Introduction

After model tuning during conversion through operator elimination, fusion, and partition by ATC, the computation results of Huawei proprietary operators may be different from those of third-party equivalents (for example, from Caffe, TensorFlow, and ONNX). A tool must be designed to compare the computation results for developers to quickly resolve the operator accuracy issues.

Model Accuracy Analyzer analyzes model accuracy by comparing the computation results of Huawei proprietary operators and those of standard Caffe, TensorFlow, and ONNX operators to locate the faults.

In this sample, the standard data is the .npy data of a TensorFlow model, and the data to be compared is the dump data generated when an offline model runs on the Ascend AI Processor.