本页尚未翻译

本页内容仍为英文原文，源自官方文档 integrations/embedding/huggingface。中文翻译正在分批进行中。

HuggingFace

We offer support for all HuggingFace models (which can be loaded via transformers library). The default model is colbert-ir/colbertv2.0 which also has its own special callout - registry.get("colbert"). Some HuggingFace models might require custom models defined on the HuggingFace Hub in their own modeling files. You may enable this by setting trust_remote_code=True. This option should only be set to True for repositories you trust and in which you have read the code, as it will execute code present on the Hub on your local machine. Pin revision to a commit hash, and note that tables using such a model need it approved with get_registry().allow_remote_code(...) when opened in a new process. See Trusting remote code and custom endpoints.

Example usage:

import tempfile
from pathlib import Path

import lancedb
import pandas as pd
from lancedb.embeddings import get_registry
from lancedb.pydantic import LanceModel, Vector

db = lancedb.connect(str(Path(tempfile.mkdtemp()) / "huggingface-demo"))
model = get_registry().get("huggingface").create(name="facebook/bart-base")

class Words(LanceModel):
    text: str = model.SourceField()
    vector: Vector(model.ndims()) = model.VectorField()

df = pd.DataFrame({"text": ["hi hello sayonara", "goodbye world"]})
table = db.create_table("greets", schema=Words)
table.add(df)
query = "old greeting"
actual = table.search(query).limit(1).to_pydantic(Words)[0]
print(actual.text)