Update post-23.09 release (#6367)

* Update README and versions for 23.09 branch (#6280) * Update `Dockerfile` and `build.py` (#6281) * Update configuration for Windows Dockerfile (#6256) * Adding installation of docker and docker-buildx * Enable '--expt-relaxed-constexpr' flag for custom ops models * Upate Dockerfile version * Disable unit tests for Jetson * Update condition (#6285) * removing Whitespaces (#6293) * removing Whitespaces * removing whitespaces
triton-inference-server · Sep 29, 2023 · f4d4278 · f4d4278
1 parent 1f4af6d
commit f4d4278
Show file tree

Hide file tree

Showing 25 changed files with 95 additions and 92 deletions.
diff --git a/Dockerfile.sdk b/Dockerfile.sdk
@@ -29,7 +29,7 @@
 #
 
 # Base image on the minimum Triton container
-ARG BASE_IMAGE=nvcr.io/nvidia/tritonserver:23.08-py3-min
+ARG BASE_IMAGE=nvcr.io/nvidia/tritonserver:23.09-py3-min
 
 ARG TRITON_CLIENT_REPO_SUBDIR=clientrepo
 ARG TRITON_COMMON_REPO_TAG=main

diff --git a/Dockerfile.win10.min b/Dockerfile.win10.min
@@ -39,6 +39,23 @@ RUN powershell.exe Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope Loca
 RUN powershell.exe [Net.ServicePointManager]::Expect100Continue=$true;[Net.ServicePointManager]::SecurityProtocol=[Net.SecurityProtocolType]::Tls,[Net.SecurityProtocolType]::Tls11,[Net.SecurityProtocolType]::Tls12,[Net.SecurityProtocolType]::Ssl3;Invoke-Expression( New-Object System.Net.WebClient ).DownloadString('https://chocolatey.org/install.ps1')
 RUN choco install git docker unzip -y
 
+#
+# Installing CMake
+#
+ARG CMAKE_VERSION=3.27.1
+ARG CMAKE_FILE=cmake-${CMAKE_VERSION}-windows-x86_64
+ARG CMAKE_SOURCE=https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/${CMAKE_FILE}.zip
+
+ADD ${CMAKE_SOURCE} ${CMAKE_FILE}.zip
+RUN unzip %CMAKE_FILE%.zip
+RUN move %CMAKE_FILE% "c:\CMake"
+RUN setx PATH "c:\CMake\bin;%PATH%"
+
+ENV CMAKE_TOOLCHAIN_FILE /vcpkg/scripts/buildsystems/vcpkg.cmake
+ENV VCPKG_TARGET_TRIPLET x64-windows
+
+LABEL CMAKE_VERSION=${CMAKE_VERSION}
+
 # Be aware that pip can interact badly with VS cmd shell so need to pip install before
 # vsdevcmd.bat (see https://bugs.python.org/issue38989)
 ARG PYTHON_VERSION=3.8.10
@@ -51,44 +68,31 @@ RUN pip install grpcio-tools
 
