Optimizing Self-Driving DNNs on NVIDIA DRIVE PX with TensorRT
With the introduction of the Python API for NVIDIA TensorRT, researchers and developers can now optimize and serialize their deep neural networks (DNNs) using familiar Python-based code. In this webinar replay, we will demonstrate a workflow for deploying a trained DNN on NVIDIA DRIVE PX using INT8 precision, to achieve high throughput and low latency, while meeting accuracy requirements.
- Overview of TensorRT optimizer and runtime engine
- Overview of TensorRT 3 Python API
- Deep dive into INT8 calibration workflow with TensorRT 3 Python API
- Demo deployment of an automotive DNN in INT8 precision