Qwen

/

Qwen2-72B-Instruct-GGUF

like 30

Follow

Qwen 2,088

can you solve this problem ?

I downloaded this file TWICE to verify that the problem is not on my side:

qwen2-72b-instruct-q5_k_m-00002-of-00002.gguf

Sadly, when i inspect it, it shows as 0B in properties...
When trying to merge i get

./llama-gguf-split --merge qwen2-72b-instruct-q5_k_m-00001-of-00002.gguf qwen2-72b-instruct-q5_k_m-00002-of-00002.gguf
gguf_merge: qwen2-72b-instruct-q5_k_m-00001-of-00002.gguf -> qwen2-72b-instruct-q5_k_m-00002-of-00002.gguf
gguf_merge: reading metadata qwen2-72b-instruct-q5_k_m-00001-of-00002.gguf done
gguf_merge: reading metadata qwen2-72b-instruct-q5_k_m-00002-of-00002.gguf ...gguf_init_from_file: invalid magic characters '�'

gguf_merge: failed to load input GGUF from qwen2-72b-instruct-q5_k_m-00001-of-00002.gguf

merging both files with CAT does not work properly aswell.

can you solve this problem ?

You don't generally need to join the files at all. Just load the 00001 as if it was the whole model, and llama.cpp will find and load the 00002 for you.

You don't generally need to join the files at all. Just load the 00001 as if it was the whole model, and llama.cpp will find and load the 00002 for you.

This is not true for me as i have tried it.

I think there is something bad in the file itself too.
The problem is NOT combining GGUF files in general! The problem is invalid magic characters '�'

Nevermind, do what you want with this comment, i just chose a different quantation