TF2 : AttributeError: Can't set the attribute "trainable_weights", likely because it conflicts with an existing read-only @property of the object. Please choose a different name.

Implementation on TF2 gives me the error : AttributeError: Can't set the attribute "trainable_weights", likely because it conflicts with an existing read-only @property of the object. Please choose a different name. at the ssd_300 or ssd_500 model creation. –> model = ssd_300(...) with nothing changed to the parameters.

I have changed all the keras... to tensorflow.keras..., and the keras.engine... ones to tensorflow.keras.layers... tf.__version__ : 2.0.0 from scipy.misc import imread changed to from matplotlib.pyplot import imread

Anyone with a solution for this?

If more info is required, I would be glad to provide.

(I’m sorry for very brief explanation, I don’t have much time on my hands at this moment)

[faisalshahbaz]

I actually found the solution, You need to set self._trainable_weights instead of self.trainable_weights

[jessebr2]

Have the same problem with TimeSeriesAttention

def build(self, input_shape):
    dimensions = input_shape[2]
    timed = keras.models.Sequential(name='per_time_step')
    timed.add(
        keras.layers.Dense(dimensions, input_shape=(dimensions,), kernel_initializer='zeros', activation='softmax', name='att_dense1')
    )
    timed.add(keras.layers.Dense(dimensions, kernel_initializer='glorot_normal', activation='tanh', name='att_dense2'))
    self.timed = keras.layers.TimeDistributed(timed, name='att_td')
    self.timed.build(input_shape)
    self.trainable_weights = self.timed.trainable_weights
    self.non_trainable_weights = self.timed.non_trainable_weights
    self.built = True

Change the:

self.trainable_weights = self.timed.trainable_weights

For:

self.trainable_weights.append(self.timed.trainable_weights)