NVIDIA Nsight Deep Learning (DL) Designer is an integrated development environment that helps developers efficiently design and optimize deep neural networks for high-performance inference. Built atop the industry-standard ONNX model format and popular inference solutions like TensorRT™, DL Designer provides a comprehensive suite of tools for model design, optimization, and performance profiling.

Features

• GUI for ONNX Model Design and Optimization: DL Designer offers a GUI-based tool for editing and creating ONNX models, making the process visible and intuitive. It integrates with other tools, including user-defined ones, for quick and easy whole model transformations.
• Built-in Profiler for Performance Evaluations: Ships with both an ONNX Runtime profiler and a TensorRT profiler, allowing developers to evaluate a model’s inference performance profile while making changes.
• Integration with TensorRT: Includes NVIDIA’s TensorRT (10.7) inference engine, enabling easy conversion of ONNX models into TensorRT engines with a GUI frontend.
• Efficient Model Design Without Coding: Supports ONNX version 1.15 (opset 20) and features like Local Functions and FP8. Advanced users can create models from scratch entirely within DL Designer.
• Inference Performance Profiling with GPU Metrics: Provides detailed profiling data based on GPU metrics like SM utilization, Tensor Core utilization, and SM occupancy, helping developers optimize models for specific GPU families.
• Easy Model Export for Deployment: Allows models to be exported as ONNX files for platform-independent deployment or as TensorRT engines for high-performance inferencing.

Benefits

• Enhanced Model Design: Simplifies the design and optimization of deep neural networks with a user-friendly GUI.
• Comprehensive Profiling: Offers detailed performance evaluations and profiling data to identify and address bottlenecks.
• Seamless Integration: Integrates with TensorRT and other ONNX tools for streamlined model transformations and optimizations.
• Flexible Deployment: Supports flexible model export options for various deployment scenarios.
• Developer-Friendly: Provides a full-fledged editor for ONNX models, reducing the need for coding and enabling quick iterations.