I am on the latest stable release (0.4.1). I get an error module 'torch' has no attribute 'mul_'
when I run the following code:
import torch
a = torch.rand([5, 5])
torch.mul_(a, -1)
print('Done!!!')
What might be causing this? Any ideas?
I am on the latest stable release (0.4.1). I get an error module 'torch' has no attribute 'mul_'
when I run the following code:
import torch
a = torch.rand([5, 5])
torch.mul_(a, -1)
print('Done!!!')
What might be causing this? Any ideas?
Hi,
All the operation that finish with _
are inplace operations, this means you will need a Tensor to perform this operation on and they as not available as torch.*
functions, you can do a.mul_(-1)
though.
Thanks, your suggestion works. But, then this behavior is not consistent throughout. For instance, torch.abs_(a)
works but torch.mul_(a, -1)
does not. Any reason why is this the case?
There are some lingering from old versions.
As you can see in the doc for torch.*, no inplace function is listed. They are only listed in the doc for Tensors.
Got it. Thanks for the explanation.