Incorporating samples into convolution operation

Is it possible to feed a five-dimensional tensor into the convolution operation? That is, normally we input [batch_size, channels, featuremapx, featuremapy] to the torch.Conv2D operation. Is there any way to expands the allowed dimensions possible to feed a vector: [batch_size, samples, channels, featuremapx, featuremapy] ? For example, you feed the tensor of dimensions: [100,150,10, 64,64] and output the tensor of size [100,150,20,32,32]?

Maybe Conv3d is what you are looking for