prompt-free testing

[yo3nglau]

Hi @ShellRedia ,

As stated in #5 and even early #4, what should be done if one would like to test without any prompts? I tried to add constant points like [-100, -100, 0], which is obviously a negative point prompt, but it did not work (an empty mask was generated). Could you provide guidance on implementing no prompt testing in the paper?

Thanks for your contribution. I look forward to hearing back from you soon.

Best, Yo3ng

[ShellRedia]

That's a good question. From a practical standpoint, it indeed shouldn't be necessary to spread tasks across different weights. I will try to re-fine-tune the weights soon and provide relevant explanations, aiming to simplify usage issues as much as possible 这是一个好问题。从实际用途而言，的确应该是不需要把不同类型的任务散布在不同权重当中，我会在近期尝试重新微调一份权重并给出相关说明，作为示例争取尽可能地简化使用上的问题。

[yo3nglau]

Hi @ShellRedia ,

Thanks for your quick response. Does your answer mean to release a generalist model? It would be great to witness it. Please allow me to state my question again: I want to know how to infer without given prompts generated from GT labels (e.g., for a custom dataset that has no GT labels).

[ShellRedia]

I have added the prediction code, weights, and relevant instructions in the README. I hope that running the code will help you find the answers you're looking for. 我已经增添了预测代码，权重，和README中的相关说明。希望通过运行代码可以找到您要的答案。