Weights[0] in-place operations

Part 1 (2020)
1

Hello everyone, I’m stuck at lesson 4 with a error on the following line:

weights[0] *= 1.0001

I get the following error:

RuntimeError: a view of a leaf Variable that requires grad is being used in an in-place operation.

From the pytorch forum:

PyTorch doesn’t allow in-place operations on leaf variables that have requires_grad=True (such as parameters of your model) because the developers could not decide how such an operation should behave.

So, how are you people doing that?
Thank you

2

Hi Midst

Which page in the book. Weights is a matrix. Try weights.shape because weights[0] seems wrong.

Regards Conwyn

1 Like
3

hi, this should work

     with torch.no_grad():
         weights[0] *= 1.0001
7 Likes
4

it is in the The MNIST Loss Function paragraph. Thank you Conwyn

5

Hi Mugnaio, I tried and it works thanks. Any idea why this no gradient isn’t specified in the lesson?

1 Like
6

Hi Midst

You can report it here

Regards Conwyn

7

Hi,
I tried:

weights[0].data *= 1.0001

and it also works
BR,
Raviv

3 Likes
9

I met the same problem and solved by this post.
but can anybody help to explaine why we can not do this “weights[0] *= 1.0001” directly?
thank you