prompt:
  template:
    - role: user
      content:
        - type: image_url
          image_url:
            url: "{{image_url}}"
        - type: text
          text: "Please detect and label all objects in the following image and mark their positions."
  template_variables:
    - image_url
  metadata:
    model: https://huggingface.co/collections/OpenGVLab/internvl-20-667d3961ab5eb12c7ed1463e
  client_parameters: {}
  custom_data: {}