README.md

metadata

language:
  - en
license: mit

This model was trained with Neural-Cherche. You can find details on how to fine-tune it in the Neural-Cherche repository.

pip install neural-cherche

Retriever

from neural_cherche import models, retrieve
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"
batch_size = 32

documents = [
    {"id": 0, "document": "Food"},
    {"id": 1, "document": "Sports"},
    {"id": 2, "document": "Cinema"},
]

queries = ["Food", "Sports", "Cinema"]

model = models.SparseEmbed(
    model_name_or_path="raphaelsty/neural-cherche-sparse-embed",
    device=device,
)

retriever = retrieve.SparseEmbed(
    key="id",
    on=["document"],
    model=model,
)

documents_embeddings = retriever.encode_documents(
    documents=documents,
    batch_size=batch_size,
)

retriever = retriever.add(
    documents_embeddings=documents_embeddings,
)

queries_embeddings = retriever.encode_queries(
    queries=queries,
    batch_size=batch_size,
)

scores = retriever(
    queries_embeddings=queries_embeddings,
    batch_size=batch_size,
    k=100,
)

scores