cosyvoice3-0.5B loading error

[ericosmic]

Describe the bug
when I loading Fun-CosyVoice3-0.5B, it throw out a error:
RuntimeError: Calculated padded input size per channel: (3). Kernel size: (4). Kernel size can't be greater than actual input size

the entirely error info:

To Reproduce
Steps to reproduce the behavior:

1. enter a docker container that based on bobui/cosyvoice image
2. running webui.py from FunAudioLLM/CosyVoice
3. choose cross_lingual inference in gradio

Desktop (please complete the following information):

• OS: linux
• Browser chrome
• Version [e.g. 22]

[hpfs0]

同样的问题，求大佬解决