 LABEL PYTHON_VERSION=${PYTHON_VERSION}
 
-# Download and install Build Tools for Visual Studio. The use of
-# powershell for the install seems to be required to make the command
-# wait for the install to complete before continuing. To avoid failures
-# caused by VS regressions we want to stick with a working
-# compiler. Currently this is 16.11.21. This page contains download
-# links for buildtools.
-# https://docs.microsoft.com/en-us/visualstudio/releases/2019/history#release-dates-and-build-numbers
-ARG BUILDTOOLS_VERSION=16.11.21
-ARG BUILDTOOLS_SOURCE=https://download.visualstudio.microsoft.com/download/pr/8f1eb024-006a-43f6-a372-0721f71058b3/cc5cc690ac094fbfa78dfb8e40089ba52056026579e8d8dc31e95e8ea5466df5/vs_BuildTools.exe
-ADD ${BUILDTOOLS_SOURCE} vs_buildtools.exe
+#
+# Installing Visual Studio BuildTools: VS17 2022
+#
+ARG BUILDTOOLS_VERSION
+# Download collect.exe in case of an install failure.
+ADD https://aka.ms/vscollect.exe "C:\tmp\collect.exe"
+
+# Use the latest release channel. For more control, specify the location of an internal layout.
+ARG CHANNEL_URL=https://aka.ms/vs/17/release/channel
+ADD ${CHANNEL_URL} "C:\tmp\VisualStudio.chman"
+# Download the Build Tools bootstrapper.
+ARG BUILD_TOOLS_SOURCE=https://aka.ms/vs/17/release/vs_buildtools.exe
+ADD ${BUILD_TOOLS_SOURCE} vs_buildtools.exe
+# Install Build Tools with the Microsoft.VisualStudio.Workload.VCTools workload, including recommended.
 ARG VS_INSTALL_PATH_WP="C:\BuildTools"
-RUN powershell.exe Start-Process -FilePath vs_buildtools.exe -ArgumentList "--wait","--quiet","--norestart","--nocache","--installPath","%VS_INSTALL_PATH_WP%","--channelUri","C:\tmp\doesnotexist.chman","--addProductLang","En-us","--add","Microsoft.VisualStudio.Workload.VCTools`;includeRecommended","--add","Microsoft.Component.MSBuild" -Wait -PassThru
+RUN vs_buildtools.exe --quiet --wait --norestart --nocache install --installPath %VS_INSTALL_PATH_WP% --channelUri "C:\tmp\VisualStudio.chman" --installChannelUri "C:\tmp\VisualStudio.chman" --add Microsoft.VisualStudio.Workload.VCTools --includeRecommended --locale "En-us"
 
 LABEL BUILDTOOLS_VERSION=${BUILDTOOLS_VERSION}
 
 WORKDIR /
 
-#
-# Installing CMake
-#
-ARG CMAKE_VERSION=3.26.1
-ARG CMAKE_FILE=cmake-${CMAKE_VERSION}-windows-x86_64
-ARG CMAKE_SOURCE=https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/${CMAKE_FILE}.zip
-
-ADD ${CMAKE_SOURCE} ${CMAKE_FILE}.zip
-RUN unzip %CMAKE_FILE%.zip
-RUN move %CMAKE_FILE% CMake
-RUN setx PATH "c:\CMake\bin;%PATH%"
-
-ENV CMAKE_TOOLCHAIN_FILE /vcpkg/scripts/buildsystems/vcpkg.cmake
-ENV VCPKG_TARGET_TRIPLET x64-windows
-
-LABEL CMAKE_VERSION=${CMAKE_VERSION}
-
 #
 # Installing Vcpkg
 #
-ARG VCPGK_VERSION=2022.11.14
+ARG VCPGK_VERSION=2023.07.21
 RUN git clone --single-branch --depth=1 -b %VCPGK_VERSION% https://github.com/microsoft/vcpkg.git
 WORKDIR /vcpkg
 RUN bootstrap-vcpkg.bat
@@ -100,13 +104,12 @@ LABEL VCPGK_VERSION=${VCPGK_VERSION}
 
 WORKDIR /
 
-
 #
 # Installing CUDA
 #
 ARG CUDA_MAJOR=12
 ARG CUDA_MINOR=2
-ARG CUDA_PATCH=0
+ARG CUDA_PATCH=1
 ARG CUDA_VERSION=${CUDA_MAJOR}.${CUDA_MINOR}.${CUDA_PATCH}
 ARG CUDA_PACKAGES="nvcc_${CUDA_MAJOR}.${CUDA_MINOR} \
                    cudart_${CUDA_MAJOR}.${CUDA_MINOR} \
@@ -127,14 +130,14 @@ ADD ${CUDA_SOURCE} cuda_${CUDA_VERSION}_windows_network.exe
 RUN cuda_%CUDA_VERSION%_windows_network.exe -s %CUDA_PACKAGES%
 # Copy the CUDA visualstudio integration from where it was installed
 # into the appropriate place in BuildTools
-RUN copy "%CUDA_INSTALL_ROOT_WP%\extras\visual_studio_integration\MSBuildExtensions\*" "%VS_INSTALL_PATH_WP%\MSBuild\Microsoft\VC\v160\BuildCustomizations"
+RUN copy "%CUDA_INSTALL_ROOT_WP%\extras\visual_studio_integration\MSBuildExtensions\*" "%VS_INSTALL_PATH_WP%\MSBuild\Microsoft\VC\v170\BuildCustomizations"
 
