Merge remote-tracking branch 'origin/main' into fpetrini-dim-shape-apis

Dockerfile.sdk

@@ -29,7 +29,7 @@
 #
 
 # Base image on the minimum Triton container
-ARG BASE_IMAGE=nvcr.io/nvidia/tritonserver:23.07-py3-min
+ARG BASE_IMAGE=nvcr.io/nvidia/tritonserver:23.08-py3-min
 
 ARG TRITON_CLIENT_REPO_SUBDIR=clientrepo
 ARG TRITON_COMMON_REPO_TAG=main

Dockerfile.win10.min

@@ -64,22 +64,7 @@ ADD ${BUILDTOOLS_SOURCE} vs_buildtools.exe
 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
 
-LABEL BUILDTOOLS_VERSION=16.11.21
-
-WORKDIR /
-
-#
-# Installing Vcpkg
-#
-ARG VCPGK_VERSION=2022.11.14
-RUN git clone --single-branch --depth=1 -b %VCPGK_VERSION% https://github.com/microsoft/vcpkg.git
-WORKDIR /vcpkg
-RUN bootstrap-vcpkg.bat
-RUN vcpkg.exe update
-RUN vcpkg.exe install openssl:x64-windows openssl-windows:x64-windows rapidjson:x64-windows re2:x64-windows boost-interprocess:x64-windows boost-stacktrace:x64-windows zlib:x64-windows pthread:x64-windows b64:x64-windows
-RUN vcpkg.exe integrate install
-
-LABEL VCPGK_VERSION=${VCPGK_VERSION}
+LABEL BUILDTOOLS_VERSION=${BUILDTOOLS_VERSION}
 
 WORKDIR /
 
@@ -100,12 +85,28 @@ ENV VCPKG_TARGET_TRIPLET x64-windows
 
 LABEL CMAKE_VERSION=${CMAKE_VERSION}
 
+#
+# Installing Vcpkg
+#
+ARG VCPGK_VERSION=2022.11.14
+RUN git clone --single-branch --depth=1 -b %VCPGK_VERSION% https://github.com/microsoft/vcpkg.git
+WORKDIR /vcpkg
+RUN bootstrap-vcpkg.bat
+RUN vcpkg.exe update
+RUN vcpkg.exe install openssl:x64-windows openssl-windows:x64-windows rapidjson:x64-windows re2:x64-windows boost-interprocess:x64-windows boost-stacktrace:x64-windows zlib:x64-windows pthread:x64-windows b64:x64-windows
+RUN vcpkg.exe integrate install
+
+LABEL VCPGK_VERSION=${VCPGK_VERSION}
+
+WORKDIR /
+
+
 #
 # Installing CUDA
 #
 ARG CUDA_MAJOR=12
-ARG CUDA_MINOR=1
-ARG CUDA_PATCH=1
+ARG CUDA_MINOR=2
+ARG CUDA_PATCH=0
 ARG CUDA_VERSION=${CUDA_MAJOR}.${CUDA_MINOR}.${CUDA_PATCH}
 ARG CUDA_PACKAGES="nvcc_${CUDA_MAJOR}.${CUDA_MINOR} \
                    cudart_${CUDA_MAJOR}.${CUDA_MINOR} \
@@ -153,7 +154,7 @@ LABEL TENSORRT_VERSION="${TENSORRT_VERSION}"
 #
 # Installing CUDNN
 #
-ARG CUDNN_VERSION=8.9.3.28
+ARG CUDNN_VERSION=8.9.4.25
 ARG CUDNN_ZIP=cudnn-windows-x86_64-${CUDNN_VERSION}_cuda12-archive.zip
 ARG CUDNN_SOURCE=${CUDNN_ZIP}
 

README.md

@@ -32,8 +32,8 @@
 
 **LATEST RELEASE: You are currently on the main branch which tracks
 under-development progress towards the next release. The current release is
-version [2.36.0](https://github.com/triton-inference-server/server/tree/r23.07)
-and corresponds to the 23.07 container release on
+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).**
 
