yolov5 face mnn版本，在mnn1.0.2版本运行报错。
Don't support broadcast for binaryOp,
due to the internal logic of MNN, if your MNN model doesn't have input shape, you may ignore this 'Resize error' information:
** Resize error for [BinaryOp], BinaryOp719, code=3 **
it will work after you set the input tensor shape in MNN, and then resize the Session
后来自己把onnx转为mnn，报错：
Output: output: ========================================
Execute Error for [StridedSlice], 434, code=2
Error for StridedSliceComputer: inputDimension>=5: 5
-----------------------------------------------------------------------------------------------------------------------------
due to the internal logic of MNN, if your MNN m
