hflocal 0.1.0

A library to save, load, and use Hugging Face models locally.

hflocal

hflocal is a Python library designed to simplify the process of saving, loading, and using Hugging Face models locally. This library provides a user-friendly interface for handling pre-trained models from the Hugging Face repository.

Author

Anish KM

• GitHub: @AnishKMBtech

Features

• Save pre-trained models and tokenizers to a local directory
• Load pre-trained models and tokenizers from a local directory
• Use models with a simple pipeline interface for various NLP tasks

Installation

You can install the hflocal library using pip:

pip install hflocal

Usage

Saving a Model

To save a pre-trained model and its tokenizer to a local directory, use the save_model function:

from hflocal import save_model

# Save the 'bert-base-uncased' model to the './saved_model' directory
save_model('bert-base-uncased', './saved_model')

Loading a Model

To load a pre-trained model and its tokenizer from a local directory, use the load_model function:

from hflocal import load_model

# Load the model and tokenizer from the './saved_model' directory
model, tokenizer = load_model('./saved_model')

Using the Model Pipeline

To use a model with a simple pipeline interface for various NLP tasks, use the ModelPipeline class:

from hflocal import ModelPipeline

# Initialize the pipeline with the saved model
pipeline = ModelPipeline('./saved_model')

# Use the pipeline for inference
result = pipeline("Your input text here")
print(result)

Development

Setting Up the Development Environment

1. Clone the repository:

git clone https://github.com/AnishKMBtech/hflocal.git
cd hflocal

2. Create a virtual environment and activate it:

python -m venv venv
# On Unix/macOS:
source venv/bin/activate
# On Windows:
venv\Scripts\activate

3. Install the required dependencies:

pip install -r requirements.txt

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

• Hugging Face for providing the pre-trained models and tokenizers
• Transformers library for the model and tokenizer implementations