Facebook Fuses Research and Production AI Projects through PyTorch 1.0

Facebook announced the upcoming research and production versions of PyTorch 1.0. The artificial intelligence framework combines production capabilities of Caffe2 and ONNX with PyTorch's research focused design. Together, developers gain a quick, frictionless path from prototyping to production for AI implementations.

Facebook will launch beta versions of PyTorch 1.0 in the coming months. Alongside the beta, Facebook will publish tools, libraries, pre-trained models, and datasets for research and production environments. Although PyTorch 1.0 is not yet in beta for the broader developer community, Facebook already uses the framework to power many Facebook products (e.g. 6 billion text translations each day).

Moving AI projects from research to production has been a challenge. PyTorch has been immensely useful for research level projects. However, for production, projects needed to move to Caffe2 to run at scale. This transition was never easy, and the Open Neural Network Exchange (ONNX) was one attempt to ease the process. PyTorch 1.0 fuses both processes into a single framework.

To keep up to date with Facebook's AI progress, stay tuned to its AI developers site. Further, those interested in all that PyTorch has to currently offer, visit the PyTorch blog. The Caffe2 blog also contains helpful updates and insights.

Be sure to read the next Artificial Intelligence article: Kairos Introduces On-Premise Facial Recognition API

 

Comments (0)