Text Generation
Transformers
PyTorch
English
llama
text-generation-inference
Inference Endpoints

Change prompt format to Vicuna v1.1

#7
by Kenshiro-28 - opened

Hi! This model is very promising, I would like to suggest changing the prompt format to Vicuna v1.1:

USER: Hi
ASSISTANT:

This uses less tokens than Alpaca format and is the one used in popular models like Vicuna, WizardLM, Wizard-Vicuna or Airoboros.

What is the current format ?

According to https://huggingface.co/TheBloke/OpenOrca-Preview1-13B-GGML it uses the Alpaca format:

### Instruction: {prompt}

### Response:
Kenshiro-28 changed discussion status to closed

Sign up or log in to comment