 RUN setx PATH "%CUDA_INSTALL_ROOT_WP%\bin;%PATH%"
 
 LABEL CUDA_VERSION="${CUDA_VERSION}"
 
 #
-# Installing Tensorrt
+# Installing TensorRT
 #
 ARG TENSORRT_VERSION=8.6.1.6
 ARG TENSORRT_ZIP="TensorRT-${TENSORRT_VERSION}.Windows10.x86_64.cuda-12.0.zip"
@@ -152,9 +155,9 @@ LABEL TENSORRT_VERSION="${TENSORRT_VERSION}"
 
 
 #
-# Installing CUDNN
+# Installing cuDNN
 #
-ARG CUDNN_VERSION=8.9.4.25
+ARG CUDNN_VERSION=8.9.5.27
 ARG CUDNN_ZIP=cudnn-windows-x86_64-${CUDNN_VERSION}_cuda12-archive.zip
 ARG CUDNN_SOURCE=${CUDNN_ZIP}
 

diff --git a/README.md b/README.md
@@ -30,12 +30,6 @@
 
 [![License](https://img.shields.io/badge/License-BSD3-lightgrey.svg)](https://opensource.org/licenses/BSD-3-Clause)
 
-**LATEST RELEASE: You are currently on the main branch which tracks
-under-development progress towards the next release. The current release is
-version [2.37.0](https://github.com/triton-inference-server/server/tree/r23.08)
-and corresponds to the 23.08 container release on
-[NVIDIA GPU Cloud (NGC)](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/tritonserver).**
-
 ----
 Triton Inference Server is an open source inference serving software that
 streamlines AI inferencing. Triton enables teams to deploy any AI model from
@@ -91,16 +85,16 @@ Inference Server with the
 
 ```bash
 # Step 1: Create the example model repository
-git clone -b r23.08 https://github.com/triton-inference-server/server.git
+git clone -b r23.09 https://github.com/triton-inference-server/server.git
 cd server/docs/examples
 ./fetch_models.sh
 
 # Step 2: Launch triton from the NGC Triton container
-docker run --gpus=1 --rm --net=host -v ${PWD}/model_repository:/models nvcr.io/nvidia/tritonserver:23.08-py3 tritonserver --model-repository=/models
+docker run --gpus=1 --rm --net=host -v ${PWD}/model_repository:/models nvcr.io/nvidia/tritonserver:23.09-py3 tritonserver --model-repository=/models
 
 # Step 3: Sending an Inference Request
 # In a separate console, launch the image_client example from the NGC Triton SDK container
-docker run -it --rm --net=host nvcr.io/nvidia/tritonserver:23.08-py3-sdk
+docker run -it --rm --net=host nvcr.io/nvidia/tritonserver:23.09-py3-sdk
 /workspace/install/bin/image_client -m densenet_onnx -c 3 -s INCEPTION /workspace/images/mug.jpg
 
 # Inference should return the following

diff --git a/build.py b/build.py
@@ -69,8 +69,8 @@
 TRITON_VERSION_MAP = {
     "2.39.0dev": (
         "23.10dev",  # triton container
-        "23.08",  # upstream container
-        "1.15.1",  # ORT
+        "23.09",  # upstream container
+        "1.16.0",  # ORT
         "2023.0.0",  # ORT OpenVINO
         "2023.0.0",  # Standalone OpenVINO
         "2.4.7",  # DCGM version

diff --git a/deploy/aws/values.yaml b/deploy/aws/values.yaml
