Hugging Face Transformers library provides access to thousands of pre-trained models. Pipeline API simplifies common tasks. Fine-tuning adapts models to specific domains. Model Hub enables sharing and discovery.
Using Pipelines
Pipelines abstract model loading and preprocessing. Specify task and model to get started. Handle batching and device placement automatically. Support text, image, and audio inputs.
- Use pipelines for quick prototyping and common tasks
- Specify models or let pipeline choose defaults
- Handle batching for improved throughput
- Deploy with Inference API for production
- Fine-tune for domain-specific improvements
Model Selection
Model Hub hosts thousands of models. Filter by task, language, and license. Check model cards for capabilities and limitations. Compare model sizes versus performance.