Is there a way to update GGUF version to match the llama.cpp version?

[Disura-Randunu]

Hi all

I'm using LM Studio and downloaded some models which uses the older llma.cpp versions. because of this, the following error gets thrown when loading the model:

llama.cpp error: 'missing tensor 'blk.0.ffn_down_exps.weight''

From what I was able to search, it seems to be beacuse of a version mismatch of llama.cpp which was used to make the current GGUF of the model.

Is there a way to upgrade the GGUF file ?

Thanks

[slaren]

There is no way to update these GGUF files, you would need to find the original HF model and convert it using convert_hf_to_gguf.py. However, keep in mind that this model is likely to be very old and outdated by current standards, you may want to look for a newer model instead.

[Disura-Randunu]

I found this in reddit, but I think it's not possible within the current version:

update the gguf filetype to current if older version is unsupported by another application
./quantize ./models/7B/ggml-model-q4_0.gguf ./models/7B/ggml-model-q4_0-v2.gguf COPY

https://www.reddit.com/r/LocalLLaMA/comments/16px2od/converting_outdated_gguf_to_modern_one/