Twitxy Saved

Quentin Lhoest ๐Ÿค—

@lhoestq

Datasets & Compute @huggingface since 2020 | Open Source + HF Hub platform
huggingface.co/lhoestq Joined September 2013
345 Following 5K Followers 2.8K Posts

Latest posts 5

Quentin Lhoest ๐Ÿค— @lhoestq
Loving the newly added decision models support in transformers !

My agent got great accuracy and decisions/sec on simple hardware in a few lines

Locally:
> uv run run_decision_model_imdb.py

On HF with cheap hardware
> hf jobs uv run run_decision_model_imdb.py

๐Ÿ‘๐Ÿ‘@art_zucker
Lysandre Verified account @LysandreJik Nov 10
A pretty important new primitive just landed in ๐Ÿค— Transformers: **decision models**.

I think this could matter much more than adding support for any single model.

Today, different decision models all approach "making a decision" differently:

- @typesafeaiโ€™s Jev, the OG: typed decisions + calibrated probabilities, single forward
- @AutoTrustAIโ€™s JEV builds open System One models around a learned decision block, w/ multimodal.
- @jaredpalmerโ€™s KEV uses a pointer head over the options, sharing the computation of the state across questions.
- @Cloudflareโ€™s CLEF uses a joint schema head that routes evidence between the state, questions and options, scoring the whole schema together.
- @perplexity_aiโ€™s Decider uses a separate 255-way readout on top of Qwen, with non-causal attention in its full-attention layers.
- @liquidaiโ€™s d1 is exploring small, multimodal decision models that can run all the way down to the edge.
- And @XunzhuoLiu's Semantic Router team is pushing the serving side hard with System One + the D3 family: one typed interface for choice, score and yes/no decisions, with models from sub-1B to 27B.

That experimentation is good, and needed! But without a common interface, every training framework, inference engine and evaluation tool has to understand every implementation separately.

So Transformers is introducing a shared decision-model layer:

`DecisionConfig`
`pipeline("decision")`
and, most importantly:

`model.decide(...)`

The idea is **not** to make every decision model work the same internally, but for them to speak the same language to the ecosystem around them.

This means that, without running custom or remote code:
- inference engines can serve them through one interface
- training tools can build against common primitives.
- the Hub can package them consistently.
- evaluations can compare them without model-specific glue everywhere.

This is the part of transformers I'm the most excited about: to grow alongside the ecosystem and the many ideas; and to make all of them more usable by everyone.

github.com/huggingface/trโ€ฆ
View quoted post on X
2 replies 1 reposts 11 likes 682 views View on X