MLIF

在自定义RISC-V ASIPS上嵌入机器学习的挑战

Solution: End-to-End TinyML Deployment and Benchmarking Flow • [MLIF] (Machine Learning Interface) • Framework/target-independent abstraction layers for Target SW • [MLonMCU] • Provides support for • 15+ targets (mainly RISC-V simulators) • 6 backends ([TVM] and TFLM) • Handling of Dependencies • Analysis and Exploration methods • Designed with并行性/可重复性