jinaai/jina-embeddings-v4

Note

This is a remotely-sourced model from the jina_embeddings_v4 plugin, maintained by the community. It is not part of FiftyOne core and may have special installation requirements. Please review the plugin documentation and license before use.

jina-embeddings-v4 is a universal embedding model for multimodal and multilingual retrieval. The model is specially designed for complex document retrieval, including visually rich documents with charts, tables, and illustrations..

Details

• Model name: jinaai/jina-embeddings-v4

• Model source: https://huggingface.co/jinaai/jina-embeddings-v4

• Model author: Jina AI

• Model license: Apache 2.0

• Exposes embeddings? yes

• Tags: embeddings, torch, visual-document-retrieval, zero-shot

Requirements

• Packages: huggingface-hub, transformers, torch, torchvision

• CPU support

  • yes

• GPU support

  • yes

Example usage

import fiftyone as fo
import fiftyone.zoo as foz

foz.register_zoo_model_source("https://github.com/harpreetsahota204/jina_embeddings_v4")

dataset = foz.load_zoo_dataset(
    "coco-2017",
    split="validation",
    dataset_name=fo.get_default_dataset_name(),
    max_samples=50,
    shuffle=True,
)

model = foz.load_zoo_model("jinaai/jina-embeddings-v4")

dataset.apply_model(model, label_field="predictions")

session = fo.launch_app(dataset)