[sandGorgon]

>Intel Gaudi software integrates the PyTorch framework and provides optimized Hugging Face community-based models – the most-common AI framework for GenAI developers today. This allows GenAI developers to operate at a high abstraction level for ease of use and productivity and ease of model porting across hardware types.

what is the programming interface here ? this is not CUDA right ...so how is this being done ?

[wmf]

PyTorch has a bunch of backends including CUDA, ROCm, OneAPI, etc.

[sandGorgon]

i understand. but which backend is intel committing to ? not CUDA for sure. or have they created a new backend

[singhrac]

Intel makes oneAPI. They have corresponding toolkits to cuDNN like oneMKL, oneDNN, etc.

However the Gaudi chips are built on top of SynapseAI, another API from before the Habana acquisition. I don’t know if there’s a plan to support oneAPI on Gaudi, but it doesn’t look like it at the moment.