The Facebook Artificial Intelligence Development Team has released the Preview Release of PyTorch 1.0. PyTorch is designed to make developing and testing AI applications more effective. For this, the development team has teamed up with the community and also receives support from other tech companies such as Google, Amazon and Microsoft.
What is PyTorch
PyTorch is an open-source programming library for the Python programming language designed to make developing and deploying applications in artificial intelligence and machine learning more effective and faster. Working closely with vendors such as ARM, Intel, IBM, NVIDIA or Qualcomm, developers can leverage a large ecosystem from PyTorch.
What’s new in PyTorch 1.0?
PyTorch has received a new hybrid frontend with the 1.0 Preview Release. This allows tracing and scripting of models from Eager mode to Graph mode, closing the gap between development and deployment in production. The torch.distributed library has been redesigned to allow faster training of AI in Python and C ++ environments. Last but not least, the C ++ interface for performance-critical research already known from the beta version is also available.
How exactly PyTorch works and how it works, explains Siraj Raval, Director of the “School of AI”, in the following video: