NOW FIXED: Usual bug. Failed to parse Jinja template.
Failed to parse Jinja template: Parser Error: Expected closing statement token. OpenSquareBracket !== CloseStatement.
Failed to parse Jinja template: Parser Error: Expected closing statement token. OpenSquareBracket !== CloseStatement.
EDIT: ITS FIXED! PLEASE REDOWNLOAD
Hi there seems to be a chat template issue in llama.cpp. We are going to reupload all of them!!!
Failed to parse Jinja template: Parser Error: Expected closing statement token. OpenSquareBracket !== CloseStatement.
Hi there seems to be a chat template issue in llama.cpp. We are going to reupload all of them!!!
RIP 🦥😭
I posted a fix here for LM Studio in the mean time. It's just from LM Studio's own release, their Jinja template seemed to be fine:
https://huggingface.co/unsloth/Qwen3-8B-128K-GGUF/discussions/1
There are actually 2 different template bugs, one that's LM Studio specific, the other is llama.cpp
@wapxmas
@MrDevolver
@phazei
@Cyanpink
Apologies guys! It's now FIXED!!!
Works on any single inference engine. Lmstudio llamacpp Ollama etc
Haven't tried none of the 2 but I would like to ask if this now fixed bug could eventually affect also the 128k version that I was going to download, as I'm on a capped mobile plan, tx