search

ShishirPatil / gorilla

Gorilla: Training and Evaluating LLMs for Function Calls (Tool Calls)
https://gorilla.cs.berkeley.edu/
Apache License 2.0
11.48k stars 999 forks source link

Contributing APIs #3

Open romilbhardwaj opened 1 year ago

romilbhardwaj commented 1 year ago

I'd like to contribute the SkyPilot API. What's the best way to add it to Gorilla?

ShishirPatil commented 1 year ago

Hi @romilbhardwaj Thank you for your interest! Yes, we would love to have SkyPilot integrated into Gorilla! In #9 we detail how to submit APIs. The walk through in the data/README.md should be helpful. In case you run into any issues, or have any suggestions - let us know!

Excited to welcome SkyPilot API 🚀

gottlike commented 1 year ago

@ShishirPatil Is it possible to train Gorilla myself to use non-public internal APIs? How would I best go about doing that?

ShishirPatil commented 1 year ago

Hey @gottlike @CrustyCode and @AldiRvn. Yes, indeed. Step 1 would be to create data in the following (format)[https://github.com/ShishirPatil/gorilla/tree/main/data#option-1-api-json-preferred]. Once you have this, you would then train Gorilla using the classical perplexity loss for the zero shot setting. In case you want the model to be robust to frequent changes to API, then you would also throw in a retriever and adopt Retriever Aware Training (RAT). Hope this helps, happy to answer any questions!

joemccann commented 1 year ago

@ShishirPatil from the demo video on YouTube you show an example of calling AWS services like s3. Where in the repo is AWS? I only see models?

ShishirPatil commented 1 year ago

Thanks for your interest @joemccann ! Yes, for the latest batch of models which includes the AWS, GCP, Azure, K8S, etc, we are in the process of releasing the data. You can use our models today, though. In case you haven't come across, the CLI tool is at https://github.com/gorilla-llm/gorilla-cli