TypeError: unhashable type: 'list'

[emperor1412]

Hi, I follow steps as the document, and got this error while trying to run stylize.py

C:\Project_GIT\WCT-TF\WCT-TF>python stylize.py --checkpoints models/relu5_1 models/relu4_1 models/relu3_1 models/relu2_1 models/relu1_1 --relu-targets relu5_1 relu4_1 relu3_1 relu2_1 relu1_1 --style-size 512 --alpha 0.8 --style-path C:\Project_GIT\WCT-TF\WCT-TF\style_images\ --content-path C:\Project_GIT\WCT-TF\WCT-TF\content\ --vgg-path C:\Project_GIT\WCT-TF\WCT-TF\models\vgg_normalised.t7

Using TensorFlow backend. Loading VGG up to layer relu5_1 Traceback (most recent call last): File "stylize.py", line 120, in <module> main() File "stylize.py", line 44, in main device=args.device) File "C:\Project_GIT\WCT-TF\WCT-TF\wct.py", line 27, in __init__ self.model = WCTModel(mode='test', relu_targets=relu_targets, vgg_path=vgg_path) File "C:\Project_GIT\WCT-TF\WCT-TF\model.py", line 57, in __init__ self.vgg_model = vgg_from_t7(vgg_path, target_layer=deepest_target) File "C:\Project_GIT\WCT-TF\WCT-TF\vgg_normalised.py", line 16, in vgg_from_t7 t7 = torchfile.load(t7_file, force_8bytes_long=False) File "C:\Users\emper\AppData\Local\Programs\Python\Python36\lib\site-packages\torchfile.py", line 424, in load return reader.read_obj() File "C:\Users\emper\AppData\Local\Programs\Python\Python36\lib\site-packages\torchfile.py", line 370, in read_obj obj._obj = self.read_obj() File "C:\Users\emper\AppData\Local\Programs\Python\Python36\lib\site-packages\torchfile.py", line 385, in read_obj k = self.read_obj() File "C:\Users\emper\AppData\Local\Programs\Python\Python36\lib\site-packages\torchfile.py", line 386, in read_obj v = self.read_obj() File "C:\Users\emper\AppData\Local\Programs\Python\Python36\lib\site-packages\torchfile.py", line 370, in read_obj obj._obj = self.read_obj() File "C:\Users\emper\AppData\Local\Programs\Python\Python36\lib\site-packages\torchfile.py", line 387, in read_obj obj[k] = v TypeError: unhashable type: 'list'

Do you have any idea ? Thanks a lot

[eridgd]

This appears to be due to Windows incompatibility with the torchfile package, see this issue from someone attempting to use the same vgg_normalised.t7 model bshillingford/python-torchfile#12

Unfortunately, I don't have access to Windows, so until the torchfile author makes it compatible I'd suggest using a Linux VM.

[eridgd]

I've gotten access to a Windows VM and will see if there's a simple solution to get this working. If not, I'll try to convert it to TF/Keras and circumvent torchfile.

[emperor1412]

Thanks @eridgd , what a nice project

[bradleybauer]

This pull request bshillingford/python-torchfile#13 fixes the issue on windows.

[eridgd]

Thanks @xdaimon for the fix!

@emperor1412 I've created a new branch with the modified torchfile.py bundled and one other tweak required to load the model on Windows https://github.com/eridgd/WCT-TF/tree/windows-compat. Can you please check if this works now?

[bradleybauer]

@eridgd The windows-compat branch works.
I forgot to mention the small edit in vgg_normalized.py :)