x-lowlevel-vision / lam_demo Goto Github PK

I'm trying to run this line:
interpolated_grad_numpy, result_numpy, interpolated_numpy = Path_gradient(tensor_lr.numpy(), model, attr_objective, gaus_blur_path_func, cuda=True)

and I"m getting:
TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
I tied to change the 'tensor_lr.numpy()' to 'tensor_lr.cpu().numpy()' or 'tensor_lr.detach().cpu().numpy()' and it's still the same.
I would like to get some help please.

This is the colab link of this repo.

Hello. Your paper is very interesting, so I want to apply it to my experiment. I was going to experiment with the x2 super-resolution model, but I'm not sure if I did it right. How should I change LAM code to experiment with the x2 super-resolution model? I changed the zoomin param, scale param, factor param but I not sure about it.

Is it possible to add demo support for SwinIR as well, since it is an important work in the super-resolution area.
Thanks!

Sir, please can you let me know, what steps I have to follow for getting the attribution maps for my super-resolved results.
I am thinking that the code, that you have provided on Google Colab , it takes state of the art pre-trained models (rcan,edsr), what if I want to check on my model's output.

Sir please help me.

Hello author, I like your LAM work very much. If you have time, can you update colab. And I would like to ask about how I need to modify my own SR model for testing, thank you very much, and I am very sorry if I bother you

I am following the step-by-step reasoning in the Colab page you provided and I get the following error when Calculate the Local Attribution Maps.
TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

Hello, as a student from mainland china, I am not able to reach the Google Colab Demo, would any of you be so kind to share the jupyter notebook version of the Colab Demo on the github? Thank you.