FileNotFoundError: [Errno 2] No such file or directory: '/fs/scratch/rng_cr_bcai_dl/lyu7rng/pretrained/vgg16-tv.pth'

[MISSEY]

Where to get the vgg16-tv.pth ? If it is download from pytorch hub, there is no configuration to set from where to download

[MISSEY]

When I load the vgg.pth from weights/v0.1/vgg.pth, it gave me following error :

RuntimeError: Error(s) in loading state_dict for VGG:
    Missing key(s) in state_dict: "features.0.weight", "features.0.bias", "features.2.weight", "features.2.bias", "features.5.weight", "features.5.bias", "features.7.weight", "features.7.bias", "features.10.weight", "features.10.bias", "features.12.weight", "features.12.bias", "features.14.weight", "features.14.bias", "features.17.weight", "features.17.bias", "features.19.weight", "features.19.bias", "features.21.weight", "features.21.bias", "features.24.weight", "features.24.bias", "features.26.weight", "features.26.bias", "features.28.weight", "features.28.bias", "classifier.0.weight", "classifier.0.bias", "classifier.3.weight", "classifier.3.bias", "classifier.6.weight", "classifier.6.bias". 
    Unexpected key(s) in state_dict: "lin0.model.1.weight", "lin1.model.1.weight", "lin2.model.1.weight", "lin3.model.1.weight", "lin4.model.1.weight". 

[YumengLi007]

Hi, from here, 'https://download.pytorch.org/models/vgg16-397923af.pth' is the one for VGG16

[MISSEY]

Thank you so much for your promt response :)

[MISSEY]

Could you explian why this line is commented out #self.parameters = list(self.net.parameters()) here https://github.com/boschresearch/ISSA/blob/main/training/lpips/dist_model.py#L77

gives me this error

TypeError: unsupported operand type(s) for +=: 'method' and 'list'

here https://github.com/boschresearch/ISSA/blob/main/training/lpips/dist_model.py#L82

Also found out that in rosinality repo line is not commented out https://github.com/rosinality/id-gan-pytorch/blob/master/stylegan2/lpips/dist_model.py#L86

Thankyou

[YumengLi007]

are you passing 'is_train=False'? Since we don't want to train this part, it's just for feature extraction and loss computation.

[MISSEY]

Yes, I pass the is_train = True. I thought, I need to train the dist_model too. Ah I see, so for training the encoder, do not need to train dist_model ?

[YumengLi007]

right, you don't need to train VGG16, it's just for LPIPS in the loss

[MISSEY]

Thnak you