update summery

[vanAken]

Hi there Your idea of using the transformer for event-based depth estimation is great.

I'm working on Python 3.9 and CUDA 11.8 and trying to update the environment.

The first is the summery function in train.py error discription:

https://discuss.pytorch.org/t/lstm-tuple-object-has-no-attribute-size/100397

from torchsummary import summary

from torchinfo import summary

#summary(model, input_size=[(c)], batch_size=1, device='cpu')
batch_size=1
summary(model, (batch_size, 3, 224, 224))

In model/loss.py, PerceptualSimilarity is changed to a newer version called LPIPS. https://github.com/richzhang/PerceptualSimilarity

from PerceptualSimilarity import models

import lpips

and in line 180ff

class perceptual_loss(): def init(self, net='vgg', use_gpu=True): """ Wrapper for PerceptualSimilarity.models.PerceptualLoss import lpips loss_fn_alex = lpips.LPIPS(net='alex') # best forward scores loss_fn_vgg = lpips.LPIPS(net='vgg') # closer to "traditional" perceptual loss, when used for optimization """ self.model = lpips.LPIPS(net='vgg') # volker models.PerceptualLoss(net=net, use_gpu=use_gpu)

self.weight = weight

In config/*.json I set the pretrained model path to None in line 72, which is null in json:
"pretrained": null

but now the error

File "/.../EReFormer/model/swin.py", line 326, in forward
x = x.view(B, C, H, W)
RuntimeError: shape '[1, 1536, 8, 11]' is invalid for input of size 75264

 x.shape was input as 

 torch.Size([1, 49, 768])

 and increase and transposed before the error occurred.

 1 1536 49

Are the 2 updates I am using doing the right thing? How do I get x into the right shape, and what should it be for the data given?

Before I dive deeper into the code, it would be nice if you could give me some advice on how to overcome this problem.

[Kingtorah]

Hi,have you solved the problem?I have exactly the same problem as you.Can your code run now?If so,please send me a copy,thank you.

[Kingtorah]

Can we discuss it?My email address is loiroyaletorah@gmail.com

[vanAken]

I stop working on this, but i'm still interested in. I wait for an update. I'm on vacation for some weeks ...