Breaking Gradients: x += 1 vs x = x + 1 in PyTorch

Quiz time: When doing automatic differentiation in PyTorch why x += 1 has very different consequences than x = x + 1?

Discussion