Using vLLM

vLLM supports the following usage patterns:

• Inference and Serving: Run a single instance of a model.
• Deployment: Scale up model instances for production.
• Training: Train or fine-tune a model.