In python I can instantiate a ragged tensor placeholder as below.
inputs_1 = keras.Input(shape=[None,], dtype='int64', ragged=True, name = 'message_A')
inputs_1
tf.RaggedTensor(values=Tensor("Placeholder:0", shape=(None,), dtype=int64), row_splits=Tensor("Placeholder_1:0", shape=(None,), dtype=int64))
Tensorflow responds that the placeholder is ragged with a shape=(None,). But using R-Keras & R-Tensorflow its seems I cannot do the same. I can request an input placeholder and the parameter to make it ragged exists but the response suggests it is not in fact ragged. Additionally, you are forced to declare some shape rather than keep the shape 'flexible' for the ragged tensor.
inputs_1 = tf$keras$Input(shape=c(100L), ragged = TRUE)
inputs_1
Tensor("Placeholder:0", shape=(None, 100), dtype=float32)
Can you show me how to create a ragged tensor placeholder using R-Keras or R-Tensorflow?