post(TRI-1872): advance main to 26.09 upstreams

Dockerfile.sdk

@@ -29,7 +29,7 @@
 #
 
 # Base image on the minimum Triton container
-ARG BASE_IMAGE=nvcr.io/nvidia/tritonserver:26.08-py3-min
+ARG BASE_IMAGE=nvcr.io/nvidia/tritonserver:26.09-py3-min
 
 ARG TRITON_CLIENT_REPO_SUBDIR=clientrepo
 ARG TRITON_REPO_ORGANIZATION=http://github.com/triton-inference-server

README.md

@@ -29,8 +29,8 @@
 
 >[!WARNING]
 >You are currently on the `main` branch which tracks under-development progress
->towards the next release. The current release is version [2.72.0](https://github.com/triton-inference-server/server/releases/latest)
->and corresponds to the 26.08 container release on NVIDIA GPU Cloud (NGC).
+>towards the next release. The current release is version [2.73.0](https://github.com/triton-inference-server/server/releases/latest)
+>and corresponds to the 26.09 container release on NVIDIA GPU Cloud (NGC).
 
 # Triton Inference Server
 
@@ -90,16 +90,16 @@ Inference Server with the
 
 ```bash
 # Step 1: Create the example model repository
-git clone -b r26.08 https://github.com/triton-inference-server/server.git
+git clone -b r26.09 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:26.08-py3 tritonserver --model-repository=/models --model-control-mode explicit --load-model densenet_onnx
+docker run --gpus=1 --rm --net=host -v ${PWD}/model_repository:/models nvcr.io/nvidia/tritonserver:26.09-py3 tritonserver --model-repository=/models --model-control-mode explicit --load-model densenet_onnx
 
 # 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:26.08-py3-sdk /workspace/install/bin/image_client -m densenet_onnx -c 3 -s INCEPTION /workspace/images/mug.jpg
+docker run -it --rm --net=host nvcr.io/nvidia/tritonserver:26.09-py3-sdk /workspace/install/bin/image_client -m densenet_onnx -c 3 -s INCEPTION /workspace/images/mug.jpg
 
 # Inference should return the following
 Image '/workspace/images/mug.jpg':

build.py

@@ -74,7 +74,7 @@
 DEFAULT_TRITON_VERSION_MAP = {
     "release_version": "2.74.0dev",
     "triton_container_version": "26.10dev",
-    "upstream_container_version": "26.08",
+    "upstream_container_version": "26.09",
     "ort_version": "1.30.0",
     "ort_openvino_version": "2026.3.1",
     "standalone_openvino_version": "2026.3.1",

deploy/aws/values.yaml

@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:26.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:26.09-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: 2.72.0
+appVersion: 2.73.0
 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:26.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:26.09-py3
   pullPolicy: IfNotPresent
   numGpus: 1
   serverCommand: tritonserver
@@ -47,13 +47,13 @@ image:
     #
     # To set model control mode, uncomment and configure below
     # TODO: Fix the following url, it is invalid
-    # See https://github.com/triton-inference-server/server/blob/r26.08/docs/user_guide/model_management.md
+    # See https://github.com/triton-inference-server/server/blob/r26.09/docs/user_guide/model_management.md
     #  for more details
     #- --model-control-mode=explicit|poll|none
     #
     # Additional server args
     #
-    # see https://github.com/triton-inference-server/server/blob/r26.08/README.md
+    # see https://github.com/triton-inference-server/server/blob/r26.09/README.md
     #  for more details
 
 service:

deploy/gcp/values.yaml

@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:26.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:26.09-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:26.08-py3-sdk
+  - image: nvcr.io/nvidia/tritonserver:26.09-py3-sdk
     imagePullPolicy: Always
     name: nv-triton-client
     securityContext:

deploy/gke-marketplace-app/server-deployer/build_and_push.sh

@@ -28,8 +28,8 @@
 export REGISTRY=gcr.io/$(gcloud config get-value project | tr ':' '/')
 export APP_NAME=tritonserver
 export MAJOR_VERSION=2.67
-export MINOR_VERSION=2.72.0
-export NGC_VERSION=26.08-py3
+export MINOR_VERSION=2.73.0
+export NGC_VERSION=26.09-py3
 
 docker pull nvcr.io/nvidia/$APP_NAME:$NGC_VERSION
 

deploy/gke-marketplace-app/server-deployer/chart/triton/Chart.yaml

@@ -28,4 +28,4 @@ apiVersion: v1
 appVersion: "2.68"
 description: Triton Inference Server
 name: triton-inference-server
-version: 2.72.0
+version: 2.73.0

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/26.08
-publishedVersion: '2.72.0'
+modelRepositoryPath: gs://triton_sample_models/26.09
