unsloth

/

gpt-oss-120b-GGUF

like 99

Follow

Unsloth AI 8.25k

Using the command: llama-gguf-split --merge gpt-oss-120b-UD-Q8_K_XL-00001-of-00002.gguf gpt-oss-120b-UD-Q8_K_XL.gguf

gguf_merge: reading metadata gpt-oss-120b-UD-Q8_K_XL-00001-of-00002.gguf ...gguf_init_from_file_impl: tensor 'blk.0.ffn_down_exps.weight' has invalid ggml type 39 (NONE

version:

llama-gguf-split --version 
version: 5992 (ce111d39)
built with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu

Does the 20B model work? Did you recompile llama.cpp and update it?

I didn't recompile llama.cpp, that seemed like a lot of work. I used the model in the ollama repository instead.

ollama probably doesnt include the related PRs yet, your only option may be to compile llama.cpp with latest commits yourself