03-07-2021 06:59 PM
03-09-2021 07:38 AM
MobileNet model has 27 Convolutions layers which includes 13 depthwise Convolution, 1 Average Pool layer, 1 Fully Connected layer and 1 Softmax Layer.
Depthwise convolution is not supported in DPUCVDX8G (please refer to UG1414, pag 90)
03-09-2021 03:11 PM - edited 03-09-2021 03:12 PM
Hi, gguasti. Thanks for your response.
As you said, I was asking a question after reading UG1414, page 90.
I think the VCK190's DPU will perform better than the ZCU102's DPU.
However, I don't understand that depthwise-conv2d supported by the DPU of ZCU102 is not supported by the DPU of VCK190.
Does the VCK190's DPU have any plans to support depthwise-conv2d in the future?