A GPU-accelerated deep learning library for Python using CUDA via PyCUDA, implementing neural networks with various training methods.
Hebel is a GPU-accelerated deep learning library for Python that implements neural network models using CUDA via PyCUDA. It provides feed-forward neural networks for classification and regression tasks along with various training methods like momentum, dropout, and early stopping. The library aims to make GPU-accelerated deep learning more accessible to Python developers.
Python developers and researchers who need GPU-accelerated neural network training for classification and regression tasks, particularly those working with CUDA-capable hardware.
Developers choose Hebel for its straightforward implementation of essential deep learning functionality with GPU acceleration, offering practical training methods and regularization techniques in a single Python package.
GPU-Accelerated Deep Learning Library in Python
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Leverages CUDA via PyCUDA to significantly speed up neural network training on compatible hardware, as emphasized in the library's core feature set.
Implements essential optimization techniques like stochastic gradient descent with momentum, Nesterov momentum, and regularization methods including dropout and weight decay for effective model training.
Provides a clear and accessible deep learning toolkit focused on core functionality, making GPU-accelerated neural networks more approachable for Python developers.
Compatible with Linux, Windows, and likely macOS, ensuring broad accessibility across different operating systems without major restrictions.
The author explicitly states that active development has stopped and recommends using Chainer instead, meaning no bug fixes, security updates, or new features will be added.
Only implements feed-forward neural networks; planned models like autoencoders and CNNs were never added, restricting its use to basic classification and regression tasks.
Requires PyCUDA, which depends on CUDA toolkits and drivers, making installation and configuration more challenging compared to modern frameworks with simpler GPU integration.