I would like to adapt the BST model "BEHAVIOR SEQUENCE TRANSFORMER" so that it accepts variable-length sequences as input, I tried to use RaggedTensor, but the problem is how create the Input layer of movie_id and rating sequence
"sequence_movie_ids": layers.Input(
name="sequence_movie_ids", shape=(sequence_length - 1,), dtype=tf.string
),
"target_movie_id": layers.Input(
name="target_movie_id", shape=(1,), dtype=tf.string
),
"sequence_ratings": layers.Input(
name="sequence_ratings", shape=(sequence_length - 1,), dtype=tf.float32
),
"sequence_movie_ids": layers.Input( name="sequence_movie_ids", shape=(none), dtype=tf.string ), "target_movie_id": layers.Input( name="target_movie_id", shape=(1,), dtype=tf.string ), "sequence_ratings": layers.Input( name="sequence_ratings", shape=(none,), dtype=tf.float32 ),