TensorLayer

Developed at Imperial College London, it was designed to be simultaneously usable by researchers who need low-level control and practitioners who need high-level convenience offering both explicit layer composition and simpler model construction APIs.

The library covers supervised learning (classification, regression, sequence modeling), unsupervised learning (autoencoders, VAEs, GANs), and reinforcement learning (DQN, A3C, PPO implementations), with utilities for data preprocessing, model serialization, visualization, and distributed training.

TensorLayer integrates with OpenAI Gym environments for reinforcement learning experiments and provides computer vision and NLP layer implementations aligned with state-of-the-art architectures in each domain.

TensorLayer is open-source under the Apache 2.0 license and was published with an associated research paper at ACM Multimedia 2017.

While the core ML ecosystem has consolidated significantly around PyTorch and high-level Keras APIs since TensorLayer's peak adoption, the library remains relevant for teams with existing TensorLayer codebases and for researchers who want a TensorFlow-based library with explicit reinforcement learning primitives.

The codebase is available on GitHub and includes comprehensive tutorials and example implementations.