datawhalechina/learn-nlp-with-transformers

Structured around foundational theory and applied practice, the curriculum progresses from attention mechanisms and transformer architectures (with PyTorch implementations) through BERT and GPT variants, then to downstream tasks including text classification, sequence tagging, extractive QA, and generation tasks like machine translation and summarization. Leverages the Hugging Face Transformers library as the primary implementation framework, with all content and examples delivered in Chinese to lower barriers for Mandarin-speaking NLP practitioners.

3,143 stars. No commits in the last 6 months.