We simplify and optimize neural network inference for Arm Cortex-M55 and Ethos-U55, Xilinx Kria based Edge ML solutions. We leverage TVM, uTVM, TF Lite, Vela, Vitis AI, and LLVM. We accomplish this via ML compiler tuning and fine manual optimization on the assembler level.
Network-based Transfer Learning
Semiconductors – TVM
Translator for AI accelerator (Tensor processor)
TVM optimization for ARMv7-M+