dipanjans / explainable_artificial_intelligence

Slides, code and resources for model interpretation methods in machine learning and deep learning

explainable_artificial_intelligence

Slides, code and resources for model interpretation methods in machine learning and deep learning

explainable_artificial_intelligence's Issues

Getting an error with Interpreting_Vision_Models_1; using SHAP to explain how 7th layer of VGG16 Model impacts prediction

I've been trying to replicate this code and I have been stuck on this particular issue. When I try to run shap.GradientExplainer() for the 7th layer, I keep getting the same error:

1 def map2layer(x, layer):
----> 2     feed_dict = dict(zip([model.layers[7].input], [preprocess_input(x.copy())]))
      3     print(feed_dict.keys().type)
      4     return K.get_session().run(model.layers[layer].input, feed_dict)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/keras_tensor.py in __hash__(self)
    259   def __hash__(self):
    260     raise TypeError('Tensors are unhashable. (%s)'
--> 261                     'Instead, use tensor.ref() as the key.' % self)
    262 
    263   # Note: This enables the KerasTensor's overloaded "right" binary

TypeError: Tensors are unhashable. (KerasTensor(type_spec=TensorSpec(shape=(None, 56, 56, 128), dtype=tf.float32, name=None), name='block2_pool/MaxPool:0', description="created by layer 'block2_pool'"))Instead, use tensor.ref() as the key.

I have tried downgrading my tensorflow version to below 2.0 but it still hasn't worked. Would love to hear a fix for this.

explainer = shap.DeepExplainer(dnn_model, data=X_train)

Getting error message from here.

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/input_spec.py in assert_input_compatibility(input_spec, inputs, layer_name)
    230                          ', found ndim=' + str(ndim))
    231     if spec.min_ndim is not None:
--> 232       ndim = x.shape.rank
    233       if ndim is not None and ndim < spec.min_ndim:
    234         raise ValueError('Input ' + str(input_index) + ' of layer ' +

AttributeError: 'tuple' object has no attribute 'rank'