Under the Hood - A PyTorch Engine for Neural Networks in Wolfram Language

Explore the technical integration of PyTorch as a backend engine for Wolfram Language's neural network framework in this 27-minute conference talk. Discover how the Wolfram Neural Net Repository is being migrated to leverage PyTorch's capabilities, resulting in enhanced hardware support and improved performance for both training and inference operations. Learn about the sophisticated process of connecting PyTorch with Wolfram Language's symbolic neural network definitions, which enables hybrid execution models and creates more streamlined workflows for researchers working at the intersection of symbolic computation and deep learning.