ModuleNotFoundError: No module named 'models.arch_util' about diffusion-posterior-sampling

[Siddharth-Shrivastava7]

Hi @erhathaway,
you need to make following changes to work non-linear deblur task work ( I am assuming that you cloned non linear deblur repository as bkse name and moved inside diffusion posterior sampling repo) :

1. In diffusion-posterior-sampling/bkse/models/kernel_encoding/kernel_wizard.py :

• Change import models.arch_util as arch_util into import bkse.models.arch_util as arch_util
• Change from models.backbones.resnet import ResidualBlock_noBN, ResnetBlock into from bkse.models.backbones.resnet import ResidualBlock_noBN, ResnetBlock
• Change from models.backbones.unet_parts import UnetSkipConnectionBlock into from bkse.models.backbones.unet_parts import UnetSkipConnectionBlock

2. In diffusion-posterior-sampling/bkse/models/backbones/resnet.py :

• Change from models.arch_util import initialize_weights into from bkse.models.arch_util import initialize_weights

Download Pretrained GOPRO_wVAE.pth from blur-kernel-space-exploring repo (link: GOPRO_wVAE.pth and then move it to diffusion-posterior-sampling/bkse/experiments/pretrained/.

3. In diffusion-posterior-sampling/bkse/options/generate_blur/default.yml :

• Change experiments/pretrained/kernel.pth into $HOME/bkse/experiments/pretrained/GOPRO_wVAE.pth

4. In diffusion-posterior-sampling/bkse/options/generic_deblur/default.yml:

• Change experiments/pretrained/kernel.pth into $HOME/diffusion-posterior-sampling/bkse/experiments/pretrained/kernel.pth
• Change experiments/pretrained/GOPRO_wVAE.pth into $HOME/diffusion-posterior-sampling/bkse/experiments/pretrained/GOPRO_wVAE.pth

Hope this helps! :~)

[tomtom1103]

Hi @erhathaway, you need to make following changes to work non-linear deblur task work ( I am assuming that you cloned non linear deblur repository as bkse name and moved inside diffusion posterior sampling repo) :

1. In diffusion-posterior-sampling/bkse/models/kernel_encoding/kernel_wizard.py :

• Change import models.arch_util as arch_util into import bkse.models.arch_util as arch_util
• Change from models.backbones.resnet import ResidualBlock_noBN, ResnetBlock into from bkse.models.backbones.resnet import ResidualBlock_noBN, ResnetBlock
• Change from models.backbones.unet_parts import UnetSkipConnectionBlock into from bkse.models.backbones.unet_parts import UnetSkipConnectionBlock

2. In diffusion-posterior-sampling/bkse/models/backbones/resnet.py :

• Change from models.arch_util import initialize_weights into from bkse.models.arch_util import initialize_weights

Download Pretrained GOPRO_wVAE.pth from blur-kernel-space-exploring repo (link: GOPRO_wVAE.pth and then move it to diffusion-posterior-sampling/bkse/experiments/pretrained/.

3. In diffusion-posterior-sampling/bkse/options/generate_blur/default.yml :

• Change experiments/pretrained/kernel.pth into $HOME/bkse/experiments/pretrained/GOPRO_wVAE.pth

4. In diffusion-posterior-sampling/bkse/options/generic_deblur/default.yml:

• Change experiments/pretrained/kernel.pth into $HOME/diffusion-posterior-sampling/bkse/experiments/pretrained/kernel.pth
• Change experiments/pretrained/GOPRO_wVAE.pth into $HOME/diffusion-posterior-sampling/bkse/experiments/pretrained/GOPRO_wVAE.pth

Hope this helps! :~)

works perfectly! thank you.

[HJ-harry]

Thanks @Siddharth-Shrivastava7 for your kind answer! Closing the issue as there are no more activities.

[DPS2022]

Hello, did you clone two repositories in the readme? For non-linear deblur, we used an external repository to simulate. Please check it.