File size: 278 Bytes
f84647c
33d9e9a
f84647c
264f9aa
33d9e9a
 
1
2
3
4
5
6
7
This fixes a template bug in llama.cpp and ollama that is causing a performance degradation in all GLM models.

The problem is caused by a wrong template detection in both projects.

https://github.com/ollama/ollama/issues/10408
https://github.com/ggml-org/llama.cpp/pull/13099