Nsight Deep Learning Designer 2021.1
Today NVIDIA announced Nsight DL Designer – the first in-class integrated development environment to support efficient design of deep neural networks for in-app inference. Download Now!
This tool aims at streamlining the often iterative process of designing deep neural network models for in-app inferencing by providing efficient support at every stage of the process.
Nsight DL Designer is a GUI-based tool for model design and has integrated profiling capabilities that are based on GPU metrics. It provides a convenient way to import models into Pytorch for training. A visual analysis mode allows developers to dive deep into the inference process in real-time and in an interactive manner, with flexible options to export the finalized mode for inference deployment.
End-to-end Nsight DL Designer Workflow
Developers start by designing their deep neural network models inside Nsight DL Designer, using a built-in set of high-level neural network layers implemented by NVIDIA as the NvNeural inference engine. After creating the model, you can do performance profiling to get some basic idea whether your model meets the allocated timing budget. The profiling can be done early, even before you spend time on training the network.
For the training phase, Nsight DL Designer provides a variety of Python scripts that automatically converts a Nsight DL Designer model into a Pytorch model that can be easily added to your training loop. When training is done, you can save the learned weights data from your model into Numpy files. You can go back to DL Designer, load the weights file and enter the analysis mode to examine the inference results. The analysis mode also allows developers to dive deep into the inference process, visually inspect what’s happening at each inference step. This feedback can potentially guide developers to optimize their network model for improved quality and performance.
Once you are satisfied with both the quality and performance of your model, it’s time for deployment. Nsight DL Designer provides several ways to support deploying a model. One way is to export your model as an ONNX file. With the ONNX file, you can deploy your model on any platform where ONNX runtime is supported.
To learn more, watch the SIGGRAPH session: Announcing Nsight Deep Learning Designer – Optimize Your Neural Network for Quality and High Performance.
Key features for this release include:
- GUI based and NvNeural inference engine for model design
- Inference performance profiling with GPU metrics
- Interface with training framework – PyTorch
- Interactive visual analysis of the inference process
- Automated model export and code generation for deployment
Resources:
- Learn More and download DL Designer today.
- Find getting started documentation here.
- Post comments on Forums.
