[Solved] 'y = x' or 'y = x.detach()' in inference mode?

[whr94621 (Wei Haoran)]

After update to pytorch 0.4.0, I am confused about the keyward detach
What is the difference between

y = x
y = x.detach()
y = x.clone()

Especially when I do not need backward computation.

[ptrblck]

I think @richard provided a nice explanationhere:

Clone and detach in v0.4.0

We’ll provide a migration guide when 0.4.0 is officially released. Here are the answers to your questions: tensor.detach() creates a tensor that shares storage with tensor that does not require grad. tensor.clone()creates a copy of tensor that imitates the original tensor's requires_grad field. You should use detach() when attempting to remove a tensor from a computation graph, and clone as a way to copy the tensor while still keeping the copy as a part of the computation graph it came from. …

Does this answer your question?

[whr94621 (Wei Haoran)]

Yeah that answers my question. Also thank you for your link.