Triton Inference Server Release 23.11

The Triton Inference Server container image, release 23.11, is available on NGC and is open source on GitHub.

Contents of the Triton Inference Server container

The Triton Inference Server Docker image contains the inference server executable and related shared libraries in /opt/tritonserver.

For a complete list of what the container includes, refer to Deep Learning Frameworks Support Matrix.

The container also includes the following:

Driver Requirements

Release 23.11 is based on CUDA 12.3.0, which requires NVIDIA Driver release 545 or later. However, if you are running on a data center GPU (for example, T4 or any other data center GPU), you can use NVIDIA driver release 450.51 (or later R450), 470.57 (or later R470), 510.47 (or later R510), or 525.85 (or later R525) 535.86 (or later R535), or 545.23 (or later R545).

The CUDA driver's compatibility package only supports particular drivers. Thus, users should upgrade from all R418, R440, R460, and R520 drivers, which are not forward-compatible with CUDA 12.3. For a complete list of supported drivers, see the CUDA Application Compatibility topic. For more information, see CUDA Compatibility and Upgrades.

GPU Requirements

Release 23.11 supports CUDA compute capability 6.0 and later. This corresponds to GPUs in the NVIDIA Pascal, NVIDIA Volta™, NVIDIA Turing™, NVIDIA Ampere architecture, NVIDIA Hopper™, and NVIDIA Ada Lovelace architecture families. For a list of GPUs to which this compute capability corresponds, see CUDA GPUs. For additional support details, see Deep Learning Frameworks Support Matrix.

Key Features and Enhancements

This Inference Server release includes the following key features and enhancements.

  • Starting with the 23.11 release, Triton containers supporting iGPU architectures are published, and run on Jetson devices. Please refer to the Frameworks Support Matrix for information regarding which iGPU hardware/software is supported by which container.
  • Implicit state management has been enhanced to support growing buffers and use a single buffer for both input and output states.
  • Sequence batcher has been enhanced to support iterative scheduling.
  • The backend API has been enhanced to support rescheduling a request.Currently, only Python backend and Custom C++ backends support request rescheduling.
  • TRT-LLM backend now supports request cancellation.
  • Configuration of a vLLM backend model can now be auto-completed by Triton. The user just needs to pass backend: "vllm" to leverage the auto-complete feature.
  • Python backend now supports parameters in BLS requests.
  • Python backend GPU tensor support has been improved to provide better performance.
  • A new tutorial demonstrating how to deploy LLaMa2 using TRT-LLM has been added.
  • Added benchmarking script for profiling LLMs using Perf Analyzer
  • The HTTP endpoint has been enhanced to support access restriction.
  • Secure Deployment Guide has been added to provide guidance on deploying Triton securely.
  • The client model loading API no longer allows uploading files outside the model repository.
  • DCGM version has been upgraded to 3.2.6.
  • The Kubernetes Deploy example now supports Kubernetes’ new StartupProbe to allow Triton pods time to finish startup before running health probes.

NVIDIA Triton Inference Server Container Versions

The following table shows what versions of Ubuntu, CUDA, Triton Inference Server, and NVIDIA TensorRT™ are supported in each of the NVIDIA containers for Triton Inference Server. For older container versions, refer to the Frameworks Support Matrix.

Container Version Triton Inference Server Ubuntu CUDA Toolkit TensorRT
23.11 2.40 22.04 NVIDIA CUDA 12.3.0 TensorRT
23.10 2.39.0 NVIDIA CUDA 12.2.2
23.09 2.38.0 NVIDIA CUDA 12.2.1
23.08 2.37.0
23.07 2.36.0 NVIDIA CUDA 12.1.1
23.06 2.35.0
23.05 2.34.0 TensorRT
23.04 2.33.0 20.04 NVIDIA CUDA 12.1.0 TensorRT 8.6.1
23.03 2.32.0 TensorRT 8.5.3
23.02 2.31.0 NVIDIA CUDA 12.0.1
23.01 2.30.0 TensorRT
22.12 2.29.0 NVIDIA CUDA 11.8.0 TensorRT 8.5.1
22.11 2.28.0
22.10 2.27.0 TensorRT 8.5 EA
22.09 2.26.0
22.08 2.25.0 NVIDIA CUDA 11.7.1 TensorRT
22.07 2.24.0 NVIDIA CUDA 11.7 Update 1 Preview TensorRT 8.4.1
22.06 2.23.0 TensorRT 8.2.5
22.05 2.22.0 NVIDIA CUDA 11.7.0
22.04 2.21.0 NVIDIA CUDA 11.6.2 TensorRT and

for x86 Linux and SBSA

TensorRT 8.4.0 for JetPack/Jetson

22.03 2.20.0 NVIDIA CUDA 11.6.1 TensorRT 8.2.3 and

for x86 Linux and SBSA

TensorRT 8.4.0 for JetPack/Jetson

22.02 2.19.0 NVIDIA CUDA 11.6.0 TensorRT 8.2.3
22.01 2.18.0 TensorRT 8.2.2
21.12 2.17.0 NVIDIA CUDA 11.5.0 TensorRT
21.11 2.16.0

