TheBloke

/

Llama-2-70B-Chat-GPTQ

like 257

This is an example of in-context learing in other model,**history is two-dimensional list**，each list in history describes the implementing prepared conversations between user and assistant:
    history = [[**User:**："Summarize this Ethereum transaction in one sentence:{......}",
    Assistant："DwarfPool sent 4.79 Ether to 0x555be1...0702b5Ed with a transaction fee of 0.001 Ether at Oct-14-2015 07:22:38 AM +UTC."],]
    prompt="Summarize this Ethereum transaction in one sentence:{......}”
    response, history = model.chat(tokenizer, prompt,**history=history**,max_length=1024)
I feel like it's important for the model that I can fine-tune the output of the model without using lora.Do you know how to insert some  code to achieve this function in llama2，thanks！