Triton Inference Server

The Triton Inference Server provides a cloud inferencing solution optimized for both CPUs and GPUs. The server provides an inference service via an HTTP or GRPC endpoint, allowing remote clients to request inferencing for any model being managed by the server. For edge deployments, Triton Server is also available as a shared library with an API that allows the full functionality of the server to be included directly in an application.

What's New in 2.32.0

• Added the Parameters Extension which allows an inference request to provide custom parameters that cannot be provided as inputs. These parameters can be used in the python backend as described here.

• Added support for models that use decoupled API for Business Scripting Logic (BLS) in Python backend. Examples can be found here.

• The same model name can be used across different repositories if the --model-namespacing flag is set.

• Triton’s Response Cache feature has been converted internally to a shared library implementation of the new TRITONCACHE APIs, similar to how backends and repo agents are used today. The default cache implementation is local_cache, which is equivalent to the fixed-size in-memory buffer implementation used before. The --response-cache-byte-size flag will continue to function in the same way, but the --cache-config flag will be the preferred method of cache configuration moving forward. For more information, see the cache documentation here.

• Triton’s trace tool now supports tracing for request_id.

• Refer to the 23.03 column of the Frameworks Support Matrix for container image versions on which the 23.03 inference server container is based.

Known Issues

• Support for TensorFlow1 will be removed starting from 23.04.

• Triton Inferentia guide is out of date. Some users have reported issues with running Triton on AWS Inferentia instances.

• 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 is installed in the Triton container and can be used by specifying the library in LD_PRELOAD.

• 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: pytorch/pytorch#38273

• 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.

Client Libraries and Examples

Ubuntu 20.04 builds of the client libraries and examples are included in this release in the attached v2.32.0_ubuntu2004.clients.tar.gz file. The SDK is also available for as an Ubuntu 20.04 based NGC Container. The SDK container includes the client libraries and examples, Performance Analyzer and Model Analyzer. Some components are also available in the tritonclient pip package. See Getting the Client Libraries for more information on each of these options.

For windows, the client libraries and some examples are available in the attached tritonserver2.32.0-sdk-win.zip file.

Windows Support

A beta release of Triton for Windows is provided in the attached file:tritonserver2.32.0-win.zip. This is a beta release so functionality is limited and performance is not optimized. Additional features and improved performance will be provided in future releases. Specifically in this release:

• HTTP/REST and GRPC endpoints are supported.

• ONNX models are supported by the ONNX Runtime backend. The ONNX Runtime version is 1.14.1. The CPU, CUDA, and TensorRT execution providers are supported. The OpenVINO execution provider is not supported.

• OpenVINO models are supported. The OpenVINO version is 2021.4.

• Prometheus metrics endpoint is not supported.

• System and CUDA shared memory are not supported.

To use the Windows version of Triton, you must install all the necessary dependencies on your Windows system. These dependencies are available in the Dockerfile.win10.min. The Dockerfile includes the following CUDA-related components:

• CUDA 11.8.0

• cuDNN 8.8.1.3

• TensorRT 8.5.3.1

Jetson Jetpack Support

A release of Triton for JetPack is provided in the attached tar file: tritonserver2.32.0-jetpack5.1.tgz.

• This release supports TensorFlow 2.11.0, TensorFlow 1.15.5, TensorRT 8.5.2.2, Onnx Runtime 1.14.1, PyTorch2.0.0, Python 3.8 and as well as ensembles.
• Onnx Runtime backend does not support the OpenVino and TensorRT execution providers. The CUDA execution provider is in Beta.
• System shared memory is supported on Jetson. CUDA shared memory is not supported.
• GPU metrics, GCS storage, S3 storage and Azure storage are not supported.

The tar file contains the Triton server executable and shared libraries and also the C++ and Python client libraries and examples. For more information on how to install and use Triton on JetPack refer to jetson.md.

The wheel for the Python client library is present in the tar file and can be installed by running the following command:

python3 -m pip install --upgrade clients/python/tritonclient-2.32.0-py3-none-manylinux2014_aarch64.whl[all]