KeyError: Ty('p')

[FernandoFS18]

Hi Lambeq support,

I am trying to train a binary classification model using my own dataset, which I built following the same structure as in the example datasets. The problem is that when trying to parameterise the string diagrams I get the error: KeyError: Ty('p').
Could someone tell why I am getting it and how can I fix it?

Thanks a lot for your help

[nikhilkhatri]

Hi @Ferchu18
It would be useful to see the code snippet which raises this error.
My guess is that for your dataset, the parser produces diagrams which contain the PREPOSITIONAL_PHRASE type, denoted by 'p'.
When parameterising a string diagram using an ansatz, it is necessary to provide a map from all the types in the string diagram to the target representation (number of qubits in the quantum case.)

The tutorial docs here demonstrate how to parameterise string diagrams containing only noun and sentence types.
For a dataset containing the preposition phrase type, this can be modified as follows:

# Define atomic types
N = AtomicType.NOUN
S = AtomicType.SENTENCE
P = AtomicType.PREPOSITIONAL_PHRASE

# Convert string diagram to quantum circuit
ansatz = IQPAnsatz({N: 1, S: 1, P: 1}, n_layers=2)

[FernandoFS18]

Hi @nikhilkhatri
My dataset is composed of sentences like the following:

And for the code that raises the error is the following:

I assume that the problem is what you commented about the PREPOSITIONAL_PHRASE type, since I do not specify it for the ansatz like you tell me there.

I will try to implement the changed you wrote and see if it works.
Thanks a lot!