Is there any case to verify depthwise convolution in CLDNN?

[axinging]

Hi, I noticed that you have implemented several depthwise convolution related cl kernels(see below). But I didn't see any related case to use these kernel.

Also, you mentioned the validated topologies include ": AlexNet*, VGG(16,19), GoogleNet(v1,v2,v3), ResNet(50,101,152)* Faster R-CNN*, Squeezenet*, SSD_googlenet*, SSD_VGG*, PVANET*, PVANET_REID*, age_gender*, FCN* and yolo*.", but not include mobilenet.

So does it mean the depthwise convolution cl kernels haven't been verified yet? If not, do you have any samples to share?

depthwise convolution related cl kernels: https://github.com/intel/clDNN/blob/master/kernel_selector/core/cl_kernels/convolution_gpu_bfyx_f16_depthwise.cl https://github.com/intel/clDNN/blob/master/kernel_selector/core/cl_kernels/convolution_gpu_bfyx_depthwise_weights_lwg.cl https://github.com/intel/clDNN/blob/master/kernel_selector/core/cl_kernels/convolution_gpu_byxf_af32_depthwise.cl