@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.09-py3
   pullPolicy: IfNotPresent
   modelRepositoryPath: s3://triton-inference-server-repository/model_repository
   numGpus: 1

diff --git a/deploy/fleetcommand/Chart.yaml b/deploy/fleetcommand/Chart.yaml
@@ -26,7 +26,7 @@
 
 apiVersion: v1
 # appVersion is the Triton version; update when changing release
-appVersion: "2.37.0"
+appVersion: "2.38.0"
 description: Triton Inference Server (Fleet Command)
 name: triton-inference-server
 # version is the Chart version; update when changing anything in the chart

diff --git a/deploy/fleetcommand/values.yaml b/deploy/fleetcommand/values.yaml
@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.09-py3
   pullPolicy: IfNotPresent
   numGpus: 1
   serverCommand: tritonserver
@@ -46,13 +46,13 @@ image:
     # Model Control Mode (Optional, default: none)
     #
     # To set model control mode, uncomment and configure below
-    # See https://github.com/triton-inference-server/server/blob/r23.08/docs/model_management.md
+    # See https://github.com/triton-inference-server/server/blob/r23.09/docs/model_management.md
     #  for more details
     #- --model-control-mode=explicit|poll|none
     #
     # Additional server args
     #
-    # see https://github.com/triton-inference-server/server/blob/r23.08/README.md
+    # see https://github.com/triton-inference-server/server/blob/r23.09/README.md
     #  for more details
 
 service:

diff --git a/deploy/gcp/values.yaml b/deploy/gcp/values.yaml
@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.09-py3
   pullPolicy: IfNotPresent
   modelRepositoryPath: gs://triton-inference-server-repository/model_repository
   numGpus: 1

diff --git a/deploy/gke-marketplace-app/benchmark/perf-analyzer-script/triton_client.yaml b/deploy/gke-marketplace-app/benchmark/perf-analyzer-script/triton_client.yaml
@@ -33,7 +33,7 @@ metadata:
   namespace: default
 spec:
   containers:
-  - image: nvcr.io/nvidia/tritonserver:23.08-py3-sdk
+  - image: nvcr.io/nvidia/tritonserver:23.09-py3-sdk
     imagePullPolicy: Always
     name: nv-triton-client
     securityContext:

diff --git a/deploy/gke-marketplace-app/server-deployer/build_and_push.sh b/deploy/gke-marketplace-app/server-deployer/build_and_push.sh
@@ -27,9 +27,9 @@
 
 export REGISTRY=gcr.io/$(gcloud config get-value project | tr ':' '/')
 export APP_NAME=tritonserver
-export MAJOR_VERSION=2.37
-export MINOR_VERSION=2.37.0
-export NGC_VERSION=23.08-py3
+export MAJOR_VERSION=2.38
+export MINOR_VERSION=2.38.0
+export NGC_VERSION=23.09-py3
 
 docker pull nvcr.io/nvidia/$APP_NAME:$NGC_VERSION
 

diff --git a/deploy/gke-marketplace-app/server-deployer/chart/triton/Chart.yaml b/deploy/gke-marketplace-app/server-deployer/chart/triton/Chart.yaml
