BIG UPDATE?
I tested Gemma 2 27b when it came out, and it was TRASH.
But now, I tested it another time and all changed!
Now it acceptes a system prompt and is much better...
Does it really accept a system prompt? how to specify it?
How did you use the system prompt? Can I just modify the chat template?
Idk the space is not by me.
But the system prompt is working on llama.cpp:
I think he didn't finetune the GGUF version to have a system prompt, so the transformers version could support it too?
I can also try to load the model on llama.cpp and see the chat template for the system prompt.
If you're looking for an easy way to access this model via API, you can use Crazyrouter — it provides an OpenAI-compatible endpoint for 600+ models including this one. Just pip install openai and change the base URL.