Load this checkpoint as a Transformers schema-extraction model.
#7
by MattThomas-fastino - opened
Point this checkpoint at the native Transformers schema-extraction model.
model_type is gliner2 and architectures is Gliner2ForSchemaExtraction, so pipeline("schema-extraction", model=...) can select the class. The encoder config is inlined from encoder_config/config.json. That directory, the weight file, and the existing gliner2 fields (model_name, span_head or boundary_head) stay, so AutoExtractor.from_pretrained still loads this repo.
The model card front matter now says library_name: transformers and pipeline_tag: schema-extraction. The card body is unchanged.
urchade changed pull request status to merged