Hi Jacques
Just an update fyi… After working on some priorities, now back to looking at this. Can’t find a way of passing something to the Tensorflow Serving pre-built container to contain the RewriterConfig, so didn’t try that. Refactoring the code slightly, we can get slightly different error messages, but the issue of the while node argument mismatch remains. So now, we’re refactoring to remove the DenseHashTable and do a vectorised tf.equal of the rails across the list that we created the DenseHashTable from. It will be faster, but more memory…