# torch.Tensor.detach_¶

Tensor.detach_()

Detaches the Tensor from the graph that created it, making it a leaf. Views cannot be detached in-place.

This method also affects forward mode AD gradients and the result will never have forward mode AD gradients.