3 ms·
Where can I see an example of an actual prompt that is produced?
by Kiro 3y ago
Where can I see an example of an actual prompt that is produced?
- lbeurerkellner 3y agoFor a showcase of different LMQL queries and resulting model output, have a look at https://lmql.ai https://lmql.ai. Per query, the LMQL runtime calls the underlying LM several times, to execute the complete specified (multi-part) query program. It does not only translate to only one LM prompt, but rather a sequence of prompts, where during generation additional constraining of the LM is applied, to ensure the LM behaves according to the provided template. That's how it support control-flow and external function calls during generation, it actually executes the queries with a proper runtime and only uses LLMs on the backend.