🐛 Bug

Running the c++ example fails when running the forward calculation on the generator:

what(): Calculated padded input size per channel: (1 x 1). Kernel size: (4 x 4). Kernel size can't be greater than actual input size (check_shape_forward at /pytorch/aten/src/ATen/native/Convolution.cpp:436)

Is this due to recent changes to use NNPACK for strided convolutions..?

(There's also a warning about batchnorm being deprecated,
but that's a simpler fix.)

Thanks

cc @yf225