[Bug]远程计算机拒绝网络连接

[wenluo-1128]

发生了什么

在配置[的消息配置器中选择的是个人微信]，然后出现了“远程计算机拒绝网络连接“

如何复现？

打开配置的消息配置器中选择个人微信，然后重启astrbot，就出现了这个问题，其中

AstrBot 版本、部署方式（如 Windows Docker Desktop 部署）、使用的提供商、使用的消息平台适配器

版本是v3.4.35，部署方式是windows本地部署，不是用docker部署

操作系统

Windows

报错日志

以下是报错日志：

你愿意提交 PR 吗？

• [x] 是的，我愿意提交 PR!

Code of Conduct

• [x] 我已阅读并同意遵守该项目的 行为准则。

[LHamnett]

@ztfmars , I was also experiencing this issue, and I think I managed to resolve it or at least find a work around. In the config file, the default value for cfg.model.pretrained is likely set to 'pretrain/vit_large_p16.pth' . For me I set the variable cfg.load_from to the correct location of the .pth file for my model, and I set cfg.model.pretrained to None. And this allowed the model to get further in the training process.

Possibly there needs to be some logic to check if both cfg.model.pretrained and cfg.load_from have been set, or possibly it is currently set that cfg.model.pretrained takes precedence over cfg.load_from.

Hope this is useful information.

[wyk1998]

I'm having the same problem, have you solved the problem now, any solutions?

[mbhurtel]

If you see the base config in base/models/setr_naive, then you find pretrained='pretrain/jx_vit_large_p16_384-b3be5167.pth' on line 14.

This exact model can be downloaded from https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p16_384-b3be5167.pth

I used the pretrained weights from the above model, and it works perfectly!