RAPIDS 24.02 Upgrade (#1468)

Requires nv-morpheus/utilities#66 Closes #1428 Closes #1430 Closes #1433 ## By Submitting this PR I confirm: - I am familiar with the [Contributing Guidelines](https://github.com/nv-morpheus/Morpheus/blob/main/docs/source/developer_guide/contributing.md). - When the PR is ready for review, new or existing tests cover these changes. - When the PR is ready for review, the documentation is up to date with these changes. Authors: - Christopher Harris (https://github.com/cwharris) Approvers: - Michael Demoret (https://github.com/mdemoret-nv) URL: #1468
nv-morpheus · Feb 5, 2024 · 4c95c1d · 4c95c1d
1 parent 16ca5df
commit 4c95c1d
Show file tree

Hide file tree

Showing 76 changed files with 865 additions and 1,224 deletions.
diff --git a/.clang-tidy b/.clang-tidy
@@ -5,7 +5,17 @@
 # Disable modernize-use-trailing-return-type (too many hits)
 # Disable modernize-use-nodiscard (why isnt [[nodiscard]] the default?)
 # Enable identifier in TODO comments
-Checks: -*,readability-identifier-naming,modernize-*,-modernize-use-trailing-return-type,-modernize-use-nodiscard,google-readability-todo
+Checks: >
+  -*,
+  readability-identifier-naming,
+  modernize-*,
+  -modernize-use-trailing-return-type,
+  -modernize-use-nodiscard,
+  google-readability-todo
+
+WarningsAsErrors: >
+  *,
+  -clang-diagnostic-unused-command-line-argument
 
 #WarningsAsErrors: '*'
 HeaderFilterRegex: '.*\/include\/morpheus\/.*'

diff --git a/.devcontainer/Dockerfile b/.devcontainer/Dockerfile
@@ -13,6 +13,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-FROM rapidsai/devcontainers:23.10-cpp-cuda11.8-mambaforge-ubuntu22.04 AS base
+FROM rapidsai/devcontainers:23.12-cpp-cuda12.1-mambaforge-ubuntu22.04 AS base
 
 ENV PATH="${PATH}:/workspaces/morpheus/.devcontainer/bin"
diff --git a/.devcontainer/opt/morpheus/bin/post-attach-command.sh b/.devcontainer/opt/morpheus/bin/post-attach-command.sh
@@ -27,6 +27,6 @@ ENV_NAME=${ENV_NAME:-morpheus}
 sed -ri "s/conda activate base/conda activate $ENV_NAME/g" ~/.bashrc;
 
 if conda_env_find "${ENV_NAME}" ; \
-then mamba env update --name ${ENV_NAME} -f ${MORPHEUS_ROOT}/conda/environments/all_cuda-118_arch-x86_64.yaml --prune; \
-else mamba env create --name ${ENV_NAME} -f ${MORPHEUS_ROOT}/conda/environments/all_cuda-118_arch-x86_64.yaml; \
+then mamba env update --name ${ENV_NAME} -f ${MORPHEUS_ROOT}/conda/environments/all_cuda-121_arch-x86_64.yaml --prune; \
+else mamba env create --name ${ENV_NAME} -f ${MORPHEUS_ROOT}/conda/environments/all_cuda-121_arch-x86_64.yaml; \
 fi
diff --git a/.github/workflows/ci_pipe.yml b/.github/workflows/ci_pipe.yml
@@ -42,7 +42,7 @@ on:
 env:
   CHANGE_TARGET: "${{ github.base_ref }}"
   CUDA_PATH: "/usr/local/cuda/"
-  CUDA_VER: "11.8"
+  CUDA_VER: "12.1"
   GH_TOKEN: "${{ github.token }}"
   GIT_COMMIT: "${{ github.sha }}"
   MORPHEUS_ROOT: "${{ github.workspace }}/morpheus"

diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml
@@ -94,8 +94,7 @@ jobs:
       # Disable conda upload for now, once we have morpheus packages in conda forge set the value to
       # !fromJSON(needs.prepare.outputs.is_pr) && (fromJSON(needs.prepare.outputs.is_main_branch) && 'main' || 'dev') || ''
       conda_upload_label: ""
-      container: nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-build-230913
-      test_container: nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-test-230913
-      # Info about the PR. Empty for non PR branches. Useful for extracting PR number, title, etc.
+      container: nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-build-240129
+      test_container: nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-test-240129
     secrets:
       NGC_API_KEY: ${{ secrets.NGC_API_KEY }}
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -36,7 +36,7 @@ option(MORPHEUS_USE_IWYU "Enable running include-what-you-use as part of the bui
 
 set(MORPHEUS_PY_INSTALL_DIR "${CMAKE_CURRENT_BINARY_DIR}/wheel" CACHE STRING "Location to install the python directory")
 
-set(MORPHEUS_RAPIDS_VERSION "23.06" CACHE STRING "Sets default versions for RAPIDS libraries.")
+set(MORPHEUS_RAPIDS_VERSION "24.02" CACHE STRING "Sets default versions for RAPIDS libraries.")
 set(MORPHEUS_CACHE_DIR "${CMAKE_SOURCE_DIR}/.cache" CACHE PATH "Directory to contain all CPM and CCache data")
 mark_as_advanced(MORPHEUS_CACHE_DIR)
 

diff --git a/ci/conda/recipes/morpheus/conda_build_config.yaml b/ci/conda/recipes/morpheus/conda_build_config.yaml
@@ -23,13 +23,10 @@ cuda_compiler:
   - cuda-nvcc
 
