Using official gguf package for constants

[ExtReMLapin]

No description provided.

[ExtReMLapin]

Work isn't finish, working on it at the office, and right now it's not the priority anymore.

[99991]

I'd prefer if this package did not require any more dependencies. But renaming gguf.py might be a good idea to avoid collision with the gguf package. Perhaps pygguf would be a better name.

[ExtReMLapin]

Using the official repository would make it automatically support future quantization types

[99991]

Yes, but why not use the official repository directly then? The goal of this library is to dequantize in Python (or NumPy, technically).

[99991]

I only looked at the gguf PyPi package, but apparently it is severely out of date. I just noticed that the official repository implements dequantization in NumPy now.

ggml-org/llama.cpp@4134999

That's great to hear! Then I do not need to maintain this repository anymore and can just redirect everyone to https://github.com/ggerganov/llama.cpp/tree/master/gguf-py