How do i stop the ai from talking to itself??

[Asory2010]

It sometimes just talks to itself for example:

###Human: Hi ###Assistant: Hello, how can i assist you?

(i am runing the latest release with vicuna mode)

[hungerf3]

Take a look at the --reverse-prompt option; it lets you indicate when you want to be able to enter your input.

[andresdelcampo]

Great to know! Is there a way to prevent the reverse prompt ('### Human:") from being printed in the screen?

> Hi there
Hello! How can I assist you today?
### Human:
>

[ungarscool1]

On a personal project, I detect the first # and : or \n. If between these, it contains 'Human', so I stop the execution and find&replace ### Human: to blank.

[Alumniminium]

-r '###' works sufficiently well and catches all the other various things like ### Instruction, ### Next Prompt, ### Question, ... the model occasionally creates.

You can pass multiple -r by the way. I also have a -r "Im sorry," and -r "However," to cut it off from either lecturing me about ethics or overexplaining things.

[IIIIIIIllllllllIIIII]

doesnt work at all on mac - super weird....

./main -m ./models/ggml-vicuna-7b-1.0-uncensored-q4_0.bin -n 256 --repeat_penalty 1.1 --color -i -r '### Human:' -r '##' -r '###' -f prompts/chat-with-vicuna.txt

[ibndias]

This works for me for vicuna prompt trained model

./main -m ../../Downloads/wizardLM-7B.ggml.q4_0.bin -n 2048 --repeat_penalty 1.0 --color -i -r "### Human:" -f prompts/chat-with-vicuna.txt

chat-with-vicuna.txt

[github-actions[bot]]

This issue was closed because it has been inactive for 14 days since being marked as stale.