Creating Custom, Production-Ready AI Models Faster with NVIDIA TAO

Learn about the latest updates to NVIDIA TAO, an AI-model-adaptation framework, and NVIDIA TAO toolkit, a CLI and Jupyter notebook-based version of TAO.

All AI applications are powered by models. Models can help spot defects in parts, detect the early onset of disease, translate languages, and much more. But building custom models for a specific use requires mountains of data and an army of data scientists. 

NVIDIA TAO, an AI-model-adaptation framework, simplifies and accelerates the creation of AI models. By fine-tuning state-of-the-art, pretrained models, you can create custom, production-ready computer vision and conversational AI models. This can be done in hours rather than months, eliminating the need for large training data or AI expertise.

The latest version of the TAO toolkit is now available for download. The TAO toolkit, a CLI and Jupyter notebook-based version of TAO, brings together several new capabilities to help you speed up your model creation process. 

Key highlights 

  • New computer vision model for 2D and 3D Action Recognition. 
  • New text-to-speech models: HifiGAN and FastPitch.
  • New computer vision model architectures: EfficientDet and YoloV4 Tiny.
  • Increased utilization of GPUs for faster, more efficient training.

We are also taking TAO to the next level and making it a lot easier to create custom, production-ready models. A graphical user interface version of TAO is currently under development that epitomizes a zero-code model development solution. This creates the ability to train, adapt, and optimize computer vision and conversational AI models without writing a single line of code. 

Early access is slated for early 2022. Sign up today!

Source:: NVIDIA