We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
A differentiable argmax function in PyTorch, for your gradient descending convenience.