NVIDIA TensorRT Inference Server

The NVIDIA TensorRT Inference Server provides a cloud inferencing solution optimized for NVIDIA 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.

What’s New in 1.11.0

  • The TensorRT backend is improved to have significantly better performance. Improvements include reducing thread contention, using pinned memory for faster CPU<->GPU transfers, and increasing compute and memory copy overlap on GPUs.

  • Reduce memory usage of TensorRT models in many cases by sharing weights across multiple model instances.

  • Boolean data-type and shape tensors are now supported for TensorRT models.

  • A new model configuration option allows the dynamic batcher to create “ragged” batches for custom backend models. A ragged batch is a batch where one or more of the input/output tensors have different shapes in different batch entries.

  • Local S3 storage endpoints are now supported for model repositories. A local S3 endpoint is specified as s3://host:port/path/to/repository.

  • The Helm chart showing an example Kubernetes deployment is updated to include Prometheus and Grafana support so that inference server metrics can be collected and visualized.

  • The inference server container no longer sets LD_LIBRARY_PATH, instead the server uses RUNPATH to locate its shared libraries.

  • Python 2 is end-of-life so all support has been removed. Python 3 is still supported.

Features

  • Multiple framework support. The server can manage any number and mix of models (limited by system disk and memory resources). Supports TensorRT, TensorFlow GraphDef, TensorFlow SavedModel, ONNX, PyTorch, and Caffe2 NetDef model formats. Also supports TensorFlow-TensorRT integrated models. Variable-size input and output tensors are allowed if supported by the framework. See Capabilities for detailed support information for each framework.

  • Concurrent model execution support. Multiple models (or multiple instances of the same model) can run simultaneously on the same GPU.

  • Batching support. For models that support batching, the server can accept requests for a batch of inputs and respond with the corresponding batch of outputs. The inference server also supports multiple scheduling and batching algorithms that combine individual inference requests together to improve inference throughput. These scheduling and batching decisions are transparent to the client requesting inference.

  • Custom backend support. The inference server allows individual models to be implemented with custom backends instead of by a deep-learning framework. With a custom backend a model can implement any logic desired, while still benefiting from the GPU support, concurrent execution, dynamic batching and other features provided by the server.

  • Ensemble support. An ensemble represents a pipeline of one or more models and the connection of input and output tensors between those models. A single inference request to an ensemble will trigger the execution of the entire pipeline.

  • Multi-GPU support. The server can distribute inferencing across all system GPUs.

  • The inference server provides multiple modes for model management. These model management modes allow for both implicit and explicit loading and unloading of models without requiring a server restart.

  • Model repositories may reside on a locally accessible file system (e.g. NFS), in Google Cloud Storage or in Amazon S3.

  • Readiness and liveness health endpoints suitable for any orchestration or deployment framework, such as Kubernetes.

  • Metrics indicating GPU utilization, server throughput, and server latency.

  • C library inferface allows the full functionality of the inference server to be included directly in an application.

Indices and tables