News: 0001527938

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

Intel PyTorch Extnesion 2.6 Brings More Xeon 6 Optimizations

([Intel] 3 Hours Ago Intel PyTorch Extension)


Intel software engineers overnight released the newest version of the Intel Extension for PyTorch. The v2.6 release brings new optimizations for current-generation Xeon 6 processors as well as new large language model (LLM) performance optimizations.

The Intel PyTorch Extension 2.6 release builds upon the Xeon 6 optimizations originally offered in the prior v2.5 series. With the v2.6 release there are more comprehensive optimizations for AI models like Stable Diffusion, DLRM, BERT Large, GPT-J, Llama, and others. These optimizations will help with PyTorch on current P-core [1]Granite Rapids server processors and to a lesser extent the E-core [2]Sierra Forest processors.

The Intel PyTorch Extension 2.6 release also brings large language model optimizations for newer LLM models like Falcon 3, DeepSeek v2.5, and Jamba. Plus this update has some INT4 improvements, correctness fixes, upgrading to oneDNN 3.6.2, and other fixes and performance optimizations.

Downloads and more details on the Intel PyTorch Extension 2.6 release via [3]GitHub .



[1] https://www.phoronix.com/search/Granite+Rapids

[2] https://www.phoronix.com/search/Sierra+Forest

[3] https://github.com/intel/intel-extension-for-pytorch/releases/tag/v2.6.0%2Bcpu



phoronix

There is no hunting like the hunting of man, and those who have hunted
armed men long enough and liked it, never care for anything else thereafter.
-- Ernest Hemingway