license: apache-2.0 datasets:


Overview

2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference of digitalpipelines/llama2_7b_chat_uncensored

Prompt style

The model was trained with the following prompt style:

### HUMAN:
Hello

### RESPONSE:
Hi, how are you?

### HUMAN:
I'm fine.

### RESPONSE:
How can I help you?
...