search

FudanDISC / ReForm-Eval

An benchmark for evaluating the capabilities of large vision-language models (LVLMs)
Apache License 2.0
32 stars 4 forks source link

[Usage] This image does not exist, please check #2

Closed yongxinwang-ai closed 8 months ago

yongxinwang-ai commented 10 months ago

Thanks for your good work! I believe the work will make a great contribution to the community.

And my question is, the images are in the format of base64, why the images don't exist in designated path when I use hf?

Should I load the original images and modify the configs?

Aweminus commented 10 months ago

You can try to download our Reform-Eval Data to the root directory through the provided link, and then use --offline_hf, please refer to the Load Data section in Readme.md