would love to be able to use the ollama instance running on machine with dedicated GPU via ollama API