What problem does it solve? Working with pre-trained transformer models involves repetitive boilerplate for loading models, tokenizing inputs, configuring generation, and setting up training loops. This Skill provides structured guidance and code patterns for the Hugging Face Transformers library across NLP, vision, audio, and multimodal tasks. ## Core Features & Use Cases - Pipeline Inference: Run text generation, classification, question answering, translation, summarization, image classification, object detection, and speech recognition with the Pipeline API. - Model Management: Load models with control over device placement, precision (float16/bfloat16), quantization (8-bit/4-bit), and attention implementations. - Fine-Tuning: Train models on custom datasets using the Trainer API with mixed precision, gradient accumulation, LoRA/PEFT, and DeepSpeed support. - Use Case: Fine-tune a BERT classifier on your own labeled dataset, then generate text with a causal language model using sampling parameters like temperature and top-p. ## Quick Start Use the transformers skill to load a pre-trained model and run text classification on my input text.