 cuda_compiler_version:
-  - 11.8
+  - 12.1
 
 python:
   - 3.10
 
-boost:
-  - 1.82
-
 rapids_version:
-  - 23.06
+  - 24.02
diff --git a/ci/conda/recipes/morpheus/meta.yaml b/ci/conda/recipes/morpheus/meta.yaml
@@ -46,81 +46,82 @@ outputs:
     requirements:
       build:
         - {{ compiler("c") }}
-        - {{ compiler("cxx") }}
         - {{ compiler("cuda") }}
-        - automake 1.16.5 # Needed for DOCA build
+        - {{ compiler("cxx") }}
+        - automake =1.16.5 # Needed for DOCA build
         - ccache
-        - cmake 3.25
-        - cuda-cudart-dev {{ cuda_compiler_version }} # Needed by CMake to compile a test application
+        - cmake =3.27
+        - cuda-cudart-dev {{ cuda_compiler_version }}.* # Needed by CMake to compile a test application
+        - cuda-version {{ cuda_compiler_version }}.*
         - libtool # Needed for DOCA build
-        - ninja
-        - pkg-config # for mrc cmake
+        - ninja =1.11
+        - pkg-config =0.29 # for mrc cmake
+        - sysroot_linux-64 =2.17
       host:
-        - cuda-toolkit {{ cuda_compiler_version }}
+        - cuda-cudart-dev {{ cuda_compiler_version }}.*
+        - cuda-libraries-dev {{ cuda_compiler_version }}.* # required by matx for libcublas-dev
+        - cuda-nvml-dev {{ cuda_compiler_version }}.*
+        - cuda-nvrtc-dev {{ cuda_compiler_version }}.*
+        - cuda-nvtx-dev {{ cuda_compiler_version }}.*
+        - cuda-version {{ cuda_compiler_version }}.*
         - cudf {{ rapids_version }}
-        - cython >=0.29,<0.30
+        - cython =3.0
         - libcudf {{ rapids_version }}
-        - librdkafka 1.9.2
+        - librdkafka >=1.9.2,<1.10.0a0
         - mrc {{ minor_version }}
         - pip
-        - pyarrow * *_cuda # Ensure we get a CUDA build. Version determined by cuDF
-        - pyarrow-hotfix # CVE-2023-47248
-        - pybind11-stubgen 0.10.5
+        - pybind11-stubgen =0.10.5
         - python {{ python }}
-        - rapidjson 1.1
-        - rdma-core >=48 # Needed for DOCA. Compatible with UCX 1.14.1
-        - scikit-build 0.17.1
+        - rapidjson =1.1.0
+        - rdma-core >=48 # Needed for DOCA.
+        - scikit-build =0.17.6
         - versioneer-518
+        - zlib # required to build triton client
       run:
         # Runtime only requirements. This + setup.py is the definitive runtime requirement list
-        - {{ pin_compatible('cuda-cudart', min_pin='x.x', max_pin='x') }}
         - click >=8
-        - configargparse 1.5.*
+        - configargparse =1.5.*
+        - cuda-version {{ cuda_compiler_version }}.*
         - cudf
-        - cudf_kafka {{ rapids_version }}.*
         - cupy # Version determined from cudf
-        - datacompy 0.8.*
+        - datacompy =0.10.*
         - dill
         - distributed
-        - docker-py 5.0.*
+        - docker-py =5.0.*
         - grpcio # Version determined from cudf
         - libmrc
         - libwebp>=1.3.2 # Required for CVE mitigation: https://nvd.nist.gov/vuln/detail/CVE-2023-4863
         - mlflow>=2.2.1,<3
         - mrc
         - networkx>=2.8
-        - numpydoc 1.4.*
-        - nvtabular {{ rapids_version }}.*
-        - pandas 1.3.*
-        - pluggy 1.0.*
+        - numpydoc =1.5.*
+        - nvtabular =23.08.00
+        - pandas
+        - pluggy =1.3.*
         - pyarrow * *_cuda # Ensure we get a CUDA build. Version determined by cuDF
-        - pyarrow-hotfix # CVE-2023-47248
         - python
-        - python-confluent-kafka 1.9.2
-        - pytorch 2.0.1
+        - python-confluent-kafka >=1.9.2,<1.10.0a0
+        - pytorch
         - pytorch-cuda
-        - requests=2.31
-        - requests-cache=1.1
-        - scikit-learn 1.2.2.*
+        - requests
+        - requests-cache =1.1
+        - scikit-learn =1.3.2.*
         - sqlalchemy <2.0 # 2.0 is incompatible with pandas=1.3
-        - tqdm 4.*
-        - tritonclient 2.26.*
-        - typing_utils 0.1.*
-        - watchdog 2.1.*
+        - tqdm =4.*
+        - tritonclient =2.34.*
+        - typing_utils =0.1.*
+        - watchdog =3.0.*
         - websockets
-      run_constrained:
-        # Since we dont explicitly require this but other packages might, constrain the versions.
-        - {{ pin_compatible('cudatoolkit', min_pin='x.x', max_pin='x') }}
     test:
       requires:
-        - pynvml
-        - pytest
-        - pytest-cov
-        - pytest-benchmark
         # test that cuml can be installed in the env
         - cuml {{ rapids_version }}.*
+        - pynvml
+        - pytest-asyncio
+        - pytest-benchmark
+        - pytest-cov
+        - pytest =7.4.4
       source_files:
-        - docker/conda/environments/*
         - pyproject.toml
         - scripts/fetch_data.py
         - tests/*

diff --git a/ci/conda/recipes/run_conda_build.sh b/ci/conda/recipes/run_conda_build.sh
@@ -89,13 +89,15 @@ if [[ "${CONDA_COMMAND}" == "mambabuild" || "${CONDA_COMMAND}" == "build" ]]; th
    CONDA_ARGS_ARRAY+=("--build-id-pat" "{n}-{v}")
 fi
 
-# And default channels (with optional channel alias)
+# And default channels (should match dependencies.yaml) with optional channel alias
+CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}conda-forge")
+CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}huggingface")
 CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}rapidsai")
-CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}nvidia/label/cuda-11.8.0")
+CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}rapidsai-nightly")
 CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}nvidia")
 CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}nvidia/label/dev")
 CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}pytorch")
-CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}conda-forge")
+CONDA_ARGS_ARRAY+=("-c" "${CONDA_CHANNEL_ALIAS:+"${CONDA_CHANNEL_ALIAS%/}/"}defaults")
 
 if hasArg morpheus; then
    # Set GIT_VERSION to set the project version inside of meta.yaml

diff --git a/ci/runner/Dockerfile b/ci/runner/Dockerfile
@@ -16,14 +16,15 @@
 # limitations under the License.
 
