Is anyone able to run Hermes with Gemma 4?
Posted by hvs69@reddit | LocalLLaMA | View on Reddit | 7 comments
I am using Gemma31b (ollama). Hermes installs just fine but cannot even do basic tasks like reading my project folder. It goes into some kind of hallucination when I ask it to read my project folder.
Is anyone successful ?
Euphoric_Emotion5397@reddit
ya.. gemma 4 is just a chat bot for me until someone technical can fix it. My LM studio and Ollmaa ... qwen 3.5 works perfectly. But Gemma 4.. wow.. real piece of work.. Everyone talks about how great it is.. I only see a chat bot now unfortunately. There is now way to make it tool calll or do function calling successfully.
hvs69@reddit (OP)
Thanks for confirming
Double_Cause4609@reddit
Have you verified with upstream LlamaCPP (which ollama inherits from) with the --jinja flag, or with vLLM?
If it can't read the folder it might be unable to execute functions. I'm not sure what Ollama does for function calling so I can't help there.
sleepingsysadmin@reddit
Is your ollama fully up to date? Might just need to patch it.
hvs69@reddit (OP)
yeah, everything is up to date.
Ashamed-Honey1202@reddit
Yo lo usé con 26b y pude configurar todo bien, hasta el acceso por Telegram
hvs69@reddit (OP)
good to know