Skip to content

Latest commit

 

History

History
36 lines (27 loc) · 615 Bytes

File metadata and controls

36 lines (27 loc) · 615 Bytes

HuggingFace Ingestion

IceFrame allows you to ingest datasets directly from the HuggingFace Hub.

Usage

from iceframe import IceFrame
from iceframe.utils import load_catalog_config_from_env

config = load_catalog_config_from_env()
ice = IceFrame(config)

# Read a dataset
ice.create_table_from_huggingface(
    "my_namespace.imdb", 
    "imdb", 
    split="train"
)

# With specific configuration
ice.create_table_from_huggingface(
    "my_namespace.glue_mrpc",
    "glue",
    name="mrpc",
    split="validation"
)

Dependencies

Requires datasets.

pip install "iceframe[hf]"