We use TIDL convert caffe models(*.prototxt + *.caffemodel)，the tidl_log.txt display the following information:
TIDL algInit returned with error -1048!
============================= TIDL ERROR! ==============================
Pooling layer error, invalid stride width!
My SDK version is PROCESSOR_SDK_VISION_03_08_00_00. I refer to the documentation "TIDeepLearningLibrary_UserGuide.pdf":
Spatial Pooling Layer
– Average and Max Pooling are supported with stride 1, 2, 4 and kernel sizes of 2x2,3x3,4x4 etc. STOCHASTIC Pooling not supported.
My *.prototxt ，the "Pooling" layer, the stride have "stride: 2" and "stride: 4", the kernel_size have "kernel_size: 3" and "kernel_size: 5", all of these accord with TIDL Limitation content "Pooling are supported with stride 1, 2, 4 and kernel sizes of 2x2,3x3,4x4 etc."
So ,why appear the above error message?
