Jinze Bai

youtube face是对齐好的数据。如果是用自己的数据集训练的话，需要通过人脸校准算法，如SDM算法寻找landmark来把所有的人脸对齐～然后剪裁～ @Alexyuanmu

Thanks for your attention. Sure we hope to make it available to more users. We will upload the dependent files to s3 or HF hub asap. We would appreciate it...

Could you please try to replace the OSS link with an HTTP link? To see if this is helpful in alleviating the timeout problem? for example, replace https://github.com/OFA-Sys/OFASys/blob/f09a3f55292487084021947bc10c9344e1f3f06e/ofasys/adaptor/image_vqgan.py#L50 to ```...

First, update the ofasys package to the latest version: ``` pip install -U http://ofasys.oss-cn-zhangjiakou.aliyuncs.com/pkg/ofasys-0.1.0-py3-none-any.whl ``` Then, try to use a remote link to initialize the model, as in the [documents](https://ofasys-doc.readthedocs.io/en/latest/start/quickstart.html#inference-with-all-kinds-of-tasks-with-one-checkpoint):...

可以参考文档：https://help.aliyun.com/zh/dashscope/developer-reference/vl-plus-quick-start 示例： ``` from http import HTTPStatus import dashscope def simple_multimodal_conversation_call(): """Simple single round multimodal conversation call. """ messages = [ { "role": "user", "content": [ {"image": "https://dashscope.oss-cn-beijing.aliyuncs.com/images/dog_and_girl.jpeg"}, {"text": "这是什么?"}...

主要是因为这个脚本可以同时评测Qwen-VL和Qwen-VL-Chat模型哈，因为Qwen-VL的预训练模型没有ChatML格式。