+publishedVersion: '2.73.0'
 gcpMarketplace: true
 
 image:
   registry: gcr.io
   repository: nvidia-ngc-public/tritonserver
-  tag: 26.08-py3
+  tag: 26.09-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: '2.72.0'
+  publishedVersion: '2.73.0'
   publishedVersionMetadata:
     releaseNote: >-
       Initial release.

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

@@ -27,7 +27,7 @@
 x-google-marketplace:
   schemaVersion: v2
   applicationApiVersion: v1beta1
-  publishedVersion: '2.72.0'
+  publishedVersion: '2.73.0'
   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/26.08
+    default: gs://triton_sample_models/26.09
   image.ldPreloadPath:
     type: string
     title: Leave this empty by default. Triton allows users to create custom layers for backend such as TensorRT plugin, 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:26.08-py3
+    -v ~:/scripts nvcr.io/nvidia/tensorrt:26.09-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/26.08/bert/1/model.plan
+gsutil cp bert_large_int8_bs1_s128.engine gs://triton_sample_models/26.09/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/26.08/` 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/26.09/` should be updated accordingly with the correct version.

deploy/k8s-onprem/values.yaml

@@ -30,7 +30,7 @@ tags:
   openshift: false
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:26.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:26.09-py3
   pullPolicy: IfNotPresent
   modelRepositoryServer: < Replace with the IP Address of your file server >
   modelRepositoryPath: /srv/models

deploy/oci/values.yaml

@@ -27,7 +27,7 @@
 replicaCount: 1
 
 image:
-  imageName: nvcr.io/nvidia/tritonserver:26.08-py3
+  imageName: nvcr.io/nvidia/tritonserver:26.09-py3
   pullPolicy: IfNotPresent
   modelRepositoryPath: s3://https://<OCI_NAMESPACE>.compat.objectstorage.<OCI_REGION>.oraclecloud.com:443/triton-inference-server-repository
   numGpus: 1

docs/customization_guide/compose.md

@@ -46,8 +46,8 @@ The `compose.py` script can be found in the
 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
- [r26.08](https://github.com/triton-inference-server/server/tree/r26.08)
-should be used to create a image based on the NGC 26.08 Triton release.
+ [r26.09](https://github.com/triton-inference-server/server/tree/r26.09)
+should be used to create a image based on the NGC 26.09 Triton release.
 
 `compose.py` provides `--backend`, `--repoagent` options that allow you to
 specify which backends and repository agents to include in the custom image.
@@ -78,20 +78,20 @@ For example, running
 ```
 python3 compose.py --backend pytorch --repoagent checksum
 ```
-on branch [r26.08](https://github.com/triton-inference-server/server/tree/r26.08) pulls:
-- `min` container `nvcr.io/nvidia/tritonserver:26.08-py3-min`
-- `full` container `nvcr.io/nvidia/tritonserver:26.08-py3`
+on branch [r26.09](https://github.com/triton-inference-server/server/tree/r26.09) pulls:
+- `min` container `nvcr.io/nvidia/tritonserver:26.09-py3-min`
+- `full` container `nvcr.io/nvidia/tritonserver:26.09-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 pytorch --repoagent checksum --container-version 26.08
+python3 compose.py --backend pytorch --repoagent checksum --container-version 26.09
 ```
 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 pytorch --repoagent checksum --image min,nvcr.io/nvidia/tritonserver:26.08-py3-min --image full,nvcr.io/nvidia/tritonserver:26.08-py3
+python3 compose.py --backend pytorch --repoagent checksum --image min,nvcr.io/nvidia/tritonserver:26.09-py3-min --image full,nvcr.io/nvidia/tritonserver:26.09-py3
 ```
 Method 1 and 2 will result in the same composed container. Furthermore,
 `--image` flag overrides the `--container-version` flag when both are specified.
@@ -102,8 +102,8 @@ Note:
 2. vLLM and TensorRT-LLM backends are currently not supported backends for
 `compose.py`. If you want to build additional backends on top of these backends,
 it would be better to [build it yourself](#build-it-yourself) by using
-`nvcr.io/nvidia/tritonserver:26.08-vllm-python-py3` or
-`nvcr.io/nvidia/tritonserver:26.08-trtllm-python-py3` as a `min` container.
+`nvcr.io/nvidia/tritonserver:26.09-vllm-python-py3` or
+`nvcr.io/nvidia/tritonserver:26.09-trtllm-python-py3` as a `min` container.
 
 
 ### CPU-only container composition

docs/getting_started/llm.md

@@ -55,7 +55,7 @@ as the example model.
 ### 1. Launch the container
 
