Why did the RNN after picking the learning rate did not use Bidirectional Layer?

Course Q&A TensorFlow Developer Professional Sequences, Time Series and Prediction
1

In Course 4, W4, Lab1, during the model that runs 100 epoches to find the optimal learning rate, the infrastructure includes:

tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(32, return_sequences=True)),
tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(32, return_sequences=True)),

However, in the following model that trains for 500 epoches after picking the optimal running rate only has the regular LSTM layers:
tf.keras.layer

  tf.keras.layers.LSTM(32, return_sequences=True),
  tf.keras.layers.LSTM(32, return_sequences=True),

Is there a reason for this?

2

One of the reasons I can think of is that the model is trained for more no of epochs and hence learns better and doesnt need the bi directional feature
I think the learning rate (optimal one)would be more or less similar for both the cases
See if this answer clears your question…
Thanks