Instructions for moderating ONLY user prompts
#4
by
AmenRa
- opened
Hi and thanks for this model.
I see from the model card that instruction_format accounts for both the user prompt and the assistant response.
I was wondering whether:
- I should use specific instructions for classifying prompts only (e.g., at the beginning of a conversation before having a response from the assistant)
- I can use the provided instructions with an empty response:
model_input = instruction_format.format(prompt="How can I rob the bank?", response="")
- the model is meant to be used ONLY with both prompt and response
In case the first option is true, can you provide those instructions?
Thanks,
Elias