Inconsistent sizes?
#1
by micsthepick - opened
Qwen2.5-Math-72B-Instruct.Q4_0.gguf Q4_0 38.4GB
Qwen2.5-Math-72B-Instruct.IQ4_NL.gguf IQ4_NL 19.8GB
Unless mistaken, according to https://github.com/ggml-org/llama.cpp/pull/5590 - the above two rows should have approximately the same size in GB? Please advise.
additionally, at least with the HF Hub, the metadata fails to load: "Error: not a valid gguf file: not starting with GGUF magic number"
Yeah, seems like just a broken file
I need to go through every file in every repo. Occasionally the files are just broken or server crashed and skipped repo as a whole
Eventually I will do it, but not now. I dont have time to code it