Microsoft deepens cooperation with Meta on AI and PyTorch

Microsoft announced the further expansion of its AI partnership with Meta, which has selected Azure as a strategic cloud provider to help accelerate AI research and development. In 2017, Microsoft and Meta (also known as Facebook at the time) jointly launched ONNX (Open Neural Network Exchange), an open ecosystem of deep learning development tools designed to allow developers to leverage different AI frameworks. Inter-shift deep learning model. In 2018, Microsoft announced the open source of ONNX Runtime, the inference engine for ONNX format models.

As part of this deepening cooperation, Meta plans to use Azure to accelerate the research and development of the entire Meta AI. Meta will leverage the latest virtual machine (VM) series in Azure (NDm A100 v4 series with NVIDIA A100 Tensor Core 80GB GPU) to use a dedicated Azure cluster of 5400 GPUs for some of their large-scale AI research workloads . As early as last year, Meta had begun using Azure VMs to conduct some large-scale AI research, especially distributed AI training.

Additionally, Meta and Microsoft will collaborate to expand PyTorch adoption on Azure and accelerate the developer journey from experimentation to production. "Azure provides PyTorch users with a comprehensive top-down stack through best-in-class hardware (NDv4 and Infiniband)."

Microsoft said they will build new PyTorch development in the next few months Accelerator to facilitate rapid implementation of PyTorch-based solutions on Azure. and continues to provide enterprise-grade support for PyTorch, enabling customers and partners to deploy PyTorch models in production environments in the cloud and edge.

According to the introduction, AI development has been an important theme at this week’s Microsoft Build conference. One of the biggest features of Project Volterra, the developer device Microsoft created in partnership with Qualcomm that we previously reported on, is its ability to run AI workloads faster and more efficiently on Arm than other processors.

Microsoft also announced a new Power Apps feature called Express Design at the conference, which allows developers to convert images into images by using Microsoft cognitive AI technology to scan input and generate working app controls backed by data storage. , documents, Figma design files and PowerPoint into apps. As well as the launch of the Azure OpenAI Service preview, users invited to the preview will be able to access different models from OpenAI; including the GPT-3 base series (Ada, Babbage, Curie and DaVinci), Codex series and embedded models, and connect them with Azure Enterprise Combination of abilities.

Currently, Microsoft’s AI model set includes Turing for rich language understanding, Z-Code for language translation, and Florence for visual recognition. Its Azure Cognitive Services family includes building-block services for vision, speech, language, and decision-making, as well as its OpenAI service. Existing Azure Applied AI Services include cognitive search, form recognizer, immersive reader, bot service, and video analyzer.

This article is reproduced from OSCHINA

The title of this article: Microsoft deepens cooperation with Meta on AI and PyTorch

The address of this article: https://www.oschina.net/news/197531/ microsoft-meta-ai-pytorch

The above is the detailed content of Microsoft deepens cooperation with Meta on AI and PyTorch. For more information, please follow other related articles on the PHP Chinese website!