Understanding...

• Understanding torch.autograd.grad

• Understanding np.einsum

• Understanding CNN layer sizes

• Transpose conv vs. upsample + conv

pytorch tensor manipulation:

• Roll 2D tensor with variable step size per-row

• Add blocks of values to a tensor at specific locations

• Transform matrix columns into wrapping diagonal matrices

• Creating a positional tensor from a given tensor

• Mapping a tensor in PyTorch

• Place ones for all columns before certain index given per rows