Understanding rate-distortion performance in distributed transformer inference
Split computing for language models, extending the theory of usable information.
Transformers
An architecture for accelerated large-scale inference of transformer-based language models
Unified batch and online transformer inference.