 # Args used in FROM commands must come first
-ARG FROM_IMAGE="rapidsai/ci"
-ARG CUDA_PKG_VER=11-8
-ARG CUDA_SHORT_VER=11.8
-ARG CUDA_VER=11.8.0
+ARG FROM_IMAGE="rapidsai/ci-conda"
+ARG CUDA_PKG_VER=12-0
+ARG CUDA_SHORT_VER=12.1
+ARG CUDA_VER=12.1.1
 ARG LINUX_DISTRO=ubuntu
-ARG LINUX_VER=20.04
+ARG LINUX_VER=22.04
 ARG PROJ_NAME=morpheus
 ARG PYTHON_VER=3.10
+ARG ARCH=x86_64
 
 # Configure the base docker img
 FROM ${FROM_IMAGE}:cuda${CUDA_VER}-${LINUX_DISTRO}${LINUX_VER}-py${PYTHON_VER} AS base
@@ -34,27 +35,36 @@ ARG CUDA_SHORT_VER
 SHELL ["/bin/bash",  "-c"]
 
 # Create conda environment
-COPY ./docker/conda/environments/* /tmp/conda/
-
-RUN CONDA_ALWAYS_YES=true /opt/conda/bin/mamba env create -n ${PROJ_NAME} -q --file /tmp/conda/cuda${CUDA_SHORT_VER}_dev.yml && \
-    /opt/conda/bin/mamba install -n morpheus -c conda-forge "conda-merge>=0.2" && \
-    sed -i "s/conda activate base/conda activate ${PROJ_NAME}/g" ~/.bashrc && \
-    conda clean -afy && \
-    rm -rf /tmp/conda
+COPY ./dependencies.yaml /tmp/conda/
 
 
 # ============ build ==================
 FROM base as build
 
-# Add any build only dependencies here.
+# Add any build only dependencies here.ARG ARCH
+ARG CUDA_SHORT_VER
+ARG PROJ_NAME
+ARG PYTHON_VER
+
+RUN rapids-dependency-file-generator \
+      --config /tmp/conda/dependencies.yaml \
+      --output conda \
+      --file_key build \
+      --matrix "cuda=${CUDA_SHORT_VER};arch=${ARCH};py=${PYTHON_VER}" > /tmp/conda/env.yaml && \
+    CONDA_ALWAYS_YES=true /opt/conda/bin/conda env create -n ${PROJ_NAME} -q --file /tmp/conda/env.yaml && \
+    sed -i "s/conda activate base/conda activate ${PROJ_NAME}/g" ~/.bashrc && \
+    conda clean -afy && \
+    rm -rf /tmp/conda
 
 # ============ test ==================
 FROM base as test
 
 # Add any test only dependencies here.
 
-ARG PROJ_NAME
+ARG ARCH
 ARG CUDA_SHORT_VER
+ARG PROJ_NAME
+ARG PYTHON_VER
 
 RUN apt update && \
     DEBIAN_FRONTEND=noninteractive TZ=Etc/UTC \
@@ -63,7 +73,16 @@ RUN apt update && \
     apt clean && \
     rm -rf /var/lib/apt/lists/*
 
