This is
openbmb/UltraLM-13b recovered with
huggyllama/llama-13b and quantized to 4bit GPTQ with the following config:
1quantize_config = BaseQuantizeConfig(
2 bits=4,
3 group_size=32,
4 desc_act=True,
5)
This is UltraLM-13b delta weights, a chat language model trained upon
UltraChatThe model is fine-tuned based on LLaMA-13b with a multi-turn chat-format template as below
User: instruction 1<eos_token>
Assistant: response 1<eos_token>
User: instruction 2<eos_token>
Assistant: response 2<eos_token>
...
To use this model, you need to
recover the full model from the delta weights and perform inference following the template below:
[Optional]User: system prompt<eos_token>
User: user input<eos_token>
Assistant:
