features: The output of the featurizer_final state.
sequence_features: The output of the featurizer at each timestep.
initial_shape = [a or -1 for a in X.get_shape().as_list()]
X = tf.reshape(X, shape=[-1] + initial_shape[-2:])
with tf.variable_scope("model/featurizer", reuse=reuse):
After Change
features: The output of the featurizer_final state.
sequence_features: The output of the featurizer at each timestep.
initial_shape = tf.shape(X)
X = tf.reshape(X, shape=[-1] + initial_shape[-2:])
X = tf.reshape(X, shape=tf.concat(([-1], initial_shape[-2:]), 0))
with tf.variable_scope("model/featurizer", reuse=reuse):
embed_weights = tf.get_variable(
name="we",