@@ -25,7 +25,7 @@
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 apiVersion: v1
-appVersion: "2.37"
+appVersion: "2.38"
 description: Triton Inference Server
 name: triton-inference-server
-version: 2.37.0
+version: 2.38.0
diff --git a/deploy/gke-marketplace-app/server-deployer/chart/triton/values.yaml b/deploy/gke-marketplace-app/server-deployer/chart/triton/values.yaml
@@ -31,14 +31,14 @@ maxReplicaCount: 3
 tritonProtocol: HTTP
 # HPA GPU utilization autoscaling target
 HPATargetAverageValue: 85
-modelRepositoryPath: gs://triton_sample_models/23_08
-publishedVersion: '2.37.0'
+modelRepositoryPath: gs://triton_sample_models/23_09
+publishedVersion: '2.38.0'
 gcpMarketplace: true
 
 image:
   registry: gcr.io
   repository: nvidia-ngc-public/tritonserver
-  tag: 23.08-py3
+  tag: 23.09-py3
   pullPolicy: IfNotPresent
   # modify the model repository here to match your GCP storage bucket
   numGpus: 1

diff --git a/deploy/gke-marketplace-app/server-deployer/data-test/schema.yaml b/deploy/gke-marketplace-app/server-deployer/data-test/schema.yaml
@@ -27,7 +27,7 @@
 x-google-marketplace:
   schemaVersion: v2
   applicationApiVersion: v1beta1
-  publishedVersion: '2.37.0'
+  publishedVersion: '2.38.0'
   publishedVersionMetadata:
     releaseNote: >-
       Initial release.

diff --git a/deploy/gke-marketplace-app/server-deployer/schema.yaml b/deploy/gke-marketplace-app/server-deployer/schema.yaml
@@ -27,7 +27,7 @@
 x-google-marketplace:
   schemaVersion: v2
   applicationApiVersion: v1beta1
-  publishedVersion: '2.37.0'
+  publishedVersion: '2.38.0'
   publishedVersionMetadata:
     releaseNote: >-
       Initial release.
@@ -89,7 +89,7 @@ properties:
   modelRepositoryPath:
     type: string
     title: Bucket where models are stored. Please make sure the user/service account to create the GKE app has permission to this GCS bucket. Read Triton documentation on configs and formatting details, supporting TensorRT, TensorFlow, Pytorch, Onnx ... etc.
-    default: gs://triton_sample_models/23_08
+    default: gs://triton_sample_models/23_09
   image.ldPreloadPath:
     type: string
     title: Leave this empty by default. Triton allows users to create custom layers for backend such as TensorRT plugin or Tensorflow custom ops, the compiled shared library must be provided via LD_PRELOAD environment variable.

diff --git a/deploy/gke-marketplace-app/trt-engine/README.md b/deploy/gke-marketplace-app/trt-engine/README.md
@@ -33,7 +33,7 @@
 ```
 docker run --gpus all -it --network host \
     --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 \
-    -v ~:/scripts nvcr.io/nvidia/tensorrt:23.08-py3
+    -v ~:/scripts nvcr.io/nvidia/tensorrt:23.09-py3
 
