unsloth

/

DeepSeek-R1-0528-Qwen3-8B-GGUF

like 313

Follow

Unsloth AI 9.35k

Where can I find chagelog ?

I cannot reliably answer your questions guys as I'm just a user, but:

• I don't think there are changelogs available, but I read in another Unsloth GGUF model thread (sorry can't find it back) that they plan to regularly update their imatrix calibration datasets and requantize the models, so it might be the reason here as the weights changed.
• From my few very basic local tests (eg. "Please give me 10 sentences ending by apple as very last word.") before/after this update, I seem to get better results with the new weights, at least on Q6_K_XL.

So if you can afford, I think you should update ;)

Actually wait in an hour or so - I'm adding trying to add native tool calling support!

Ok done! Tool calling support works now!

So why is the template not supported in lm studio or llama.cpp?

@urtuuuu What command are you running? Are you adding --jinja as suggested?

@urtuuuu llama.cpp works - lm studio I just fixed 1 hour ago!
As @owao mentioned for llama.cpp, add --jinja. lmstudio should be automatic now

@urtuuuu What command are you running? Are you adding --jinja as suggested?

In LM Studio latest version under Promt>Template(Jinja) there is an error message: Failed to parse Jinja template: Parser Error: Expected closing statement token. OpenSquareBracket !== CloseStatement.
Or wait... Was the gguf updated again? I thought i downloaded already undated version yesterday... cmon