ValueError: Tried to convert 'x' to a tensor and failed. Error: None values not supported. about keras-grad-cam HOT 9 OPEN

I made it work following piece of code from : https://github.com/tensorflow/tensorflow/issues/783

def _compute_gradients(tensor, var_list):
grads = tf.gradients(tensor, var_list)
return [grad if grad is not None else tf.zeros_like(var)
for var, grad in zip(var_list, grads)]

and transforming from
grads = normalize(K.gradients(loss, conv_output)[0])
to
grads = normalize(_compute_gradients(loss, [conv_output])[0])

from keras-grad-cam.

experiencing the same issue. Found this as well:

tensorflow/tensorflow#783

from keras-grad-cam.

For everyone who is facing the same issue: I forked this repository and fixed all errors I was facing.

#27
https://github.com/PowerOfCreation/keras-grad-cam

from keras-grad-cam.

I am also getting the same error, can this issue be fixed?

from keras-grad-cam.

osimeoni's answer worked out for me

from keras-grad-cam.

I get another error when i use：
def _compute_gradients(tensor, var_list):
def _compute_gradients(tensor, var_list):
grads = tf.gradients(tensor, var_list)
&nbsp return [grad if grad is not None else tf.zeros_like(var) for var, grad in zip(var_list, grads)]
the error is:
zip argument #1 must support iteration

from keras-grad-cam.

I finally find out where is key problem：
K.gradient return None, because it counter an op that can't gradient
I doubt that the Sequential have used to encapsulate vgg16 , that is a Model type
here is code in grad_cam funciton after alter:

    nb_classes = 1000
    target_layer = lambda x: target_category_loss(x, category_index, nb_classes)
    last = Lambda(target_layer, output_shape=target_category_loss_output_shape)(input_model.output)
    model = Model(inputs=input_model.input, outputs=last)
    loss = K.sum(model.layers[-1].output)  
    # loss = model.layers[-1].output              
    conv_output = [l for l in model.layers if l.name is layer_name][0].output    
    grads = normalize(K.gradients(loss, conv_output)[0])
    # grads = normalize(_compute_gradients(loss, conv_output)[0])    
    gradient_function = K.function([model.inputs[0]], [conv_output, grads])   

from keras-grad-cam.

@PowerOfCreation Thanks, it really helps. It works!

from keras-grad-cam.

For anyone who still experience this error with tensorflow 2.x. The default argument at tf.gradients (unconnected_gradients='none') causes this error. Set it to ='zero' could solve your problem.

from keras-grad-cam.