GLiNER is a Named Entity Recognition (NER) model capable of identifying any entity type using a bidirectional transformer encoders (BERT-like). It provides a practical alternative to traditional NER models, which are limited to predefined entities, and Large Language Models (LLMs) that, despite their flexibility, are costly and large for resource-constrained scenarios.

This particular version utilize bi-encoder architecture, where textual encoder is Sheared-Llama-1.3B and entity label encoder is sentence transformer - BGE-base-en.

This model leverages the LLM2Vec approach, transforming the initial decoder model into a bi-directional encoder. We further enhanced the model by pre-training it on the masked token prediction task using the Wikipedia corpus. This approach unlocks new capabilities for GLiNER, such as supporting flash attention, enabling a longer context window, and achieving faster inference times. Moreover, by utilizing modern decoders trained on extensive and up-to-date datasets, the model benefits from improved generalization and performance.

This version highlights the key improvements and contextual benefits more clearly.Such architecture brings several advantages over uni-encoder GLiNER:

An unlimited amount of entities can be recognized at a single time;
Faster inference if entity embeddings are preprocessed;
Better generalization to unseen entities;

However, it has some drawbacks such as a lack of inter-label interactions that make it hard for the model to disambiguate semantically similar but contextually different entities.

Installation & Usage

Install or update the gliner package:

pip install gliner -U

Once you've downloaded the GLiNER library, you can import the GLiNER class. You can then load this model using GLiNER.from_pretrained and predict entities with predict_entities.

python

from gliner import GLiNER

model = GLiNER.from_pretrained("knowledgator/gliner-bi-llama-v1.0")

text = """
Cristiano Ronaldo dos Santos Aveiro (Portuguese pronunciation: [kɾiʃˈtjɐnu ʁɔˈnaldu]; born 5 February 1985) is a Portuguese professional footballer who plays as a forward for and captains both Saudi Pro League club Al Nassr and the Portugal national team. Widely regarded as one of the greatest players of all time, Ronaldo has won five Ballon d'Or awards,[note 3] a record three UEFA Men's Player of the Year Awards, and four European Golden Shoes, the most by a European player. He has won 33 trophies in his career, including seven league titles, five UEFA Champions Leagues, the UEFA European Championship and the UEFA Nations League. Ronaldo holds the records for most appearances (183), goals (140) and assists (42) in the Champions League, goals in the European Championship (14), international goals (128) and international appearances (205). He is one of the few players to have made over 1,200 professional career appearances, the most by an outfield player, and has scored over 850 official senior career goals for club and country, making him the top goalscorer of all time.
"""

labels = ["person", "award", "date", "competitions", "teams"]

entities = model.predict_entities(text, labels, threshold=0.3)

for entity in entities:
    print(entity["text"], "=>", entity["label"])

Cristiano Ronaldo dos Santos Aveiro => person
5 February 1985 => date
Al Nassr => teams
Portugal national team => teams
Ballon d'Or => award
UEFA Men's Player of the Year Awards => award
European Golden Shoes => award
UEFA Champions Leagues => competitions
UEFA European Championship => competitions
UEFA Nations League => competitions
Champions League => competitions
European Championship => competitions

If you want to use flash attention or increase sequence length, please, check the following code:

python

from gliner import GLiNER
import torch

model = GLiNER.from_pretrained("knowledgator/gliner-bi-llama-v1.0",
                                _attn_implementation = 'flash_attention_2',
                                                max_length = 2048).to('cuda:0', dtype=torch.float16)

If you have a large amount of entities and want to pre-embed them, please, refer to the following code snippet:

python

labels = ["your entities"]
texts = ["your texts"]

entity_embeddings = model.encode_labels(labels, batch_size = 8)

outputs = model.batch_predict_with_embeds(texts, entity_embeddings, labels)

Benchmarks

Below you can see the table with benchmarking results on various named entity recognition datasets:

Dataset	Score
ACE 2004	26.8%
ACE 2005	29.2%
AnatEM	25.3%
Broad Tweet Corpus	66.8%
CoNLL 2003	60.3%
FabNER	21.2%
FindVehicle	28.3%
GENIA_NER	58.3%
HarveyNER	18.3%
MultiNERD	64.7%
Ontonotes	28.4%
PolyglotNER	45.3%
TweetNER7	35.9%
WikiANN en	53.6%
WikiNeural	73.4%
bc2gm	63.2%
bc4chemd	56.8%
bc5cdr	71.3%
ncbi	64.9%
Average	47.0%

CrossNER_AI	56.7%
CrossNER_literature	61.5%
CrossNER_music	70.2%
CrossNER_politics	75.6%
CrossNER_science	66.8%
mit-movie	39.9%
mit-restaurant	41.7%
Average (zero-shot benchmark)	58.9%

Join Our Discord

Connect with our community on Discord for news, support, and discussion about our models. Join Discord.