dglteam/label/th23_cu118 \
+  "cugraph-dgl=${RAPIDS_VERSION},>=24.12.00a1000"
 
 NBTEST="$(realpath "$(dirname "$0")/utils/nbtest.sh")"
 NOTEBOOK_LIST="$(realpath "$(dirname "$0")/notebook_list.py")"
 EXITCODE=0
 trap "EXITCODE=1" ERR
 
-
-pushd notebooks
 TOPLEVEL_NB_FOLDERS="$(find . -name "*.ipynb" | cut -d'/' -f2 | sort -u)"
 set +e
 # Always run nbtest in all TOPLEVEL_NB_FOLDERS, set EXITCODE to failure

diff --git a/ci/test_python.sh b/ci/test_python.sh
@@ -27,10 +27,120 @@ RAPIDS_TESTS_DIR=${RAPIDS_TESTS_DIR:-"${PWD}/test-results"}
 RAPIDS_COVERAGE_DIR=${RAPIDS_COVERAGE_DIR:-"${PWD}/coverage-results"}
 mkdir -p "${RAPIDS_TESTS_DIR}" "${RAPIDS_COVERAGE_DIR}"
 
+# RAPIDS_DATASET_ROOT_DIR is used by test scripts
+export RAPIDS_DATASET_ROOT_DIR="$(realpath datasets)"
+mkdir -p "${RAPIDS_DATASET_ROOT_DIR}"
+pushd "${RAPIDS_DATASET_ROOT_DIR}"
+./get_test_data.sh --benchmark
+popd
+
 EXITCODE=0
 trap "EXITCODE=1" ERR
 set +e
 
+# Test runs that include tests that use dask require
+# --import-mode=append. Those tests start a LocalCUDACluster that inherits
+# changes from pytest's modifications to PYTHONPATH (which defaults to
+# prepending source tree paths to PYTHONPATH).  This causes the
+# LocalCUDACluster subprocess to import cugraph from the source tree instead of
+# the install location, and in most cases, the source tree does not have
+# extensions built in-place and will result in ImportErrors.
+#
+# FIXME: TEMPORARILY disable MG PropertyGraph tests (experimental) tests and
+# bulk sampler IO tests (hangs in CI)
+
+if [[ "${RUNNER_ARCH}" != "ARM64" ]]; then
+  rapids-mamba-retry env create --yes -f env.yaml -n test_cugraph_dgl
+
+  # activate test_cugraph_dgl environment for dgl
+  set +u
+  conda activate test_cugraph_dgl
+  set -u
+
+  if [[ "${RAPIDS_CUDA_VERSION%%.*}" == "11" ]]; then
+    DGL_CHANNEL="dglteam/label/th23_cu118"
+  else
+    DGL_CHANNEL="dglteam/label/th23_cu121"
+  fi
+
+
+  # TODO: remove the '>=24.12.00a1000' once we start publishing nightly packages
+  #       from the 'cugraph-gnn' repo and stop publishing them from
+  #       the 'cugraph' / 'wholegraph' repos
+  rapids-mamba-retry install \
+    --channel "${CPP_CHANNEL}" \
+    --channel "${PYTHON_CHANNEL}" \
+    --channel pytorch \
+    --channel conda-forge \
+    --channel "${DGL_CHANNEL}" \
+    --channel nvidia \
+    "pylibwholegraph=${RAPIDS_VERSION},>=24.12.00a1000" \
+    "cugraph-dgl=${RAPIDS_VERSION},>=24.12.00a1000" \
+    'pytorch::pytorch>=2.3,<2.4' \
+    "ogb"
+
+  rapids-print-env
+
+  rapids-logger "Check GPU usage"
+  nvidia-smi
+
+  rapids-logger "pytest cugraph_dgl (single GPU)"
+  ./ci/run_cugraph_dgl_pytests.sh \
+    --junitxml="${RAPIDS_TESTS_DIR}/junit-cugraph-dgl.xml" \
+    --cov-config=../../.coveragerc \
+    --cov=cugraph_dgl \
+    --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cugraph-dgl-coverage.xml" \
+    --cov-report=term
+
+  # Reactivate the test environment back
+  set +u
+  conda deactivate
+  set -u
+else
+  rapids-logger "skipping cugraph_dgl pytest on ARM64"
+fi
+
+if [[ "${RUNNER_ARCH}" != "ARM64" ]]; then
+  rapids-mamba-retry env create --yes -f env.yaml -n test_cugraph_pyg
+
+  # Temporarily allow unbound variables for conda activation.
+  set +u
+  conda activate test_cugraph_pyg
+  set -u
+
+  # TODO: remove the '>=24.12.00a1000' once we start publishing nightly packages
+  #       from the 'cugraph-gnn' repo and stop publishing them from
+  #       the 'cugraph' / 'wholegraph' repos
+  rapids-mamba-retry install \
+    --channel "${CPP_CHANNEL}" \
+    --channel "${PYTHON_CHANNEL}" \
+    --channel pytorch \
+    "pylibwholegraph=${RAPIDS_VERSION},>=24.12.00a1000" \
+    "cugraph-pyg=${RAPIDS_VERSION},>=24.12.00a1000" \
+    'pytorch::pytorch>=2.3,<2.4' \
+    'ogb'
+
+  rapids-print-env
+
+  rapids-logger "Check GPU usage"
+  nvidia-smi
+
+  rapids-logger "pytest cugraph_pyg (single GPU)"
+  ./ci/run_cugraph_pyg_pytests.sh \
+    --junitxml="${RAPIDS_TESTS_DIR}/junit-cugraph-pyg.xml" \
+    --cov-config=../../.coveragerc \
+    --cov=cugraph_pyg \
+    --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cugraph-pyg-coverage.xml" \
+    --cov-report=term
+
+  # Reactivate the test environment back
+  set +u
+  conda deactivate
+  set -u
+else
+  rapids-logger "skipping cugraph_pyg pytest on ARM64"
+fi
+
 if [[ "${RUNNER_ARCH}" != "ARM64" ]]; then
   rapids-mamba-retry env create --yes -f env.yaml -n test_pylibwholegraph
 
