yjchai

Results 17 comments of yjchai

> I installed all libs, converted them with help > `jupyter nbconvert --to script *.ipynb` > And try to start all python files. They were failed with different errors. Fo...

Same problem. When I train my model, the code has great performance. While testing, there is an error KeyError: 'module.encoder.1.weight_orig'

oading EdgeModel generator... Traceback (most recent call last): File "train.py", line 2, in main(mode=1) File "/media/sdc/yjchai/edge-connect/main.py", line 50, in main model.load() File "/media/sdc/yjchai/edge-connect/src/edge_connect.py", line 53, in load self.edge_model.load() File "/media/sdc/yjchai/edge-connect/src/models.py",...

Yes, the output is right as you said. But how to fix the Key error?

encoder.1.bias encoder.1.weight_orig encoder.1.weight_u encoder.1.weight_v encoder.4.bias encoder.4.weight_orig encoder.4.weight_u encoder.4.weight_v encoder.7.bias encoder.7.weight_orig encoder.7.weight_u encoder.7.weight_v middle.0.conv_block.1.weight_orig middle.0.conv_block.1.weight_u middle.0.conv_block.1.weight_v middle.0.conv_block.5.weight_orig middle.0.conv_block.5.weight_u middle.0.conv_block.5.weight_v middle.1.conv_block.1.weight_orig middle.1.conv_block.1.weight_u middle.1.conv_block.1.weight_v middle.1.conv_block.5.weight_orig middle.1.conv_block.5.weight_u middle.1.conv_block.5.weight_v middle.2.conv_block.1.weight_orig middle.2.conv_block.1.weight_u middle.2.conv_block.1.weight_v middle.2.conv_block.5.weight_orig middle.2.conv_block.5.weight_u middle.2.conv_block.5.weight_v...

![image](https://user-images.githubusercontent.com/25547974/57215161-ac168e80-701e-11e9-8b72-e5af90527d5f.png)

Instead of directly using 'bash ./scripts/download_model.sh', I download your pre-trained model from Google drive due to the limitation of web speed. Is there something wrong in the process?

Yeah! You are right! Thank you for your patience.

BTW I suggest you put the original image(before adding mask) in your examples, therefore people who use it could easily compare the ground truth with inpainted image and use them...