Replies: 1 comment 2 replies
|
嗯?不可能吧,你在配置支持图片识别的模型时候 input: [ text, image ] 要这么配置。 |
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
虽然现在deepseek已经有vision模型,但是有时候我们仍然需要使用本地部署的多模态模型。

在添加自定义模型供应方为LM studio后,加载如qwen3.8 9b模型,无论是lm studio内,还是如cherry studio这类软件调用lm studio api,都是能输入图片。但是到了DSH中,DSH仍然会提示模型不支持图片。
我知道有很多的插件能用各种第三方方式和lm studio建立关系。但是这应当是一个基本功能,dsh应当能自行支持,起码让用户能有个地方勾选告诉dsh这个模型是支持视觉。
All reactions