Fetching the paper…

Exploiting Sparsity for Long Context Inference: Million Token Contexts on Commodity GPUs · Around