How is the output dimension of MPSCNNConvolutionTranspose calculated?
Let's say we have an input of N x C x H x W, kernelSize kH x kW, stride = 1, pad = 0,
Should it be
output_h = (H - 1) * stride + kH and
output_w = (W - 1) * stride + kW?
I got off by one result when trying out this API.