From TokenPost · · 1m
Anyscale Forms Ray Group for AI Chip Support
Anyscale launches Ray interest group to boost AI chip compatibility across hardware types.
In brief
Anyscale launches Ray interest group to boost AI chip compatibility across hardware types. Anyscale created a Ray interest group to improve testing, hardware configuration awareness, and data transfer between AI chips. Originally reported by TokenPost.
Anyscale Launches Ray Interest Group
Anyscale formed a special interest group for Ray on Oct 20 to improve testing, job placement, and data transfers across AI chips.
Ray Supports Multiple Accelerators
Ray is an open-source framework that distributes Python and AI applications across clusters of computers and already supports over a dozen accelerator types including NVIDIA GPUs, AMD GPUs, TPUs, Trainium, and Gaudi.
Three Focus Areas for Improvement
The group will address development testing, hardware configuration awareness, and data movement between devices to solve practical problems when software runs across chips with different capabilities.
New PyTorch Support for Google TPUs
Anyscale and Google Cloud introduced native PyTorch support for Ray on Google TPUs on Oct 20, assigning each Ray worker to one TPU chip and scheduling jobs together on the same TPU slice.
The short version
Anyscale created a Ray interest group to improve testing, hardware configuration awareness, and data transfer between AI chips.
