The image size parameter in the cogvlm tool does not match the parameter provided on the official Zhipu website because the CogView3Tool class in the cogview3.py file uses a predefined size_mapping dictionary to map size options to specific dimensions:

size_mapping = {
    "square": "1024x1024",
    "vertical": "1024x1792",
    "horizontal": "1792x1024",
}

This mapping may differ from the size parameters specified on the official Zhipu website, leading to interface errors. Additionally, the stepfun tool, as described in the image.yaml file, offers a different set of predefined image sizes:

options:
  - value: 256x256
    label:
      en_US: 256x256
  - value: 512x512
    label:
      en_US: 512x512
  - value: 768x768
    label:
      en_US: 768x768
  - value: 1024x1024
    label:
      en_US: 1024x1024
  - value: 1280x800
    label:
      en_US: 1280x800
  - value: 800x1280
    label:
      en_US: 800x1280

The discrepancy between these predefined sizes and the ones used in the cogvlm tool can cause interface errors when the expected parameters do not align ^[1][2].

To continue talking to Dosu, mention @dosu.

@ox01024 Hello can you update this tool?

Leave it to me. I'll take care of him as soon as possible. Don't worry.

@ox01024 Hello can you update this tool?

Leave it to me. I'll take care of him as soon as possible. Don't worry.