serving-infrastructure

Tag

Cards List
#serving-infrastructure

Lifecycle-Optimal Tokenization: Vocabulary Size as a Deployment-Regime-Dependent Infrastructure Parameter

arXiv cs.LG · 3d ago Cached

This paper argues that the optimal tokenizer vocabulary size is not fixed but depends on the deployment regime, such as batch size and inference volume. Through roofline analysis and experiments on A10G and A100 GPUs, it shows the lifecycle-optimal vocabulary can shift by up to 16x between on-device and datacenter serving, with minimal quality impact.

0 favorites 0 likes
← Back to home

Submit Feedback