@@ -39,13 +149,15 @@ if [[ "${RUNNER_ARCH}" != "ARM64" ]]; then
   conda activate test_pylibwholegraph
   set -u
 
-  # Will automatically install built dependencies of pylibwholegraph
+  # TODO: remove the '>=24.12.00a1000' once we start publishing nightly packages
+  #       from the 'cugraph-gnn' repo and stop publishing them from
+  #       the 'cugraph' / 'wholegraph' repos
   rapids-mamba-retry install \
     --channel "${CPP_CHANNEL}" \
     --channel "${PYTHON_CHANNEL}" \
     --channel pytorch \
     'mkl<2024.1.0' \
-    "pylibwholegraph=${RAPIDS_VERSION}" \
+    "pylibwholegraph=${RAPIDS_VERSION},>=24.12.00a1000" \
     'pytorch::pytorch>=2.3,<2.4' \
     'pytest-forked' \
     'ogb'

diff --git a/ci/utils/nbtest.sh b/ci/utils/nbtest.sh
@@ -60,7 +60,7 @@ for nb in $*; do
     echo --------------------------------------------------------------------------------
     echo STARTING: ${NBNAME}
     echo --------------------------------------------------------------------------------
-    jupyter nbconvert --to script ${NBFILENAME} --output ${NBTMPDIR}/${NBNAME}-test
+    jupyter nbconvert --to python ${NBFILENAME} --output ${NBTMPDIR}/${NBNAME}-test
     echo "${MAGIC_OVERRIDE_CODE}" > ${NBTMPDIR}/tmpfile
     cat ${NBTESTSCRIPT} >> ${NBTMPDIR}/tmpfile
     mv ${NBTMPDIR}/tmpfile ${NBTESTSCRIPT}

diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -21,13 +21,15 @@ dependencies:
 - doxygen
 - graphviz
 - ipython
+- nbconvert
 - nbsphinx
 - nccl>=2.19
 - ninja
 - notebook>=0.5.0
 - numba>=0.57
 - numpy>=1.23,<3.0a0
 - numpydoc
+- ogb
 - pandas
 - pre-commit
 - pydantic
@@ -54,6 +56,5 @@ dependencies:
 - sphinx<6
 - sphinxcontrib-websupport
 - torchdata
-- wget
 - wheel
 name: all_cuda-118_arch-x86_64
diff --git a/conda/environments/all_cuda-121_arch-x86_64.yaml b/conda/environments/all_cuda-121_arch-x86_64.yaml
@@ -26,13 +26,15 @@ dependencies:
 - libcurand-dev
 - libcusolver-dev
 - libcusparse-dev
+- nbconvert
 - nbsphinx
 - nccl>=2.19
 - ninja
 - notebook>=0.5.0
 - numba>=0.57
 - numpy>=1.23,<3.0a0
 - numpydoc
+- ogb
 - pandas
 - pre-commit
 - pydantic
@@ -59,6 +61,5 @@ dependencies:
 - sphinx<6
 - sphinxcontrib-websupport
 - torchdata
-- wget
 - wheel
 name: all_cuda-121_arch-x86_64