Happy Transformer

Happy Transformer is a Python package that makes it easier to fine tune and run inference with NLP transformer models. It wraps common model families behind a small, consistent interface so developers can try a model or train it on their own data without writing the surrounding boilerplate. The documentation lists the supported tasks and which of them support training. Text generation, text classification, question answering, word prediction and text to text all support both basic usage and training. Token classification and next sentence prediction are available for inference. Tutorials walk through text classification with training, hate speech detection, sentiment analysis, text generation with training using GPT Neo, and word prediction with training using DistilBERT and RoBERTa. The site also lists news and maintainers, and links to GitHub and a Discord community. It is aimed at developers, students and researchers who want a gentle entry into transformer models rather than a full training framework, and who are comfortable working in code. The project is a library and documentation site rather than a hosted service, so there is no account to create. Pricing is not published. The site has no pricing page and sells nothing. Consult the project's GitHub repository for license terms.