fix dot product for newer version of torch

[shaanrockz]

fix(min_norm_solvers.py): change torch.dot -> torch.mul followed by sum() for Matrix dot product
fix(min_norm_solvers.py): change data[0] to data.cpu()

[allenjack]

Hi, when running the code under my setting, I also got the issue "RuntimeError: 1D tensors expected, got 2D". How I solve this issue is that I flatten the original tensor gradients and make it fit into the torch.dot function.

If what I am doing is correct, we are simply doing an element-wise matrix followed by sum() instead of torch.mul.

Hope someone can clarify it.

[shaanrockz]

Hi, when running the code under my setting, I also got the issue "RuntimeError: 1D tensors expected, got 2D". How I solve this issue is that I flatten the original tensor gradients and make it fit into the torch.dot function.

If what I am doing is correct, we are simply doing an element-wise matrix followed by sum() instead of torch.mul.

Hope someone can clarify it.

Hi,
Sorry for late reply. Yes, that will also work. I however found another issue with the '.data[0]' for newer version of torch, we need to convert it to cpu if the tensors are in gpu.
torch.dot() requires 1D tensor inputs, that is why I used torch.mul() followed by sum().
Cheers

[ozansener]

Looks good to me