fulstock/gliner-nerel-finetuned

GLiNER NEREL Finetuned

GLiNER model finetuned on NEREL dataset (Russian NER, 29 entity types).

• Base model: urchade/gliner_multi-v2.1 (mdeberta-v3-base)
• Training: 100k steps, batch 16, focal_loss_alpha=0.75, lr=1e-5
• Data: NEREL (746 train / 94 val examples)

Usage

from gliner import GLiNER

model = GLiNER.from_pretrained("fulstock/gliner-nerel-finetuned")
entities = model.predict_entities(
    "Иван Иванов посетил Москву 5 января 2024 года.",
    ["PERSON", "CITY", "DATE"],
    threshold=0.5,
)
for e in entities:
    print(e["text"], "=>", e["label"])