Produced Output Stability
#13
by
luissimoes
- opened
Hi,
I have been trying the model, and I wanted to clarify one thing...
Is there a way to make outputs more predictable? I am making same prompt and obtaining different answers every time.
Is there a way to keep the answers stable and consistently producing same expected output?
Thanks
Which software are you using?
In llama.cpp main or server, you need to set --seed to a number. You should get a reproducable deterministic result for each seed.
I am using LangChain for now.
Is the seed something that is configured on the config section?