Expand description
An inference engine for GLiClass models.
These language models are efficient for zero-shot topic classification or derivatives like sentiment analysis. They can also be used for efficient re-ranking.
GLiClass stands for “Generalist and Lightweight Model for Sequence Classification”, after an original work from Knowledgator, which was itself inspired by GLiNER.
Modules§
- input
- Input steps of the pipeline
- output
- Output steps of the pipeline
- params
- GLiClass Parameters
- pipeline
- Orp pipeline for GLiClass classification
- tokenizer
- Wrapper around HuggingFace tokenizers
- util
Structs§
- GLiClass
- Convenience front-end for easy use with default runtime parameters (CPU).
For more advanced use, see examples and the
orpcrate.