UER-py (Universal Encoder Representations) is an open-source toolkit for pre-training language models on general-domain corpora and fine-tuning on downstream NLP tasks. It provides modular implementations of major transformer architectures including BERT, GPT, GPT-2, RoBERTa, ALBERT, T5, BART, ELMO, and XLM-RoBERTa. The project includes a comprehensive model zoo with pre-trained models and supports tasks such as text classification and named entity recognition.