Upgrade openai version to 1.13 and langchain to version 0.1.9 (#1529

) This PRs main focus is to update the version of the OpenAI package to 1.13 to allow using the built in retry logic and improved code. In addition, the following changes are also made: - Switch to using `langchain==0.1.9` installed via pip to get around pandas dependency issue - Sorted the lists in `dependencies.yaml` to make merging easier in the future - Improved logging for the `OpenAILlmService` to log all requests ordered to a file - New feature to add default model arguments to the service to avoid repeating arguments for each client. You can now just overwrite the default args - New feature to allow the `run_ci_local.sh` to persist any changes to the code back to the host via a volume mount. ## 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: - Michael Demoret (https://github.com/mdemoret-nv) - David Gardner (https://github.com/dagardner-nv) Approvers: - David Gardner (https://github.com/dagardner-nv) URL: #1529
nv-morpheus · Feb 27, 2024 · 594b10c · 594b10c
1 parent eeb8b51
commit 594b10c
Show file tree

Hide file tree

Showing 29 changed files with 389 additions and 230 deletions.
diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml
@@ -94,7 +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-240214
-      test_container: nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-test-240214
+      container: nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-build-240221
+      test_container: nvcr.io/ea-nvidia-morpheus/morpheus:morpheus-ci-test-240221
     secrets:
       NGC_API_KEY: ${{ secrets.NGC_API_KEY }}
diff --git a/ci/scripts/bootstrap_local_ci.sh b/ci/scripts/bootstrap_local_ci.sh
@@ -16,11 +16,16 @@
 
 export WORKSPACE_TMP="$(pwd)/ws_tmp"
 mkdir -p ${WORKSPACE_TMP}
-git clone ${GIT_URL} Morpheus
-cd Morpheus/
-git checkout ${GIT_BRANCH}
-git pull
-git checkout ${GIT_COMMIT}
+
+if [[ "${USE_HOST_GIT}" == "1" ]]; then
+    cd Morpheus/
+else
+    git clone ${GIT_URL} Morpheus
+    cd Morpheus/
+    git checkout ${GIT_BRANCH}
+    git pull
+    git checkout ${GIT_COMMIT}
+fi
 
 export MORPHEUS_ROOT=$(pwd)
 export WORKSPACE=${MORPHEUS_ROOT}

diff --git a/ci/scripts/run_ci_local.sh b/ci/scripts/run_ci_local.sh
@@ -41,6 +41,10 @@ function git_ssh_to_https()
 
 MORPHEUS_ROOT=${MORPHEUS_ROOT:-$(git rev-parse --show-toplevel)}
 
+# Specifies whether to mount the current git repo (to allow changes to be persisted) or to use a clean clone (to closely
+# match CI, the default)
+USE_HOST_GIT=${USE_HOST_GIT:-0}
+
 GIT_URL=$(git remote get-url origin)
 GIT_URL=$(git_ssh_to_https ${GIT_URL})
 
@@ -51,7 +55,7 @@ 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=${CONTAINER_VER:-240214}
+CONTAINER_VER=${CONTAINER_VER:-240221}
 CUDA_VER=${CUDA_VER:-12.1}
 DOCKER_EXTRA_ARGS=${DOCKER_EXTRA_ARGS:-""}
 
@@ -66,6 +70,7 @@ ENV_LIST="${ENV_LIST} --env GIT_COMMIT=${GIT_COMMIT}"
 ENV_LIST="${ENV_LIST} --env PARALLEL_LEVEL=$(nproc)"
 ENV_LIST="${ENV_LIST} --env CUDA_VER=${CUDA_VER}"
 ENV_LIST="${ENV_LIST} --env SKIP_CONDA_ENV_UPDATE=${SKIP_CONDA_ENV_UPDATE}"
+ENV_LIST="${ENV_LIST} --env USE_HOST_GIT=${USE_HOST_GIT}"
 
 mkdir -p ${LOCAL_CI_TMP}
 cp ${MORPHEUS_ROOT}/ci/scripts/bootstrap_local_ci.sh ${LOCAL_CI_TMP}
@@ -82,14 +87,20 @@ for STAGE in "${STAGES[@]}"; do
         DOCKER_RUN_ARGS="${DOCKER_RUN_ARGS} --runtime=runc"
     fi
 
