{"id":450,"date":"2018-10-17T11:40:59","date_gmt":"2018-10-17T11:40:59","guid":{"rendered":"https:\/\/www.appservgrid.com\/paw92\/index.php\/2018\/10\/17\/pytorch-1-0-preview-release-facebooks-newest-open-source-ai\/"},"modified":"2018-10-17T11:40:59","modified_gmt":"2018-10-17T11:40:59","slug":"pytorch-1-0-preview-release-facebooks-newest-open-source-ai","status":"publish","type":"post","link":"https:\/\/www.appservgrid.com\/paw92\/index.php\/2018\/10\/17\/pytorch-1-0-preview-release-facebooks-newest-open-source-ai\/","title":{"rendered":"PyTorch 1.0 Preview Release: Facebook\u2019s newest Open Source AI"},"content":{"rendered":"

Last updated October 4, 2018 By Avimanyu Bandyopadhyay<\/a> Leave a Comment<\/a> <\/p>\n

<\/p>\n

Facebook already uses its own Open Source AI, PyTorch quite extensively in its own artificial intelligence projects. Recently, they have gone a league ahead by releasing a pre-release preview version 1.0. <\/p>\n

For those who are not familiar, PyTorch<\/a> is a Python-based library for Scientific Computing.<\/p>\n

PyTorch harnesses the superior computational power of Graphical Processing Units (GPUs)<\/a> for carrying out complex Tensor<\/a> computations and implementing deep neural networks<\/a>. So, it is used widely across the world by numerous researchers and developers.<\/p>\n

This new ready-to-use Preview Release<\/a> was announced at the PyTorch Developer Conference<\/a> at The Midway<\/a>, San Francisco, CA on Tuesday, October 2, 2018.<\/p>\n

Highlights of PyTorch 1.0 Release Candidate<\/h2>\n

\"PyTorhc<\/p>\n

Some of the main new features in the release candidate are:<\/p>\n

1. JIT<\/h3>\n

JIT is a set of compiler tools to bring research close to production. It includes a Python-based language called Torch Script and also ways to make existing code compatible with itself.<\/p>\n

2. New torch.distributed library: \u201cC10D\u201d
<\/h3>\n

\u201cC10D\u201d enables asynchronous operation on different backends with performance improvements on slower networks and more.<\/p>\n

3. C++ frontend (experimental)
<\/h3>\n

Though it has been specifically mentioned as an unstable API (expected in a pre-release), this is a pure C++ interface to the PyTorch backend that follows the API and architecture of the established Python frontend to enable research in high performance, low latency and C++ applications installed directly on hardware.<\/p>\n

To know more, you can take a look at the complete update notes<\/a> on GitHub.<\/p>\n

The first stable version PyTorch 1.0 will be released in summer.<\/p>\n

Installing PyTorch on Linux<\/h2>\n

To install PyTorch v1.0rc0, the developers recommend using conda<\/a> while there also other ways to do that as shown on their local installation page<\/a> where they have documented everything necessary in detail.<\/p>\n

Prerequisites<\/h3>\n