AI Inference & Serving
Padding
Adding designated filler positions so sequences can meet a required shape, such as a common batch length. Padding is different from truncation, which removes content. A model needs the appropriate masks and conventions to avoid treating filler as ordinary input; those conventions depend on the task, tokenizer and implementation.
Reviewed
Sources
Free complete lesson
This definition and the complete practical lesson are free. The interactive reader loads the lesson examples without a membership.