 ----
@@ -91,16 +91,16 @@ Inference Server with the
 
 ```bash
 # Step 1: Create the example model repository
-git clone -b r23.07 https://github.com/triton-inference-server/server.git
+git clone -b r23.08 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:23.07-py3 tritonserver --model-repository=/models
+docker run --gpus=1 --rm --net=host -v ${PWD}/model_repository:/models nvcr.io/nvidia/tritonserver:23.08-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:23.07-py3-sdk
+docker run -it --rm --net=host nvcr.io/nvidia/tritonserver:23.08-py3-sdk
 /workspace/install/bin/image_client -m densenet_onnx -c 3 -s INCEPTION /workspace/images/mug.jpg
 
 # Inference should return the following
@@ -260,4 +260,4 @@ For questions, we recommend posting in our community
 ## For more information
 
 Please refer to the [NVIDIA Developer Triton page](https://developer.nvidia.com/nvidia-triton-inference-server)
-for more information.
+for more information.

build.py

@@ -69,7 +69,7 @@
 TRITON_VERSION_MAP = {
     "2.38.0dev": (
         "23.09dev",  # triton container
-        "23.07",  # upstream container
+        "23.08",  # upstream container
         "1.15.1",  # ORT
         "2023.0.0",  # ORT OpenVINO
         "2023.0.0",  # Standalone OpenVINO
@@ -1312,10 +1312,10 @@ def add_cpu_libs_to_linux_dockerfile(backends, target_machine):
 COPY --from=min_container /usr/local/cuda/lib64/stubs/libcublasLt.so /usr/local/cuda/lib64/stubs/libcublasLt.so.11
 
 RUN mkdir -p /usr/local/cuda/targets/{cuda_arch}-linux/lib
-COPY --from=min_container /usr/local/cuda-12.1/targets/{cuda_arch}-linux/lib/libcudart.so.12 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
-COPY --from=min_container /usr/local/cuda-12.1/targets/{cuda_arch}-linux/lib/libcupti.so.12 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
-COPY --from=min_container /usr/local/cuda-12.1/targets/{cuda_arch}-linux/lib/libnvToolsExt.so.1 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
-COPY --from=min_container /usr/local/cuda-12.1/targets/{cuda_arch}-linux/lib/libnvJitLink.so.12 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
+COPY --from=min_container /usr/local/cuda-12.2/targets/{cuda_arch}-linux/lib/libcudart.so.12 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
+COPY --from=min_container /usr/local/cuda-12.2/targets/{cuda_arch}-linux/lib/libcupti.so.12 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
+COPY --from=min_container /usr/local/cuda-12.2/targets/{cuda_arch}-linux/lib/libnvToolsExt.so.1 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
+COPY --from=min_container /usr/local/cuda-12.2/targets/{cuda_arch}-linux/lib/libnvJitLink.so.12 /usr/local/cuda/targets/{cuda_arch}-linux/lib/.
 
 RUN mkdir -p /opt/hpcx/ucc/lib/ /opt/hpcx/ucx/lib/
 COPY --from=min_container /opt/hpcx/ucc/lib/libucc.so.1 /opt/hpcx/ucc/lib/libucc.so.1

deploy/aws/values.yaml

@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.07-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.08-py3
   pullPolicy: IfNotPresent
   modelRepositoryPath: s3://triton-inference-server-repository/model_repository
   numGpus: 1

deploy/fleetcommand/Chart.yaml

@@ -26,7 +26,7 @@
 
 apiVersion: v1
 # appVersion is the Triton version; update when changing release
-appVersion: "2.36.0"
+appVersion: "2.37.0"
 description: Triton Inference Server (Fleet Command)
 name: triton-inference-server
 # version is the Chart version; update when changing anything in the chart

deploy/fleetcommand/values.yaml

@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.07-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.08-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/r23.07/docs/model_management.md
+    # See https://github.com/triton-inference-server/server/blob/r23.08/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/r23.07/README.md
+    # see https://github.com/triton-inference-server/server/blob/r23.08/README.md
     #  for more details
 
 service:

deploy/gcp/values.yaml

@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.07-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.08-py3
   pullPolicy: IfNotPresent
   modelRepositoryPath: gs://triton-inference-server-repository/model_repository
   numGpus: 1

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:23.07-py3-sdk
+  - image: nvcr.io/nvidia/tritonserver:23.08-py3-sdk
     imagePullPolicy: Always
     name: nv-triton-client
     securityContext:

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=2.33
-export MINOR_VERSION=2.36.0
-export NGC_VERSION=23.07-py3
+export MAJOR_VERSION=2.37
+export MINOR_VERSION=2.37.0
+export NGC_VERSION=23.08-py3
 
 docker pull nvcr.io/nvidia/$APP_NAME:$NGC_VERSION
 

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: "2.33"
+appVersion: "2.37"
 description: Triton Inference Server
 name: triton-inference-server
-version: 2.36.0
+version: 2.37.0

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_04
-publishedVersion: '2.36.0'
+modelRepositoryPath: gs://triton_sample_models/23_08
+publishedVersion: '2.37.0'
 gcpMarketplace: true
 
 image:
   registry: gcr.io
   repository: nvidia-ngc-public/tritonserver
-  tag: 23.07-py3
+  tag: 23.08-py3
   pullPolicy: IfNotPresent
   # modify the model repository here to match your GCP storage bucket
   numGpus: 1

deploy/gke-marketplace-app/server-deployer/data-test/schema.yaml

@@ -27,7 +27,7 @@
 x-google-marketplace:
   schemaVersion: v2
   applicationApiVersion: v1beta1
-  publishedVersion: '2.36.0'
+  publishedVersion: '2.37.0'
   publishedVersionMetadata:
     releaseNote: >-
       Initial release.

deploy/gke-marketplace-app/server-deployer/schema.yaml

@@ -27,7 +27,7 @@
 x-google-marketplace:
   schemaVersion: v2
   applicationApiVersion: v1beta1
-  publishedVersion: '2.36.0'
+  publishedVersion: '2.37.0'
   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_02
+    default: gs://triton_sample_models/23_08
   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.

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:23.06-py3
+    -v ~:/scripts nvcr.io/nvidia/tensorrt:23.08-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_02/bert/1/model.plan
+gsutil cp bert_large_int8_bs1_s128.engine gs://triton_sample_models/23_08/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_02/` 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_08/` should be updated accordingly with the correct version.

deploy/k8s-onprem/values.yaml

@@ -29,7 +29,7 @@ tags:
   loadBalancing: true
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:23.06-py3
+  imageName: nvcr.io/nvidia/tritonserver:23.08-py3
   pullPolicy: IfNotPresent
   modelRepositoryServer: < Replace with the IP Address of your file server >
   modelRepositoryPath: /srv/models

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
-r23.06 branch, `<container tag>` will default to r23.06. If you are
+r23.08 branch, `<container tag>` will default to r23.08. 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 r23.06 branch, '<container tag>' will default to
-r23.06. Therefore, you typically do not need to provide '<container
+are building on the r23.08 branch, '<container tag>' will default to
+r23.08. 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

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 [r23.06](https://github.com/triton-inference-server/server/tree/r23.06)
-should be used to create a image based on the NGC 23.06 Triton release.
+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.
 
 `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 [r23.06](https://github.com/triton-inference-server/server/tree/r23.06) pulls:
-- `min` container `nvcr.io/nvidia/tritonserver:23.06-py3-min`
-- `full` container `nvcr.io/nvidia/tritonserver:23.06-py3`
+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`
 
 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 23.06
+python3 compose.py --backend tensorflow1 --repoagent checksum --container-version 23.08
 ```
 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:23.06-py3-min --image full,nvcr.io/nvidia/tritonserver:23.06-py3
+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
 ```
 Method 1 and 2 will result in the same composed container. Furthermore, `--image` flag overrides the `--container-version` flag when both are specified.
 

docs/customization_guide/test.md

@@ -49,7 +49,7 @@ $ ./gen_qa_custom_ops
 ```
 
