Tensorflow 渴望执行和梯度
我试图想象CNN的激活，但面临下一个错误： RuntimeError:启用急切执行时不支持tf.gradients。改用tf.GradientTape 我怎样才能解决它Tensorflow 渴望执行和梯度,tensorflow,conv-neural-network,gradient,Tensorflow,Conv Neural Network,Gradient,我试图想象CNN的激活，但面临下一个错误： RuntimeError:启用急切执行时不支持tf.gradients。改用tf.GradientTape 我怎样才能解决它 layer_dict = dict([(layer.name,layer) for layer in baseline_model.layers[1:]]) img_input = baseline_model.input neuron_index = 1 layer_output = layer_dict['co[enter
layer_dict = dict([(layer.name,layer) for layer in baseline_model.layers[1:]])
img_input = baseline_model.input
neuron_index = 1
layer_output = layer_dict['co[enter image description here][1]nv2d_171'].output
loss = keras.backend.mean(layer_output[:,neuron_index,:,:])
gradients = keras.backend.gradients(loss,img_input)[0]
gradients = tf.linalg.normalize(gradients)
iterate = keras.function([img_input],[loss,gradents])