setting use_gpu for knn embeddings

On a Mac M1 (with Metal 4 GPU) I am trying to set autoembeddings with use_gpu=true (Manticore V25 latest)

On starting the server I get the following error:

WARNING: table ‘rt_products’: prealloc: Failed to initialize CUDA device - NOT SERVING

There is not much information on this flag so difficult to understand what are the gpu requirements (NVDIA GPUs only probably as CUDA is their layer… ?)

Any information available on this ?

Thanks

CUDA is proprietary technology that exclusively runs on NVIDIA hardware; that’s why it may not be working with the Metal 4 GPU. In theory, it might work with Metal too, but we haven’t tested that yet.