I have textual data of various lengths for which ragged tensors seems well suited. For instance my data could look as follows :

x = tf.ragged.constant([[1,2,3,4,5,6,7], [5,6,1,2]])

I would like provide this ragged tensor to a model composed by some convolutional filters, let's say at least one filter as follows :

model = Sequential([Embedding(alphabet_size, embedding_dim),
                    Conv1D(filters=10, kernel_size=10),

I tried to use tf.ragged.map_flat_values however I am not sure that it does what I would like, i.e : embedding each text line of the batch, convolving it, and then taking the max over each text line.

Is there a workaround to make this model work on (very) variable lengths texts (except of course using 3d tensors batches)?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.