Lucien

https://github.com/huggingface/huggingface-vscode#configuring Sure, you can deploy it locally then using endpoint to specific your own service.

Dose the multi-task fine-tuned model's tokenizer & model structure as same as [damo/ofa_ocr-recognition_general_base_zh](https://www.modelscope.cn/models/damo/ofa_ocr-recognition_general_base_zh)? There's only one `.pt` file in downloaded file.

是使用的 `docker-compose` 部署的吗？改一下 16 行 `80:8080` 就好了。

你好，admin 的地址需要自行反向代理 `/api/` 路径，这个当时没有留下文档，导致了现在的问题，后面有时间我再做补充。 举个例子，比如当 admin 的 `docker-compose.yml` 文件如下所示时，需要将 `localhost:18001/api/` 反向代理至 `http://pasteme.cn/backend/api/` 后，在 `api.backend` 中填入 `/backend/api/`。 ```yaml version: "3" services: pasteme-admin: image: registry.cn-hangzhou.aliyuncs.com/pasteus/pasteme-admin:0.0.2 container_name: pasteme-admin volumes: - ./log:/app/log - ./config:/app/config...

same issue

Sorry for taking so long to reply to you. In fact, I encourage everyone to submit pull requests directly, and I will carefully review each one.

Have you checked your model's hash? It looks like something goes wrong when loading config or weights. Perhaps problem from system maybe.

What does your `git status` command display? If there is any diff, that means there is something wrong maybe.

What's the reason for removing the package's version? If you want to upgrade some packages, you should specific another version instead of removing it.

Thank you for your concern. Currently, this project has not been tested for compatibility with StarCoderPlus, but it may be possible and I will test later on.