Open shailesh837 opened 6 months ago
Just double checking. Are you asking to always change the default or just for an individual run? For the latter you can set the ONEAPI_DEVICE_SELECTOR variable to something. Normally GPUs are used in order based on what’s in sycl-ls but try the environment variable if that’s not working for you.
Please can you help how can i set Intel(R) Arc(TM) A770M as default that sycl-ls instead of Device Name: Intel(R) Iris(R) Xe Graphics As llama-cpp is detecting Xe Graphics and running inference there instead of Intel Arc GPU ?