TensorRT acceleration based on deep learning OFDM channel compensation

Abstract

In order to accelerate the DNN module computing speed and significantly improve the operation efficiency of the miniaturized deep learning communication platform. TensorRT-based DNN computing acceleration technology is studied and implemented. By performing model preservation, model parsing, engine optimization and inference deployment on trained DNNs, the DNN network structure and quantized data volume are effectively optimized. The performance of the TensorRT-accelerated models is tested, and the results show that the TensorRT-based DNN computation acceleration technique has shorter computation time and higher computation throughput with high performance.