+    if [[ "${USE_HOST_GIT}" == "1" ]]; then
+        DOCKER_RUN_ARGS="${DOCKER_RUN_ARGS} -v ${MORPHEUS_ROOT}:/Morpheus"
+    fi
+
     if [[ "${STAGE}" == "bash" ]]; then
         DOCKER_RUN_CMD="bash --init-file /ci_tmp/bootstrap_local_ci.sh"
     else
         DOCKER_RUN_CMD="/ci_tmp/bootstrap_local_ci.sh"
     fi
 
     echo "Running ${STAGE} stage in ${CONTAINER}"
+    set -x
     docker run ${DOCKER_RUN_ARGS} ${DOCKER_EXTRA_ARGS} ${CONTAINER} ${DOCKER_RUN_CMD}
+    set +x
 
     STATUS=$?
     if [[ ${STATUS} -ne 0 ]]; then

diff --git a/conda/environments/all_cuda-121_arch-x86_64.yaml b/conda/environments/all_cuda-121_arch-x86_64.yaml
@@ -54,7 +54,6 @@ dependencies:
 - isort
 - jsonpatch>=1.33
 - kfp
-- langchain=0.0.190
 - librdkafka>=1.9.2,<1.10.0a0
 - libtool
 - libwebp=1.3.2
@@ -67,10 +66,11 @@ dependencies:
 - ninja=1.11
 - nlohmann_json=3.9
 - nodejs=18.*
+- numexpr
 - numpydoc=1.5
 - nvtabular=23.08.00
 - onnx
-- openai=0.28
+- openai=1.13
 - papermill=2.4.0
 - pip
 - pkg-config=0.29
@@ -119,6 +119,7 @@ dependencies:
   - dgl
   - dglgo
   - google-search-results==2.4
+  - langchain==0.1.9
   - milvus==2.3.5
   - nemollm
   - pymilvus==2.3.6

diff --git a/conda/environments/examples_cuda-121_arch-x86_64.yaml b/conda/environments/examples_cuda-121_arch-x86_64.yaml
@@ -27,16 +27,16 @@ dependencies:
 - huggingface_hub=0.20.2
 - jsonpatch>=1.33
 - kfp
-- langchain=0.0.190
 - libwebp=1.3.2
 - mlflow=2.9.2
 - networkx=2.8.8
 - newspaper3k=0.2
 - nodejs=18.*
+- numexpr
 - numpydoc=1.5
 - nvtabular=23.08.00
 - onnx
-- openai=0.28
+- openai=1.13
 - papermill=2.4.0
 - pip
 - pypdf=3.17.4
@@ -63,6 +63,7 @@ dependencies:
   - dgl
   - dglgo
   - google-search-results==2.4
+  - langchain==0.1.9
   - milvus==2.3.5
   - nemollm
   - pymilvus==2.3.6

diff --git a/dependencies.yaml b/dependencies.yaml
@@ -21,107 +21,97 @@ files:
       cuda: ["12.1"]
       arch: [x86_64]
     includes:
-      - data_retrieval
+      - benchmark_cpp
       - build_cpp
+      - checks
+      - cudatoolkit
+      - cve-mitigation
+      - data_retrieval
       - development
-      - benchmark_cpp
-      - runtime
+      - doca
+      - docs
       - example-dfp-prod
       - example-gnn
       - example-llm-agents
       - example-llm-completion
       - example-llm-rag
       - example-llm-vdb-upload
-      - test_python_morpheus
-      - docs
-      - cudatoolkit
       - python
-      - checks
-      - cve-mitigation
-      - doca
+      - runtime
+      - test_python_morpheus
 
   dev:
     output: conda
     matrix:
       cuda: ["12.1"]
       arch: [x86_64]
     includes:
-      - data_retrieval
+      - benchmark_cpp
       - build_cpp
+      - checks
+      - cudatoolkit
+      - data_retrieval
       - development
-      - benchmark_cpp
-      - runtime
-      - test_python_morpheus
       - docs
-      - cudatoolkit
       - python
-      - checks
-
-  examples:
-    output: conda
-    matrix:
-      cuda: ["12.1"]
-      arch: [x86_64]
-    includes:
-      - examples
       - runtime
-      - cudatoolkit
+      - test_python_morpheus
 
   build:
     output: none
     matrix:
       cuda: ["12.1"]
       arch: [x86_64]
     includes:
-      - data_retrieval
+      - benchmark_cpp
       - build_cpp
+      - cudatoolkit
+      - data_retrieval
       - development
-      - benchmark_cpp
+      - python
       - runtime
       - test_python_morpheus
-      - cudatoolkit
-      - python
 
