AMD and Intel to consolidate PyTorch x86 CPU optimizations into oneDNN
Overview
AMD and Intel say they will integrate their x86 CPU optimizations for PyTorch into the open-source oneDNN library, making it the default open-source acceleration library for PyTorch on x86 CPUs.
AMD will add its ZenDNN optimizations to oneDNN and join the UXL Foundation's steering committee.
The companies also plan to bring AI Compute Extensions (ACE), an x86 instruction set aimed at matrix multiplication, into PyTorch through future oneDNN updates. The announcement was made on October 20, according to IThome. No release timeline has been given.
Written by AI from the articles below · updated Oct 21, 5:45 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- IThome · AINewsAMD and Intel team up to unify PyTorch x86 CPU optimization via oneDNN
AIAMD and Intel say they will integrate their x86 CPU optimizations for PyTorch into the open-source oneDNN library, making it the default open-source library for x86 acceleration in PyTorch. AMD will add its ZenDNN optimizations to oneDNN and join the UXL Foundation's steering committee. The companies also plan to bring their AI Compute Extensions (ACE), an x86 instruction set aimed at matrix multiplication, into PyTorch through future oneDNN updates.
Heat Trend
Not enough continuous observations to show a trend yet.