xyliu

Linear-Transformer. Transformer are RNNs: Fast Autoregressive Transformer with Linear Attention

LPT. This repo contains the code for Late Prompt Tuning.

awesome-early-exiting. A curated list of Early Exiting papers, benchmarks, and misc.

RoSE. The code for paper "Making Large Language Models Better Reasoners with Orchestrated Streaming Experiences"