ggerganov/llama.cpp

convert-hf-to-gguf.py fails PR #7234

LostRuins opened this issue · 0 comments

Hi, after merging #7234 , model conversion with convert-hf-to-gguf.py is broken.

  File "E:\LLaMA\llamacpp\gguf-py\gguf\lazy.py", line 9, in <module>
    from numpy._typing import _Shape
ModuleNotFoundError: No module named 'numpy._typing'

I was using numpy-1.22.3 After force upgrading my environment to the latest numpy-1.26.4, the script finally works.

However, I am hoping that it is possible to allow it to work with numpy 1.22 as it did before this commit as a fallback? A lot of toolchains might still be on slightly older versions of numpy, forcing the use of the latest newest version may not be ideal.