lm-eval evaluates times incorrectly: When running test_see_lm using the environment given by requirements, the error shown above is not clear

[gxy0727]

[rickyang1114]

Hello, may I know if you set HF_ENDPOINT=https://hf-mirror.com? Doing so will result in error during the evaluation of lm-eval.

[gxy0727]

I've already set up the mirror
export HF_ENDPOINT=https://hf-mirror.com
And login to huggingface was set when evaluating the model
huggingface-cli login
But an error is still reported

[rickyang1114]

The root cause of the error is indeed setting export HF_ENDPOINT=https://hf-mirror.com. You may try running unset HF_ENDPOINT and using a VPN, which should resolve the issue.

This error occurs because https://hf-mirror.com restricts access to the Spaces section of HuggingFace to manage traffic. However, lm-eval requires access to this section to download evaluation scripts.

[rickyang1114]