gemma.cpp
gemma.cpp copied to clipboard
GPU Support
Is there a way to use GPU or is this on the roadmap.
GPUs are not supported yet, though we have done a bit of prototyping. Speaking personally, I believe GPUs are well-supported by other inference frameworks and less suitable here given our focus on "experimentation and research".
I'm looking into options with metal shaders/vulkan/webgpu, if anyone wants to collaborate please reach out.
Closing for now but will update here when there's more info on this.