This is very high level as much more scaffolding needs to be implemented but thing of this as an "epic" from scrum methodology.
We would like to call a fine-tuned GPT3.5 model. The model should be fine-tuned on at least the Scenic codebase and documentation. Maybe even the CARLA documentation as well but CARLA is TBD. We can open a new issue for that.
https://platform.openai.com/docs/guides/fine-tuning
- Fine-tuning on the scenic codebase:
- Function calling: https://platform.openai.com/docs/guides/gpt/function-calling
This is very high level as much more scaffolding needs to be implemented but thing of this as an "epic" from scrum methodology.
We would like to call a fine-tuned GPT3.5 model. The model should be fine-tuned on at least the Scenic codebase and documentation. Maybe even the CARLA documentation as well but CARLA is TBD. We can open a new issue for that.
https://platform.openai.com/docs/guides/fine-tuning