 ```bash
-export RELEASE=26.08
+export RELEASE=26.09
 docker run --rm -it --net host --shm-size=2g --ulimit memlock=-1 --gpus all \
     -v ~/.cache/huggingface:/root/.cache/huggingface \
     nvcr.io/nvidia/tritonserver:${RELEASE}-trtllm-python-py3 bash

docs/introduction/compatibility.md

@@ -71,6 +71,7 @@
 
 | Triton release version	 | NGC Tag	 | Python version	 | vLLM version | CUDA version | CUDA Driver version | Size |
 | --- | --- | --- | --- | --- | --- | --- |
+| 26.09 | nvcr.io/nvidia/tritonserver:26.09-vllm-python-py3 | Python 3.12.3  | 0.29.0+3b480f30.nv26.9.68265578 | 13.4.1.012 | 615.71.09 | 10.57 GB |
 | 26.08 | nvcr.io/nvidia/tritonserver:26.08-vllm-python-py3 | Python 3.12.3  | 0.27.1+b1082979.nv26.8.62699174 | 13.4.1.012 | 615.61 | 10.65 GB |
 | 26.07 | nvcr.io/nvidia/tritonserver:26.07-vllm-python-py3 | Python 3.12.3  | 0.24.0+092c4842.nv26.7.57386851| 13.3.0.035 | 610.43.02 | 10.28 GB |
 | 26.06 | nvcr.io/nvidia/tritonserver:26.06-vllm-python-py3 | Python 3.12.3  | 0.22.1+7b9cb5b7.nv26.6.55098374 | 13.3.0.035 | 610.43.02 | 10.21 GB |
@@ -105,6 +106,7 @@
 
 | Triton release version	 | ONNX Runtime	 |
 | --- | --- |
+| 26.09 | 1.30.0 |
 | 26.08 | 1.28.0 |
 | 26.07 | 1.27.0 |
 | 26.06 | 1.24.4 |

docs/introduction/release_notes.md

@@ -25,9 +25,9 @@
 # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 -->
-# [Triton Inference Server Release 26.08](https://docs.nvidia.com/deeplearning/triton-inference-server/release-notes/rel-26-08.html#rel-26-08)
+# [Triton Inference Server Release 26.09](https://docs.nvidia.com/deeplearning/triton-inference-server/release-notes/rel-26-08.html#rel-26-08)
 
