neulab / prompt2model

prompt2model - Generate Deployable Models from Natural Language Instructions
Apache License 2.0
1.96k stars 177 forks source link

Make API-based model configurable #344

Closed neubig closed 1 year ago

neubig commented 1 year ago

Description

This PR makes it possible to choose different API-based models. There are several things involved:

  1. Creating a global API agent
  2. Using the global API agent in various parts of the code
  3. Updating the notebook-based demo

There is also one somewhat unrelated change (apologies for conflating this) in making max_tokens configurable for the API-based agents. This is because I tried with my Cohere account, and was having trouble due to the default max_token setting for Cohere (20 tokens).

I still haven't been able to do a run-through of the entire pipeline with an API other than OpenAI, I'll try to do that next.

References

Blocked by