Since installing the GPU version of llama-cpp-python can be challenging, I have uploaded the JoyCaption-beta-one-hf-llava-GGUF model to Ollama. You can now use these models by installing Ollama and the ComfyUI-Ollama.
I have separated the prompt template used by the ComfyUI-joycaption-beta-one-GGUF node into this dedicated node.