Token samplers for large language models, providing strategies like top-k, top-p, and temperature sampling.