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.

New Features and Improvements

• Added base python backend functionality for Windows.

• Removed the Wait/Read(avg) and Overhead metrics for gRPC in the Trace Summary Tool to avoid displaying inaccurate readings.

• OpenTelemetry trace mode switched to Batch Span Processor, which batches completed spans and sends them in bulk. This processor supports both size and time based batching. Size-based batching is controlled by 2 parameters: bsp_max_export_batch_size and bsp_max_queue_size, while time-based batching is controlled by bsp_schedule_delay.

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

Known Issues

• ONNX Runtime backend is not included with 24.02 release due to incompatibility reasons. However iGPU and Windows build assets shipped with ONNX Runtime backend.

• TensorRT-LLM backend is installed with Triton 24.01 base container due to incompatibility reasons.

• The 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 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: 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.

• 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.

• Python backend support for Windows is limited and does not currently support the following features:

  • GPU tensors
  • CPU and GPU-related metrics
  • Custom execution environments
  • The model load/unload APIs

Client Libraries and Examples

Ubuntu 22.04 builds of the client libraries and examples are included in this release in the attached v2.43.0_ubuntu22.04.clients.tar.gz file. The SDK is also available for as an Ubuntu 22.04 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 tritonserver2.43.0-sdk-win.zip file.

Windows Support

A beta release of Triton for Windows is provided in the attached file:tritonserver2.43.0-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 ONNXRuntime backend. The ONNX Runtime version is 1.16.3. The CPU, CUDA, and TensorRT execution providers are supported. The OpenVINO execution provider is not supported.

• OpenVINO models are supported. The OpenVINO version is 2023.3.0.

• 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:

• Python 3.8.10

• CUDA 12.3.2

• cuDNN 8.9.7.29

• TensorRT 8.6.1.6

Jetson iGPU Support

A release of Triton for IGX is provided in the attached tar file: tritonserver2.43.0-igpu.tgz.

• This release supports TensorFlow 2.15.0, TensorRT 8.6.2.3, Onnx Runtime 1.16.3, PyTorch 2.3.0a0+ebedce2, Python 3.10 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-2.43.0-py3-none-manylinux2014_aarch64.whl[all]