Skip to main content

Predacons

Predacons is a Python library built on top of the Hugging Face Transformers library, designed to simplify and enhance various Natural Language Processing (NLP) tasks. It provides a user-friendly interface for data loading, model training, and text generation, enabling you to leverage the power of transfer learning with ease.

PyPI Downloads License Python Version

Purpose​

The primary goal of Predacons is to abstract away the complexities involved in working with transformer models, making them more accessible to a wider audience. Whether you're a seasoned machine learning practitioner or just getting started with NLP, Predacons offers a set of tools to streamline your workflow and accelerate your projects.

Key Features​

Predacons offers a comprehensive suite of functionalities, including:

  • Easy Data Loading: Load text data from various sources, including directories, individual files, and CSV files, with automatic handling of encoding.
  • Text Cleaning Utilities: Preprocess your text data with built-in cleaning functions to remove noise and improve model performance.
  • Simplified Model Training: Train transformer models on your custom datasets with minimal code, leveraging the power of the Hugging Face Trainer. Supports both standard and legacy training methods.
  • Versatile Text Generation: Generate text using trained models with various options, including fast generation using speculative decoding, temperature control, and chat templates.
  • Text Streaming: Stream text generation output for real-time applications.
  • Chat Generation and Streaming: Generate and stream chat responses, suitable for building conversational AI applications.
  • Data Preparation Tools: Generate synthetic datasets using OpenAI or local language models to augment your training data.
  • Model and Tokenizer Loading: Easily load pre-trained models and tokenizers from local paths or the Hugging Face Model Hub.
  • Embeddings Generation: Generate embeddings for sentences using pre-trained transformer models, with full compatibility with Langchain methods.

Installation​

To get started with Predacons, simply install it using pip:


pip install predacons


Quick Start​

Here's a basic example to demonstrate how to use Predacons for text generation:


from predacons import predacons

# Initialize Predacons (prints available functions)
predacons.rollout()

# Load a pre-trained model (replace with your model path)
model_path = "path/to/your/model"  # e.g., "bert-base-uncased"
sequence = "The quick brown fox"

# Generate text
generated_text = predacons.generate_text(model_path=model_path, sequence=sequence, max_length=50)
print(f"Generated text: {generated_text}")

# Example usage for chat generation
chat_sequence = [{"role": "user", "content": "Tell me a joke."}]
generated_chat = predacons.chat_generate(model_path=model_path, sequence=chat_sequence, max_length=100)
print(f"Generated chat: {generated_chat}")


Documentation​

For detailed information on specific functions and modules, please refer to the following documentation pages:

Contributing​

We welcome contributions to Predacons! If you have any ideas, suggestions, or bug reports, please open an issue on the GitHub repository. If you'd like to contribute code, please submit a pull request.

License​

This project is licensed under multiple licenses:

  • For free users, the project is licensed under the terms of the GNU Affero General Public License (AGPL). See LICENSE-AGPL for more details.

  • For paid users, there are two options:

Please ensure you understand and comply with the license that applies to you.