-The Triton Inference Server container image, release 26.08, is available
+The Triton Inference Server container image, release 26.09, is available
 on [NGC](https://ngc.nvidia.com/catalog/containers/nvidia:tritonserver) and
 is open source
 on [GitHub](https://github.com/triton-inference-server/server). Release notes can

docs/user_guide/performance_tuning.md

@@ -235,7 +235,7 @@ with a `tritonserver` binary.
 
 ```bash
 # Start server container
-docker run -ti --rm --gpus=all --network=host -v $PWD:/mnt --name triton-server nvcr.io/nvidia/tritonserver:26.08-py3
+docker run -ti --rm --gpus=all --network=host -v $PWD:/mnt --name triton-server nvcr.io/nvidia/tritonserver:26.09-py3
 
 # Start serving your models
 tritonserver --model-repository=/mnt/models
@@ -284,7 +284,7 @@ by setting the `-u` flag, such as `perf_analyzer -m densenet_onnx -u
 
 ```bash
 # Start the SDK container interactively
-docker run -ti --rm --gpus=all --network=host -v $PWD:/mnt --name triton-client nvcr.io/nvidia/tritonserver:26.08-py3-sdk
+docker run -ti --rm --gpus=all --network=host -v $PWD:/mnt --name triton-client nvcr.io/nvidia/tritonserver:26.09-py3-sdk
 
 # Benchmark model being served from step 3
 perf_analyzer -m densenet_onnx --concurrency-range 1:4

python/openai/README.md

@@ -46,7 +46,7 @@
 docker run -it --net=host --gpus all --rm \
   -v ${HOME}/.cache/huggingface:/root/.cache/huggingface \
   -e HF_TOKEN \
-  nvcr.io/nvidia/tritonserver:26.08-vllm-python-py3
+  nvcr.io/nvidia/tritonserver:26.09-vllm-python-py3
 ```
 
 2. Launch the OpenAI-compatible Triton Inference Server:
@@ -355,7 +355,7 @@ Currently, OpenAI-Compatible Frontend supports loading embedding models and embe
 docker run -it --net=host --gpus all --rm \
   -v ${HOME}/.cache/huggingface:/root/.cache/huggingface \
   -e HF_TOKEN \
-  nvcr.io/nvidia/tritonserver:26.08-vllm-python-py3
+  nvcr.io/nvidia/tritonserver:26.09-vllm-python-py3
 ```
 
 2. Launch the OpenAI-compatible Triton Inference Server:
@@ -451,7 +451,7 @@ docker run -it --net=host --gpus all --rm \
   -v ${HOME}/.cache/huggingface:/root/.cache/huggingface \
   -e HF_TOKEN \
   -e TRTLLM_ORCHESTRATOR=1 \
-  nvcr.io/nvidia/tritonserver:26.08-trtllm-python-py3
+  nvcr.io/nvidia/tritonserver:26.09-trtllm-python-py3
 ```
 
 2. Install dependencies inside the container:

qa/L2_build_presets/README.md

@@ -59,7 +59,7 @@ cd server/
 docker run --rm -v "$PWD":/workspace -w /workspace \
   -e TRITON_SERVER_BRANCH_NAME=<remote-branch> \
   -w /workspace/qa/L2_build_presets \
-  nvcr.io/nvidia/tritonserver:26.08-py3 python3 build_presets_test.py
+  nvcr.io/nvidia/tritonserver:26.09-py3 python3 build_presets_test.py
 ```
 
 ## Output

qa/common/gen_qa_model_repository

@@ -72,7 +72,7 @@ log_message.status "define: default values"
 # named the container train here and the semver in CI, so reading it was wrong
 # half the time.
 TRITON_CONTAINER_VERSION=${TRITON_CONTAINER_VERSION:=26.09dev}
-NVIDIA_UPSTREAM_VERSION=${NVIDIA_UPSTREAM_VERSION:=26.08}
+NVIDIA_UPSTREAM_VERSION=${NVIDIA_UPSTREAM_VERSION:=26.09}
 TRITON_SEMVER=${TRITON_SEMVER:=2.73.0dev}
 ONNX_VERSION=1.22.0 # https://github.com/microsoft/onnxruntime/blob/v1.28.0/tools/ci_build/github/linux/python/requirements.txt#L15
 # Max opset ONNX Runtime accepts: the last *released* opset of the onnx that ORT