[FIXED] Tensorflow1 DropoutWrapper to Tensorflow2

Issue

How to rewrite this one in tensorflow2? Tried using tf_upgrade_v2 it failed. Tried also with tf.keras.layers.Dropout but had some issues converting MultiRNNCell to tensor

rnn_cells = tf.compat.v1.nn.rnn_cell.MultiRNNCell(
            [lstm_cell(size_layer) for _ in range(num_layers)],
            state_is_tuple = False,
        )
drop = tf.contrib.rnn.DropoutWrapper(rnn_cells, output_keep_prob = forget_bias)

Solution

I replaced:

drop = tf.contrib.rnn.DropoutWrapper(rnn_cells, output_keep_prob = forget_bias)

With:

drop = tf.compat.v1.nn.rnn_cell.DropoutWrapper(rnn_cells, output_keep_prob = forget_bias)

And it worked

Answered By - user16965639