TurboBatch accelerates transformer inference by up to 10.2x with dynamic batching. It's lightweight, HuggingFace-compatible, and ideal for real-time NLP tasks.
