Hi, I'm using keras-vis in a jupyter notebook. Google Colab has migrated to Tensorflow 2, which provokes compatibility problems when I try to run the cells that use this library (in my case, it detects as None the penultimate layer of the model when I use the visualize_cam function)
Unable to determine penultimateConvorPoolinglayer for layer_idx: -1
I think this problem may be solved by changing the imports of keras to tensorflow.keras
from tensorflow.keras.layers.convolutional import _Conv from tensorflow.keras.layers.pooling import _Pooling1D, _Pooling2D, _Pooling3D from tensorflow.keras.layers.wrappers import Wrapper from tensorflow.keras import backend as K
I'm not sure if this could harm the rest of the code, but it is the only way to maintain this library running in Google Colab.
Hi, I'm using keras-vis in a jupyter notebook. Google Colab has migrated to Tensorflow 2, which provokes compatibility problems when I try to run the cells that use this library (in my case, it detects as
Nonethe penultimate layer of the model when I use the visualize_cam function)Unable to determine penultimateConvorPoolinglayer for layer_idx: -1I think this problem may be solved by changing the imports of keras to tensorflow.keras
from tensorflow.keras.layers.convolutional import _Conv from tensorflow.keras.layers.pooling import _Pooling1D, _Pooling2D, _Pooling3D from tensorflow.keras.layers.wrappers import Wrapper from tensorflow.keras import backend as KI'm not sure if this could harm the rest of the code, but it is the only way to maintain this library running in Google Colab.