Add cpuflow-v8-discrete model weights and model card

initial commit