Hugging Face (Meta-package)

0.0.1 · active · verified Fri Apr 17

The `huggingface` PyPI package (version 0.0.1) is a meta-package designed to simplify installation by pulling in the main Hugging Face libraries: `transformers`, `datasets`, and `huggingface_hub`. It does not provide its own direct API functionality. This package typically has a very low release cadence, primarily updating its dependencies.

Common errors

```
AttributeError: module 'huggingface' has no attribute 'pipeline'
```
cause Trying to import a function/class like `pipeline` directly from the `huggingface` meta-package.

fix Functionality like `pipeline` resides in the `transformers` library. Change your import to `from transformers import pipeline`.
```
ModuleNotFoundError: No module named 'huggingface.models'
```
cause Incorrectly assuming that the `huggingface` meta-package establishes a hierarchical module structure like `huggingface.models` or `huggingface.datasets`.

fix The core libraries (`transformers`, `datasets`, `huggingface_hub`) are installed as separate, top-level packages. You should import from them directly, e.g., `from transformers import AutoModel`.

Warnings

gotcha The `huggingface` PyPI package is a meta-package (version 0.0.1) and does not provide direct API functionality. It primarily serves to install other core Hugging Face libraries (`transformers`, `datasets`, `huggingface_hub`).
Fix: Always import classes and functions directly from `transformers`, `datasets`, or `huggingface_hub`, not `huggingface`.
gotcha Attempting to import or use functionality directly from `huggingface` (e.g., `from huggingface import pipeline`) will result in an `AttributeError` or `ModuleNotFoundError`.
Fix: Identify which specific Hugging Face library provides the desired functionality (e.g., `transformers` for `pipeline`, `datasets` for `Dataset`) and import from that package.

Install

pip install huggingface Install `huggingface` and its dependencies

Imports

pipeline
wrong:
```
from huggingface import pipeline
```
correct:
```
from transformers import pipeline
```
The `huggingface` package itself does not expose an API; functionality comes from its sub-libraries like `transformers`.
Dataset
wrong:
```
from huggingface import Dataset
```
correct:
```
from datasets import Dataset
```
For dataset management, import directly from the `datasets` library.
HfApi
wrong:
```
from huggingface import HfApi
```
correct:
```
from huggingface_hub import HfApi
```
For interacting with the Hugging Face Hub, use the `huggingface_hub` client.

Quickstart

The `huggingface` package itself does not expose an API. This quickstart demonstrates common usage of the `transformers` library (specifically the `pipeline` abstraction), which is automatically installed as a dependency by `huggingface`. All direct functionality will come from `transformers`, `datasets`, or `huggingface_hub`.

from transformers import pipeline

# The huggingface meta-package itself does not provide an API.
# This quickstart demonstrates typical usage of the `transformers` library,
# which is installed as a dependency by `huggingface`.

classifier = pipeline("sentiment-analysis")
result = classifier("I love using Hugging Face libraries!")
print(result)

# Example of using a dataset (also installed by huggingface meta-package)
from datasets import load_dataset
# dataset = load_dataset("imdb") # Uncomment to load a large dataset
# print(dataset)

view raw JSON →