Transformers

This handles Hugging Face's transformers library for working with pre-trained models across NLP, vision, and audio tasks. You get pipelines for quick inference on things like text generation, classification, and question answering, plus lower-level APIs when you need fine-tuning or custom preprocessing. The skill covers the standard workflow: load a model with AutoModel classes, tokenize inputs, run inference or training with the Trainer API. It's comprehensive without being overwhelming, though you'll still need to know which model architecture fits your task. The pipeline API is genuinely convenient for prototyping, and the references break down generation strategies, tokenization, and training patterns clearly enough to get productive fast.