XLM is a modular, research-friendly framework for developing and comparing non-autoregressive language models. Built on PyTorch and PyTorch Lightning, with Hydra for configuration management, XLM makes it effortless to experiment with cutting-edge NAR architectures.
