huggingface / huggingface/smolagents
Does smolagents offer structured outputs (constrained decoding not implicit)
- Dominant language
- Python
- Stars
- 29.3k
- Forks
- 3k
- Avg merge
- 17m
- Merged PRs (30d)
- 2
Description
@aymeric-roucher
I've been wondering what solutions HuggingFace offers with regards to structured outputs (SO) and have come short (the only thing I have found is dottxt/outlines, which is totally ok if HF intends to depend on them for SO).
This is particularly relevant for smolagents where SO compound in value.
I know this is not a very smolagents specific issue but nonetheless I think its highly consequential for smolagents.
Do you know what HF strategy towards structured outputs is? How are you guys thinking about it?
Contributor guide
Research direction
The issue names no files, tests, or entry points. First clarify the intended Hugging Face strategy for structured outputs and whether smolagents should implement or depend on a solution; done criteria are not defined in the current discussion.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ai
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100