An open-source library for hardware-aware model optimization. It learns soft gates to prune and fine-tune neural networks for specific GPUs, balancing latency and accuracy. Includes tools for laten…