TensorRT for x64 Linux

TensorRT for ARM SBSA Linux
21.10 2.15.0 NVIDIA CUDA 11.4.2 with cuBLAS
21.09 2.14.0 NVIDIA CUDA 11.4.2 TensorRT 8.0.3
21.08 2.13.0 NVIDIA CUDA 11.4.1 TensorRT
21.07 2.12.0 NVIDIA CUDA 11.4.0
21.06.1 2.11.0 NVIDIA CUDA 11.3.1 TensorRT
21.05 2.10.0 NVIDIA CUDA 11.3.0
21.04 2.9.0
21.03 2.8.0 NVIDIA CUDA 11.2.1 TensorRT
21.02 2.7.0 NVIDIA CUDA 11.2.0 TensorRT
20.12 2.6.0 NVIDIA CUDA 11.1.1 TensorRT 7.2.2
20.11 2.5.0


NVIDIA CUDA 11.1.0 TensorRT 7.2.1
20.10 2.4.0
20.09 2.3.0 NVIDIA CUDA 11.0.3 TensorRT 7.1.3


20.07 1.15.0


NVIDIA CUDA 11.0.194
20.06 1.14.0


NVIDIA CUDA 11.0.167 TensorRT 7.1.2
20.03.1 1.13.0 NVIDIA CUDA 10.2.89 TensorRT 7.0.0
20.03 1.12.0






1.9.0 TensorRT 6.0.1
19.10 1.7.0 NVIDIA CUDA 10.1.243
19.09 1.6.0
19.08 1.5.0 TensorRT 5.1.5

Known Issues

  • When using the generate streaming endpoint, Triton will segfault if the client closes the connection before all responses have been generated. The fix will be available in the next release.
  • Reuse-grpc-port and reuse-http-port are now properly parsed as booleans. 0 and 1 will continue to work as values. Any other integers will throw an error.
  • TensorRT-LLM backend provides limited support of Triton extensions and features.
  • The TensorRT-LLM backend may core dump on server shutdown. This impacts server teardown only and will not impact inferencing.
  • When using decoupled models, there is a possibility that response order as sent from the backend may not match with the order in which these responses are received by the streaming gRPC client. Note that this only applies to responses from different requests. Any responses corresponding to the same request will still be received in their expected order, relative to each other.
  • The FasterTransformer backend is only officially supported for 22.12, though it can be built for Triton container versions up to 23.07.
  • The Java CAPI is known to have intermittent segfaults we’re looking for a root cause.
  • Some systems which implement malloc() may not release memory back to the operating system right away causing a false memory leak. This can be mitigated by using a different malloc implementation. Tcmalloc and jemalloc are installed in the Triton container and can be used by specifying the library in LD_PRELOAD. We recommend experimenting with both tcmalloc and jemalloc to determine which one works better for your use case.
  • Auto-complete may cause an increase in server start time. To avoid a start time increase, users can provide the full model configuration and launch the server with --disable-auto-complete-config.
  • Auto-complete does not support PyTorch models due to lack of metadata in the model. It can only verify that the number of inputs and the input names matches what is specified in the model configuration. There is no model metadata about the number of outputs and datatypes. Related PyTorch bug:
  • Triton Client PIP wheels for ARM SBSA are not available from PyPI and pip will install an incorrect Jetson version of Triton Client library for Arm SBSA. The correct client wheel file can be pulled directly from the Arm SBSA SDK image and manually installed.
  • Traced models in PyTorch seem to create overflows when int8 tensor values are transformed to int32 on the GPU. Refer to pytorch/pytorch#66930 for more information.
  • Triton cannot retrieve GPU metrics with MIG-enabled GPU devices (A100 and A30).
  • Triton metrics might not work if the host machine is running a separate DCGM agent on bare-metal or in a container.
  • When cloud storage (AWS, GCS, AZURE) is used as a model repository and a model has multiple versions, Triton creates an extra local copy of the cloud model’s folder in the temporary directory, which is deleted upon server’s shutdown.