> only to learn that the "Iris GPU" is not supported You can use the CPU backend, which has gotten significant optimization since ~v1.2. You can probably use t

> only to learn that the "Iris GPU" is not supported You can use the CPU backend, which has gotten significant optimization since ~v1.2. You can probably use the integrated GPU with the OpenCL backend, but I am not sure it will be faster than a modern AVX2/3 CPU. > because Intel doesn't give a fuck about DL. Why do you think that? https://www.intelnervana.com/ https://www.intelnervana.com/ https://www.intel.in/content/www/in/en/events/ai-day.html https://www.intel.in/content/www/in/en/events/ai-day.html https://github.com/01org/mkl-dnn https://github.com/01org/mkl-dnn