Instructions to use Sambenja1/glirel-ghost-b-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- GLiNER
How to use Sambenja1/glirel-ghost-b-v1 with GLiNER:
from gliner import GLiNER model = GLiNER.from_pretrained("Sambenja1/glirel-ghost-b-v1") - Notebooks
- Google Colab
- Kaggle
GLiREL โ Ghost B v1 (Polymath)
Fine-tuned GLiREL relation-extraction model for Polymath's local "Ghost B" ingestion lane. Predicts the 30-predicate Ghost B schema over GLiNER-extracted entities, fully on-device (no cloud).
- Base encoder: microsoft/deberta-v3-large
- Predicates: 30 (see
labels.json) - Best-F1 threshold: 0.40
- Pairs with: GLiNER
urchade/gliner_medium-v2.1(entity pass)
Use in Polymath
python scripts/bootstrap_models.py --gliner torch --glirel-custom
This downloads the checkpoint to models/glirel_ghost_b_v1/best/, where the
extraction sidecar loads it by default. Or point GLIREL_CKPT_DIR at this repo
id directly.
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support
Model tree for Sambenja1/glirel-ghost-b-v1
Base model
microsoft/deberta-v3-large