about evaluate/perplexity

I've found that the perplexity example in the evaluation function on huggingface is always in the Runtime error state[https://huggingface.co/spaces/evaluate-metric/perplexity](url). When will it be fixed?

BTW,After referring to some tutorials, I successfully evaluated the model perplexity of GPT2, but when testing the local Llama-7B model, I encountered a lot of errors. Do you have any kind friends who can help me explain how to import models and datasets from local for testing, thank you very much!