implement better rate-limiting and sorted imports.

deep-diver / llamaduo

This project showcases an LLMOps pipeline that fine-tunes a small-size LLM model to prepare for the outage of the service LLM.

https://huggingface.co/papers/2408.13467

Apache License 2.0

286 stars 29 forks source link

Closed sayakpaul closed 6 months ago

sayakpaul commented 6 months ago

This PR implements:

Better rate-limiting.
An import utility module to prevent the installation of modules we don't need for a certain step (`alignment isn't needed for synthetic data generation).

Couldn't test it because of error code 429. Would appreciate some testing help.

deep-diver commented 6 months ago

FYI: The rate limit of Gemini 1.0 Pro has been decreased down to 15 RPM (ref)

sayakpaul commented 6 months ago

@deep-diver pushed some updates. Feel free to test and let me know.

deep-diver commented 6 months ago

@sayakpaul

pushed updates worked with some minor errors. I am sorry I should have asked you first, but I made some changes to make it work properly.

sayakpaul commented 6 months ago

No worries. Appreciate the help.