Running RIS-Kernel Long Context Inference Demo 🔬 Generate long-context text using O(N log N) sparse attention