https://developer.nvidia.com/blog/mastering-llm-techniques-inference-optimization/
Flash Attention Review