Наборы данных Hugging Face

Работа с Hugging Face

Jacob H. Marquez

Lead Data Engineer

¹ https://huggingface.co/datasets

Установка пакета Datasets

pip install datasets

🌐 Доступ
📥 Загрузка
🔧 Использование
🤝 Публикация

¹ https://huggingface.co/docs/datasets/loading

Загрузка набора данных

from datasets import load_dataset

data = load_dataset("IVN-RIN/BioBERT_Italian")

Параметр split

data = load_dataset("IVN-RIN/BioBERT_Italian", split="train")

¹ https://huggingface.co/docs/datasets/v2.15.0/loading

Форматы набора данных Apache Arrow

Набор данных Apache Arrow

¹ https://arrow.apache.org/overview/

Обработка данных

data = load_dataset("IVN-RIN/BioBERT_Italian", split="train")


# Filter for pattern " bella "
filtered = data.filter(lambda row: " bella " in row['text'])
print(filtered)

Dataset({
    features: ['text'],
    num_rows: 1122
})

¹ https://huggingface.co/docs/datasets/process#select-and-filter

Обработка данных

# Select the first two rows
sliced = filtered.select(range(2))


print(sliced)

Dataset({features: ['text'], num_rows: 2})

# Extract the 'text' for the first row
print(sliced[0]['text'])

Concentrazioni atmosferiche di PCDD/PCDF...

¹ https://huggingface.co/docs/datasets/process#select-and-filter

Давайте потренируемся!

Работа с Hugging Face