Facebook makes preview PyTorch AI framework available

Get a free Techzine subscription!

Facebook has launched a developer preview of its PyTorch Artificial Intelligence (AI) framework. The framework helps accelerate the deployment of AI-based applications, reports Silicon Angle.

PyTorch was announced in May. The framework was developed by the Facebook AI research group as a machine learning library with functions for the Python programming language. It is mainly designed for the use of deep learning. The framework is intended to help accelerate the development of AI capabilities. Earlier it was used for more realistic avatars for the Oculus headset.

PyTorch works by combining the “modular production-oriented capabilities” of the Caffe2 AI framework with the deep learning model standard Open Neural Network Exchange. The Caffe2 AI framework was also developed by Facebook. Open Neural Network Exchange was created by Facebook, Amazon and Microsoft.

PyTorch 1.0 has also received some new features according to Facebook. These are in a hybrid front-end, which makes it possible to “track and script models from eager mode to graph mode”. This would help bridge the gap between experimentation and deployment. There is also an updated torch.distrubted library, which accelerates deep learning training in Python and C++ environments.

Adoption

Meanwhile, PyTorch’s ecosystem is growing rapidly, thanks in part to its rapid adoption by the developer community. For example, the large three public cloud providers support the project in various ways. For example, the AWS SageMaker platform for training and deploying machine learning models now supports “pre-configured environments for PyTorch”.

Google offers a new virtual machine image with PyTorch 1.0 for its Deep Learning VM service. The framework is also supported by Nvidia’s graphics processing units and cloud-based tensor processing units. This hardware is designed to speed up AI workloads. Microsoft will add PyTorch support for a number of cloud services, including Azure Machine Learning, Azure Notebooks and Visual Studio Code.

Facebook has also worked closely with hardware manufacturers to ensure that the PyTorch framework is supported on various computer chips and accelerators. These include Nvidia, ARM, IBM, Intel and Qualcomm.

This news article was automatically translated from Dutch to give Techzine.eu a head start. All news articles after September 1, 2019 are written in native English and NOT translated. All our background stories are written in native English as well. For more information read our launch article.