Meta will use Microsoft Azure's latest virtual machine (VM) series with up to 5,400 GPUs

Meta has chosen Azure as its strategic cloud provider, and Microsoft has found an important partner. According to Microsoft, the multinational tech conglomerate will use its services for AI research and development.

eta will use Microsoft Azure virtual machines for the first time in 2021. Due to its faster distributed AI training, Meta uses the NVIDIA A100 80GB GPU for its large-scale AI research, such as training its OPT-175B language model. Obviously, it can do more, so Meta decided to provide more machine learning training workloads for its AI research.

"As part of this deeper relationship, Meta will expand its use of Azure's supercomputing power to accelerate AI research and development for its Meta AI team," noted Sherry Wang, senior program manager for Azure HPC and AI. "Meta will use the latest virtual machine (VM) series in Azure (NDm A100 v4 series with NVIDIA A100 Tensor Core 80GB GPU), a dedicated Azure cluster of 5,400 GPUs for some of their large scale AI research workloads."

The collaboration between Microsoft and Meta doesn't stop there. The companies also plan to expand the use of PyTorch on Azure. Microsoft will create a PyTorch development accelerator to rapidly implement PyTorch-based solutions on Azure, Wang said . In addition, the company promises to provide ongoing enterprise-level support for PyTorch "to enable customers and partners to deploy PyTorch models in production environments both in the cloud and at the edge."

"We are excited to deepen our collaboration with Azure to advance Meta's AI research, innovation and open source efforts to benefit more developers around the world," said Jerome Pesenti, VP of Artificial Intelligence at Meta. "With the computing power of Azure and 1.6TB/s interconnect bandwidth per VM, we are able to accelerate the ever-increasing training needs to better accommodate larger, more innovative AI models. Additionally, we are excited to partner with Microsoft to bring our The experience extends to customers who use PyTorch from research to production." 

Post a Comment