 pip install onnx six torch tf2onnx tensorflow
 
@@ -57,7 +57,7 @@ mkdir -p engines
 
 python3 builder.py -m models/fine-tuned/bert_tf_ckpt_large_qa_squad2_amp_128_v19.03.1/model.ckpt -o engines/bert_large_int8_bs1_s128.engine -b 1 -s 128 -c models/fine-tuned/bert_tf_ckpt_large_qa_squad2_amp_128_v19.03.1/ -v models/fine-tuned/bert_tf_ckpt_large_qa_squad2_amp_128_v19.03.1/vocab.txt --int8 --fp16 --strict --calib-num 1 -iln -imh
 
-gsutil cp bert_large_int8_bs1_s128.engine gs://triton_sample_models/23_08/bert/1/model.plan
+gsutil cp bert_large_int8_bs1_s128.engine gs://triton_sample_models/23_09/bert/1/model.plan
 ```
 
-For each Triton upgrade, container version used to generate the model, and the model path in GCS `gs://triton_sample_models/23_08/` should be updated accordingly with the correct version.
+For each Triton upgrade, container version used to generate the model, and the model path in GCS `gs://triton_sample_models/23_09/` should be updated accordingly with the correct version.
diff --git a/deploy/k8s-onprem/values.yaml b/deploy/k8s-onprem/values.yaml
@@ -29,7 +29,7 @@ tags:
   loadBalancing: true
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.09-py3
   pullPolicy: IfNotPresent
   modelRepositoryServer: < Replace with the IP Address of your file server >
   modelRepositoryPath: /srv/models

diff --git a/docs/customization_guide/build.md b/docs/customization_guide/build.md
@@ -173,7 +173,7 @@ $ ./build.py ... --repo-tag=common:<container tag> --repo-tag=core:<container ta
 
 If you are building on a release branch then `<container tag>` will
 default to the branch name. For example, if you are building on the
-r23.08 branch, `<container tag>` will default to r23.08. If you are
+r23.09 branch, `<container tag>` will default to r23.09. If you are
 building on any other branch (including the *main* branch) then
 `<container tag>` will default to "main". Therefore, you typically do
 not need to provide `<container tag>` at all (nor the preceding
@@ -334,8 +334,8 @@ python build.py --cmake-dir=<path/to/repo>/build --build-dir=/tmp/citritonbuild
 If you are building on *main* branch then '<container tag>' will
 default to "main". If you are building on a release branch then
 '<container tag>' will default to the branch name. For example, if you
-are building on the r23.08 branch, '<container tag>' will default to
-r23.08. Therefore, you typically do not need to provide '<container
+are building on the r23.09 branch, '<container tag>' will default to
+r23.09. Therefore, you typically do not need to provide '<container
 tag>' at all (nor the preceding colon). You can use a different
 '<container tag>' for a component to instead use the corresponding
 branch/tag in the build. For example, if you have a branch called

diff --git a/docs/customization_guide/compose.md b/docs/customization_guide/compose.md
@@ -44,8 +44,8 @@ from source to get more exact customization.
 The `compose.py` script can be found in the [server repository](https://github.com/triton-inference-server/server).
 Simply clone the repository and run `compose.py` to create a custom container.
 Note: Created container version will depend on the branch that was cloned.
-For example branch [r23.08](https://github.com/triton-inference-server/server/tree/r23.08)
-should be used to create a image based on the NGC 23.08 Triton release.
+For example branch [r23.09](https://github.com/triton-inference-server/server/tree/r23.09)
+should be used to create a image based on the NGC 23.09 Triton release.
 
 `compose.py` provides `--backend`, `--repoagent` options that allow you to
 specify which backends and repository agents to include in the custom image.
@@ -76,19 +76,19 @@ For example, running
 ```
 python3 compose.py --backend tensorflow1 --repoagent checksum
 ```
-on branch [r23.08](https://github.com/triton-inference-server/server/tree/r23.08) pulls:
-- `min` container `nvcr.io/nvidia/tritonserver:23.08-py3-min`
-- `full` container `nvcr.io/nvidia/tritonserver:23.08-py3`
+on branch [r23.09](https://github.com/triton-inference-server/server/tree/r23.09) pulls:
+- `min` container `nvcr.io/nvidia/tritonserver:23.09-py3-min`
+- `full` container `nvcr.io/nvidia/tritonserver:23.09-py3`
 
 Alternatively, users can specify the version of Triton container to pull from any branch by either:
 1. Adding flag `--container-version <container version>` to branch
 ```
-python3 compose.py --backend tensorflow1 --repoagent checksum --container-version 23.08
+python3 compose.py --backend tensorflow1 --repoagent checksum --container-version 23.09
 ```
 2. Specifying `--image min,<min container image name> --image full,<full container image name>`.
    The user is responsible for specifying compatible `min` and `full` containers.
 ```
-python3 compose.py --backend tensorflow1 --repoagent checksum --image min,nvcr.io/nvidia/tritonserver:23.08-py3-min --image full,nvcr.io/nvidia/tritonserver:23.08-py3
+python3 compose.py --backend tensorflow1 --repoagent checksum --image min,nvcr.io/nvidia/tritonserver:23.09-py3-min --image full,nvcr.io/nvidia/tritonserver:23.09-py3
 ```
 Method 1 and 2 will result in the same composed container. Furthermore, `--image` flag overrides the `--container-version` flag when both are specified.