-COPY ./docker/conda/environments/cuda${CUDA_SHORT_VER}_examples.yml /tmp/conda/cuda${CUDA_SHORT_VER}_examples.yml
+RUN rapids-dependency-file-generator \
+      --config /tmp/conda/dependencies.yaml \
+      --output conda \
+      --file_key test \
+      --matrix "cuda=${CUDA_SHORT_VER};arch=${ARCH};py=${PYTHON_VER}" > /tmp/conda/env.yaml && \
+    CONDA_ALWAYS_YES=true /opt/conda/bin/conda env create -n ${PROJ_NAME} -q --file /tmp/conda/env.yaml && \
+    sed -i "s/conda activate base/conda activate ${PROJ_NAME}/g" ~/.bashrc && \
+    conda clean -afy && \
+    rm -rf /tmp/conda
+
 COPY ./ci/scripts/download_kafka.py /tmp/scripts/download_kafka.py
 
 # Install camouflage needed for unittests to mock a triton server

diff --git a/ci/scripts/github/common.sh b/ci/scripts/github/common.sh
@@ -77,7 +77,8 @@ function update_conda_env() {
 
 
         # Update the packages
-        rapids-mamba-retry env update -n morpheus --prune -q --file "$1"
+        # use conda instead of mamba due to bug: https://github.com/mamba-org/mamba/issues/3059
+        rapids-conda-retry env update -n morpheus --prune -q --file "$1" --solver=libmamba
     fi
 
     # Finally, reactivate

diff --git a/ci/scripts/run_ci_local.sh b/ci/scripts/run_ci_local.sh
@@ -51,8 +51,8 @@ GIT_BRANCH=$(git branch --show-current)
 GIT_COMMIT=$(git log -n 1 --pretty=format:%H)
 
 LOCAL_CI_TMP=${LOCAL_CI_TMP:-${MORPHEUS_ROOT}/.tmp/local_ci_tmp}
-CONTAINER_VER=${CONTAINER_VER:-230913}
-CUDA_VER=${CUDA_VER:-11.8}
+CONTAINER_VER=${CONTAINER_VER:-240129}
+CUDA_VER=${CUDA_VER:-12.1}
 DOCKER_EXTRA_ARGS=${DOCKER_EXTRA_ARGS:-""}
 
 BUILD_CONTAINER="nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-build-${CONTAINER_VER}"

diff --git a/ci/scripts/run_clang_tidy_for_ci.sh b/ci/scripts/run_clang_tidy_for_ci.sh
@@ -16,11 +16,7 @@
 
 # set -x
 
-# Call clang-tidy adding warnings-as-errors option. Currently this is not
-# possible with clang-tidy-diff.py until this is merged:
-# https://reviews.llvm.org/D49864
-
 # Also add -fno-caret-diagnostics to prevent clangs own compiler warnings from
 # coming through:
 # https://github.com/llvm/llvm-project/blob/3f3faa36ff3d84af3c3ed84772d7e4278bc44ff1/libc/cmake/modules/LLVMLibCObjectRules.cmake#L226
-${CLANG_TIDY:-clang-tidy} --warnings-as-errors='*' --extra-arg=-fno-caret-diagnostics "$@"
+${CLANG_TIDY:-clang-tidy} --extra-arg=-fno-caret-diagnostics "$@"
diff --git a/cmake/dependencies.cmake b/cmake/dependencies.cmake
@@ -34,6 +34,11 @@ rapids_find_package(CUDAToolkit REQUIRED
   INSTALL_EXPORT_SET ${PROJECT_NAME}-core-exports
 )
 
+rapids_find_package(ZLIB
+  BUILD_EXPORT_SET ${PROJECT_NAME}-core-exports
+  INSTALL_EXPORT_SET ${PROJECT_NAME}-core-exports
+)
+
 if(MORPHEUS_BUILD_BENCHMARKS)
   # google benchmark
   # - Expects package to pre-exist in the build environment
@@ -62,9 +67,9 @@ if(MORPHEUS_BUILD_TESTS)
   )
 endif()
 
-# libcudacxx -- get an explicit lubcudacxx build, matx tries to pull a tag that doesn't exist.
+# cccl -- get an explicit cccl build, matx tries to pull a tag that doesn't exist.
 # =========
-morpheus_utils_configure_libcudacxx()
+morpheus_utils_configure_cccl()
 
 # matx
 # ====