Llama 3 chat template not working?
#1
by
Jdods84
- opened
When I use the Llama 3 chat template for my prompt, Lumimaid seems to repeat responses and loses its mind. When I remove the formatting and use my Alpaca format, it RPs and works fine. I am using KoboldCPP to run this LLM locally. Would it be possible to get some help from an expert with formatting my prompt correctly?
It work for me on kobold, I use the one I put on the model card (Llama-3-Instruct)
It also work unquantized, the prompt format is on the tokenizer_config.json
Can you check that and be sure you use the right one? Updated all your tools ?