[Task] Add RNN-based model apis for session-based recommendation

[rnyak]

Description

Currently there is no out-of-the-box api for creating an LSTM, BiLSTM and GRU architectures for a session-based (or sequential) task.

• For LSTM , I assume we'd use tf.keras.layers.LSTM api.
• For BiLSTM shall we use tf.keras.layers.Bidirectional?
• For GRU, can we not port GRU4Rec model code from TF4Rec to Merlin Models?

Looks like there were two tickets before but they were closed without completion:

https://github.com/NVIDIA-Merlin/models/issues/732 https://github.com/NVIDIA-Merlin/models/issues/733

[zhiruiwang]

Just a wild suggestion, maybe you can also add SAS4Rec by using tf.keras.layers.MultiHeadAttention. I assume it will be a very similar block structure once LSTM and GRU blocks are done

[rnyak]

closing this since we are creating an example for BiLSTM where we can easily use tf.keras.layers.Bidirectional?