AMD and Intel aligned on oneDNN as the default open-source library for x86 acceleration within PyTorch.

The announcement concerns a shared software-library choice. According to an Intel website update relayed by financialjuice, AMD and Intel aligned on oneDNN as the default open-source library for x86 acceleration within PyTorch. The stated scope is x86 acceleration, not a common GPU software platform.

AMD also joined the Unified Acceleration Foundation (UXL Foundation) as a steering member. The update explicitly connects that membership to advancing the oneDNN work in PyTorch. It reports both the organizational participation and the library alignment.

Implementation and performance details are limited in the supplied source. No PyTorch release number, deployment schedule, supported-processor list, or benchmark result is provided. The announcement therefore establishes the alignment without quantifying a speedup.