Find answers from the community

Updated 5 months ago

Setting system prompt with nvidia llama model

At a glance

The community member is asking how to set a system prompt when using the NVIDIA class to create an LLM, specifically for use with the FunctionCallingAgent. The comments suggest that the system prompt can be set when creating the FunctionCallingAgent, with one community member providing an example of setting the system prompt to "Talk like a pirate in every response." However, there is no explicitly marked answer in the comments.

kkere

Hello,
Does anyone know how to set a system prompt when using the NVIDIA class ?

line : llm = NVIDIA(model="meta/llama-3.1-70b-instruct")

I want to use FunctionCallingAgent with the llm above but I couldnt figure how to set a prompt that defines its general purpose.

agent = FunctionCallingAgent.from_tools(
[discovery_tool],
llm=llm,
verbose=True,
)

3 comments

LLogan M

Plain Text

agent = FunctionCallingAgent.from_tools(
    [discovery_tool],
    llm=llm,
    verbose=True, 
    system_prompt="Talk like a pirate in every response."
)

LLogan M

should be that no?

kkere

thanks a lot

Add a reply