SVD Contextual Sparsity Predictors for Fast LLM Inference — Quantapedia
Contextual sparsity is one of the approaches used to reduce computational complexity in the inference process of large language models (LLMs). Existing techniques for efficient LLM inference accelerat