#Qwen tool calling
1 messages ยท Page 1 of 1 (latest)
I've used qwen a bunch on ollama so I can help if needed
My guess is that, even though the model is compliant with the OPENAI library, it seems that it's not (for the 2.5-coder,14B)
Did you manage to get it to do some tool calling ?
yep! one sec
so one thing about all these different models is they all respond differently to prompts. So a good system prompt for gpt-4o might not be great for qwen.
Anyway, for qwen I always put - Always write changes to files to the original files in the workspace in the system prompt so that when it comes up with new files or file changes it calls the write function after
(based on the image) Completely off, based on current llm branch (dev version)
Like it currently just doesn't see my local (ollama qwen) tools
It's working fine for me on 17-llm.1 but let me pull llm.2
Is it with qwen-coder-2.5 ?
qwen2.5-coder:14b
Yeah so exactly the same ๐
Yep definitely working on my end, screenshot in a sec
one sec i'll run another prompt and screenshot the tool calling
To fix I'd try more prompting and maybe if you can share what NodeWorkspace looks like in case there's a bug with it
cc @fervent lagoon ๐
Re-compiling the engine -- interested to sync 5 min when finished recompiling, just to double chefck ๐
(it's gonna take 10min, my computer is heating ahha)
i'm on dev-audio ๐ Feel free to hop in anytime