happy-transformer

Why we included this project

Happy Transformer wraps the Hugging Face transformers library behind a small set of classes for common NLP tasks. You can load a pretrained model and run text generation, text classification, or text-to-text inference in a few lines, and the same objects expose train(), eval(), and test() methods, so fine-tuning on your own labeled CSV data is straightforward. That's a practical on-ramp for teams that need a working classifier or generator without writing training loops from scratch. Several older task classes (word prediction, question answering, next sentence prediction, token classification) are deprecated, and a 4.0 rewrite with breaking changes is in progress, so pin the version below 4.0.0 for stable behavior. If you want a thin, readable layer over transformers rather than a full framework, this is a convenient place to start.