 This will create multiple model repositories in /tmp/<version>/qa_*
-(for example /tmp/23.06/qa_model_repository).  The TensorRT models
+(for example /tmp/23.08/qa_model_repository).  The TensorRT models
 will be created for the GPU on the system that CUDA considers device 0
 (zero). If you have multiple GPUs on your system see the documentation
 in the scripts for how to target a specific GPU.

docs/user_guide/custom_operations.md

@@ -64,7 +64,7 @@ simple way to ensure you are using the correct version of TensorRT is
 to use the [NGC TensorRT
 container](https://ngc.nvidia.com/catalog/containers/nvidia:tensorrt)
 corresponding to the Triton container. For example, if you are using
-the 23.06 version of Triton, use the 23.06 version of the TensorRT
+the 23.08 version of Triton, use the 23.08 version of the TensorRT
 container.
 
 ## TensorFlow
@@ -123,7 +123,7 @@ simple way to ensure you are using the correct version of TensorFlow
 is to use the [NGC TensorFlow
 container](https://ngc.nvidia.com/catalog/containers/nvidia:tensorflow)
 corresponding to the Triton container. For example, if you are using
-the 23.06 version of Triton, use the 23.06 version of the TensorFlow
+the 23.08 version of Triton, use the 23.08 version of the TensorFlow
 container.
 
 ## PyTorch
@@ -167,7 +167,7 @@ simple way to ensure you are using the correct version of PyTorch is
 to use the [NGC PyTorch
 container](https://ngc.nvidia.com/catalog/containers/nvidia:pytorch)
 corresponding to the Triton container. For example, if you are using
-the 23.06 version of Triton, use the 23.06 version of the PyTorch
+the 23.08 version of Triton, use the 23.08 version of the PyTorch
 container.
 
 ## ONNX