AdaSeq is an all-in-one library built on ModelScope and PyTorch that enables researchers and developers to train custom models for sequence understanding tasks. It provides state-of-the-art models, training methods, and toolkits for tasks including named entity recognition (NER), part-of-speech tagging, chunking, entity typing, and relation extraction. The library supports modular registration and assembling of predefined modules for customized model development.