bartowski

/

THUDM_GLM-4-32B-0414-GGUF

like 92

And there's another fix for GLM4 chat template, it seems just a linefeed https://github.com/ggml-org/llama.cpp/releases/tag/b5250

Here are the relevant links:
https://huggingface.co/THUDM/GLM-4-32B-0414/commit/8dea913c41b543527c2860fb6f3c15c202f24aeb
https://github.com/ggml-org/llama.cpp/pull/13238

this new one i think is only on the inference side, right?

Versus the first that i regenerated for?

Yes. Inference side. But the modified jinja template is embedded in the gguf and used in llama.cpp.

ohhhhh i didn't notice that part

but that would have been picked up when i remade these a couple days ago, so no worries there :)