Start Creating Custom AI Models with the NVIDIA TAO Toolkit

NVIDIA TAO Toolkit Version 3.22.05: What’s New

NVIDIA TAO Toolkit is a low-code AI model development solution that uses the power of transfer learning to simplify and accelerate the creation of custom, production-ready AI models. The new release makes it easy to:

  • “Bring your own” ONNX models weights into TAO for fine-tuning and optimizing.
  • Deploy TAO Toolkit as-a-Service in a modern, cloud-native infrastructure on Kubernetes and integrate it with REST APIs.
  • Visualize the model training progress such as training loss, validation loss, and histogram of model weights in TensorBoard.
  • Access new vision pretrained models, including Point Cloud, Pose Action Classification, and 3D Pose Estimation.
  • Take advantage of new speech pretrained models, including HiFi Gan and FastPitch for deploying custom text-to-speech applications.

Get Started With TAO Toolkit

Introductory Whitepaper

Learn how NVIDIA TAO Toolkit and pretrained models can transform your development efforts.

Read Whitepaper

New Developer Blog

Fastrack AI-Powered Robotic Applications with Synthetic Data and Transfer Learning

Read Blog

GTC Webinar

Learn how to create and deploy custom, production-ready vision AI and conversational AI models without any expertise in AI.

Watch Now


Test Drive the NVIDIA TAO Toolkit for free on NVIDIA LaunchPad.

Sign Up Now

Developer Starter Resources

Vision AI

Developer Blogs
Sample Applications

To convert TAO Toolkit model (etlt) to an NVIDIA® TensorRT™ engine for deployment with DeepStream, select the appropriate TAO-converter for your hardware and software stack.

Download Resources

Additional Resources

Product Support

Ethical AI

NVIDIA’s platforms and application frameworks enable developers to build a wide array of AI applications. Consider potential algorithmic bias when choosing or creating the models being deployed. Also, work with the model’s developer to ensure that it meets the requirements for the relevant industry and use case; that the necessary instruction and documentation are provided to understand error rates, confidence intervals, and results; and that the model is being used under the conditions and in the manner intended.