Facebook is planning to open source PyTorch, the deep learning framework, which will assist developers to use its AI models in production systems and not just for research work.
On the first day of F8 conference(just concluded), Facebook focused on rebuilding its image and relationship with users claiming to take “broader responsibility.” On the other hand, the company used day two to emphasize on the technical advancements the company has made over the past one year.
Facebook says the release of PyTorch 1.0 will help developers to move from research to production in a single framework seamlessly.
It also claims that PyTorch’s front end allows accelerated prototyping and scope of experimentation due to its flexible programming model.
Last year, Facebook collaborated with Microsoft to create the Open Neural Network Exchange (ONNX) that was designed to facilitate smooth movement between frameworks.
Now they have integrated ONNX into PyTorch1.0 so that these platforms can be used interchangeably with other frameworks and developers can use them as per requirements.
Facebook has also revealed its plan of merging PyTorch 0.4 and Caffe2 codebases over the course of next few months to generate a consolidated framework that would support multiple features.
The beta version of PyTorch is slated to arrive later this year. It’ll be a collection of open source tools, libraries, datasets, training models, etc.