   test:
     output: none
     matrix:
       cuda: ["12.1"]
       arch: [x86_64]
     includes:
-      - data_retrieval
+      - benchmark_cpp
       - build_cpp
+      - cudatoolkit
+      - data_retrieval
       - development
-      - benchmark_cpp
-      - runtime
-      - test_python_morpheus
       - example-dfp-prod
       - example-gnn
       - example-llm-agents
       - example-llm-completion
       - example-llm-rag
       - example-llm-vdb-upload
-      - cudatoolkit
       - python
+      - runtime
+      - test_python_morpheus
 
   docs:
     output: none
     matrix:
       cuda: ["12.1"]
       arch: [x86_64]
     includes:
-      - data_retrieval
-      - build_cpp
       - benchmark_cpp
+      - build_cpp
       - cudatoolkit
-      - python
+      - data_retrieval
       - docs
-      - runtime
       - example-dfp-prod
       - example-gnn
       - example-llm-agents
       - example-llm-completion
       - example-llm-rag
       - example-llm-vdb-upload
+      - python
+      - runtime
 
   runtime:
     output: conda
@@ -138,15 +128,15 @@ files:
       cuda: ["12.1"]
       arch: [x86_64]
     includes:
-      - runtime
+      - cve-mitigation
       - example-dfp-prod
       - example-gnn
       - example-llm-agents
       - example-llm-completion
       - example-llm-rag
       - example-llm-vdb-upload
       - python
-      - cve-mitigation
+      - runtime
 
   model-utils:
     output: conda
@@ -231,18 +221,18 @@ dependencies:
     common:
       - output_types: [conda]
         packages:
+          - &click click >=8
+          - &numpydoc numpydoc=1.5
           - breathe=4.35.0
           - doxygen=1.9.2
           - exhale=0.3.6
           - ipython
           - myst-parser=0.18.1
           - nbsphinx
+          - pluggy=1.3
           - python-graphviz
           - sphinx
           - sphinx_rtd_theme
-          - &numpydoc numpydoc=1.5
-          - pluggy=1.3
-          - &click click >=8
 
   benchmark_cpp:
     common:
@@ -287,10 +277,10 @@ dependencies:
       - output_types: [conda]
         packages:
           - &nodejs nodejs=18.*
-          - pytest=7.4.4
           - pytest-asyncio
           - pytest-benchmark=4.0
           - pytest-cov
+          - pytest=7.4.4
           - python-docx==1.1.0
           - pip
           - pip:
@@ -328,12 +318,13 @@ dependencies:
       - output_types: [conda]
         packages:
           - &grpcio-status grpcio-status==1.59
-          - &langchain langchain=0.0.190
           - &transformers transformers=4.36.2 # newer versions are incompatible with our pinned version of huggingface_hub
           - huggingface_hub=0.20.2 # work-around for https://github.com/UKPLab/sentence-transformers/issues/1762
-          - pip
+          - numexpr
           - sentence-transformers
+          - pip
           - pip:
+            - &langchain langchain==0.1.9
             - nemollm
 
   example-llm-completion:
@@ -350,12 +341,12 @@ dependencies:
       - output_types: [conda]
         packages:
           - *grpcio-status
-          - *langchain
           - anyio>=3.7
           - jsonpatch>=1.33
-          - openai=0.28
+          - openai=1.13
           - pip
           - pip:
+            - *langchain
             - google-search-results==2.4
 
   example-llm-vdb-upload:
@@ -364,27 +355,27 @@ dependencies:
         packages:
           - *arxiv
           - *grpcio-status
-          - *langchain
           - *newspaper3k
           - *pypdf
           - onnx
           - pip
           - pip:
               - PyMuPDF==1.23.21
+              - *langchain
 
   model-training-tuning:
     common:
       - output_types: [conda]
         packages:
           - *cuml
+          - *scikit-learn
+          - *transformers
           - jupyterlab
           - matplotlib
           - onnx
           - pandas
-          - *scikit-learn
           - seaborn
           - seqeval=1.2.2
-          - *transformers
           - xgboost
 
   cve-mitigation:

diff --git a/docs/source/conf.py b/docs/source/conf.py
@@ -176,6 +176,7 @@
     "morpheus.cli.commands",  # Dont document the CLI in Sphinx
     "nvtabular",
     "pandas",
+    "pydantic",
     "pymilvus",
     "tensorrt",
     "torch",