Facebook introduced several new products at the second annual PyTorch Developer’s Conference in San Francisco Thursday.
The event opened with a keynote by Facebook chief technology officer Mike Schroepfer, and it featured speakers from companies including Microsoft, Salesforce, Google, Tesla and Uber, as well as from academic institutions including Massachusetts Institute of Technology, Stanford University, California Institute of Technology, Carnegie Mellon University and Harvard University.
PyTorch said in a blog post that PyTorch citations in papers on ArXiv grew 194 percent in the first half of 2019, and the number of contributors to the platform is up over 50% over the past year, to nearly 1,200.
PyTorch added, “We are now advancing the platform further with the release of PyTorch 1.3, which includes experimental support for features such as seamless model deployment to mobile devices, model quantization for better performance at inference time and front-end improvements, like the ability to name tensors and create clearer code with less need for inline comments. We’re also launching a number of additional tools and libraries to support model interpretability and bringing multimodal research to production.”
A new end-to-end machine language workflow from Python to development on iOS and Android, PyTorch Mobile, was introduced at the conference, with a demo available on-site.
PyTorch wrote, “Running ML on edge devices is growing in importance as applications continue to demand lower latency. It is also a foundational element for privacy-preserving techniques such as federated learning. To enable more efficient on-device ML, PyTorch 1.3 now supports an end-to-end workflow from Python to deployment on iOS and Android.”
An early experimental release is currently being demonstrated, and coming releases will add optimization for size, performance and coverage improvements and a high-level application-programming interface covering common preprocessing and integration tasks needed for incorporating ML in mobile apps.
Other product introductions at the event included a new tool for secure and private artificial intelligence, a new tool for interpreting complex AI models, broad availability of support for Google Cloud TPU (Tensor Processing Unit), integration with Alibaba Cloud and a new object-detection platform, which is already being used by Facebook at scale.
PyTorch also released CrypTen, a community-based research platform for advancing privacy-preserving ML.
The platform wrote in a blog post, “Despite the AI community’s tremendous recent progress in advancing the applications of ML, there currently exist only very limited tools to build ML systems capable of working with encrypted data. This constrains the use of ML in domains that must be encrypted for security, such as work that involves sensitive medical information or data that people would prefer to encrypt simply for added privacy. Building secure ML systems to address these use cases today is difficult or even impossible because powerful, easy-to-use frameworks don’t work effectively with encrypted data.”
It added that CrypTen enables ML researchers, who typically aren’t cryptography experts, to easily experiment with ML models using secure computing techniques.
And Facebook AI Research unveiled Detectron2, a ground-up rewrite of its Detectron object-detection platform, writing in a blog post, “With a new, more modular design, Detectron2 is flexible and extensible and able to provide fast training on single or multiple GPU servers. Detectron2 includes high-quality implementations of state-of-the-art object detection algorithms, including DensePose, panoptic feature pyramid networks and numerous variants of the pioneering Mask R-CNN model family also developed by FAIR. Its extensible design makes it easy to implement cutting-edge research projects without having to fork the entire codebase.”
FAIR added, “We built Detectron2 to meet the research needs of Facebook AI and to provide the foundation for object detection in production use cases at Facebook. We are now using Detectron2 to rapidly design and train the next-generation pose detection models that power Smart Camera, the AI camera system in Facebook’s Portal video-calling devices. By relying on Detectron2 as the unified library for object detection across research and production use cases, we are able to rapidly move research ideas into production models that are deployed at scale.”
Finally, Captum, a model interpretability library for PyTorch, is being open-sourced.
PyTorch wrote in a blog post, “Captum supports model interpretability across modalities such as vision and text, and its extensible design allows researchers to add new algorithms. Captum also allows researchers to quickly benchmark their work against other existing algorithms available in the library. For model developers, Captum can be used to improve and troubleshoot models by facilitating the identification of different features that contribute to a model’s output in order to improve their design and troubleshoot unexpected outputs.”