With this new partnership, Meta AI will look to use more of the supercomputing power of Azure to further its goal of AI research and development. Employing the new Virtual Machine (VM) Metal will take advantage of the 5400 GPUs of the Azure cluster to carry out their heavy AI R&D workload.
In 2021, Impressed by the performance of Azure, Meta started to use it for their research. Azure stands out from other cloud services since they offer four times more performance bandwidth among virtual machines than any other cloud service provider.
Azure’s new NDmA100 V4 VM series lets users organise clusters irrespective of their size dynamically and automatically from a couple of GPUs to 1000s. This model also allows users to pause while conducting experiments.
Furthermore, Microsoft and Meta will join hands to improve the adoption of PyTorch on the Azure platform. This would give developers the liberty to experiment and produce. Microsoft will develop new development accelerators for PyTorch to improve the fast execution of solutions based on PyTorch. Microsoft has also promised to provide extensive support and guidance for PyTorch so that users can use models of Pytorch either on edge or cloud or both.
By using the supercomputing power of Azure for AI model training for premier AI research institutes and improving the existing resources and tools for experimentation and collaboration, new opportunities will be unlocked for developers.