Can't convert openLLAMA 7b

[Darin755]

Hello,

I am trying to convert openLLAMA 7b using the convert script. I ran the command in order to convert:

python3 convert.py ../open_llama_7b_v2/ (open_llama_7b_v2 is the name of the directory with openllama)

However, I get the following error:

ValueError: unknown format: ../open_llama_7b_v2/pytorch_model-00001-of-00002.bin

I don't know how to troubleshoot this. Does anyone else have more information?

[Yeetster-cmd]

I don't think convert.py itself works with the new version of llama.cpp, but there are many other convert files, the one I think may work (though I'm not sure) may be convert-llama-hf-to-gguf.py.

[Darin755]

That script just gives me a serialization error.

[KerfuffleV2]

convert.py works with the latest version. I'm pretty sure it is the recommended tool for model conversion (at least LLaMA) models.

OP, "unknown format" sounds like the file is corrupted or something. Double check the validity of the files. If you got it from HuggingFace, HF will show you the SHA256 for each file. On Linux (maybe Mac?) you can run the sha256sum tool. Not sure how to check hashes on Windows.

[bukosabino]

Hi @Darin755 ,

Please, check this comment.

It was the solution for me.