I'm trying to use innvestigate 1.8 to analyze a transfer-learning model I built using resnet50.
I run one of the examples in the distribution:
python train_pattern_imagenet.py resnet50 >& TRA-OUT-ResNet50
and got the following message:
File "/home/dripoll/miniconda3/envs/INNV/lib/python3.6/site-packages/innvestigate-1.0.8-py3.6.egg/innvestigate/tools/pattern.py", line 367, in init
% layer)
Exception: Model contains not supported layer: <keras.layers.convolutional.ZeroPadding2D object at 0x7f12ae2fa358>
I saw that the ZeroPadding2d layer used in Rensnet50 is missing in the declaration of supported layers (SUPPORTED_LAYER_PATTERNNET) in innvestigate/analyzer/pattern_based.py
Is there any work around? -Thanks
I'm trying to use innvestigate 1.8 to analyze a transfer-learning model I built using resnet50. I run one of the examples in the distribution: python train_pattern_imagenet.py resnet50 >& TRA-OUT-ResNet50 and got the following message: File "/home/dripoll/miniconda3/envs/INNV/lib/python3.6/site-packages/innvestigate-1.0.8-py3.6.egg/innvestigate/tools/pattern.py", line 367, in init % layer) Exception: Model contains not supported layer: <keras.layers.convolutional.ZeroPadding2D object at 0x7f12ae2fa358>
I saw that the ZeroPadding2d layer used in Rensnet50 is missing in the declaration of supported layers (SUPPORTED_LAYER_PATTERNNET) in innvestigate/analyzer/pattern_based.py Is there any work around? -Thanks