Kernel and Graph Optimization for DL Model Execution