Error from_tk

[JVM1982]

Hello.

Why the following code does not work ? :

sentence = 'person runs program .'
diagram = remove_cups( parser.sentence2diagram( sentence ) )
circuit = ansatz( diagram )
print( model( [ circuit ] ) ) # OK #
print( model( [ from_tk( circuit.to_tk() ) ] ) ) # ERROR #

[[0.14473685 0.85526315]]
Unexpected exception formatting exception. Falling back to standard exception

Traceback (most recent call last):
File "/home/javier.valera/.local/lib/python3.8/site-packages/IPython/core/interactiveshell.py", line 3378, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "/tmp/ipykernel_16444/2916265434.py", line 9, in
print( model( [ from_tk( circuit.to_tk() ) ] ) )
File "/home/javier.valera/.local/lib/python3.8/site-packages/lambeq/training/model.py", line 59, in call
return self.forward(*args, **kwds)
File "/home/javier.valera/.local/lib/python3.8/site-packages/lambeq/training/tket_model.py", line 131, in forward
return self.get_diagram_output(x)
File "/home/javier.valera/.local/lib/python3.8/site-packages/lambeq/training/tket_model.py", line 101, in get_diagram_output
*[diag_f(*self.weights) for diag_f in lambdified_diagrams],
File "/home/javier.valera/.local/lib/python3.8/site-packages/lambeq/training/tket_model.py", line 101, in
*[diag_f(*self.weights) for diag_f in lambdified_diagrams],
File "/home/javier.valera/.local/lib/python3.8/site-packages/discopy/monoidal.py", line 509, in
return lambda xs: self.id(self.dom).then((
File "/home/javier.valera/.local/lib/python3.8/site-packages/discopy/monoidal.py", line 510, in
self.id(left) @ box.lambdify(*symbols, **kwargs)(*xs)
File "/home/javier.valera/.local/lib/python3.8/site-packages/discopy/quantum/gates.py", line 321, in
return lambda *xs: type(self)(c_fn(*xs), distance=self.distance)
File "/home/javier.valera/.local/lib/python3.8/site-packages/discopy/quantum/gates.py", line 448, in
return lambda *xs: type(self)(data(*xs))
File "", line 2, in _lambdifygenerated
return runs__n.r@s@n.l_0
NameError: name 'runs__n' is not defined

Thanks.

[Thommy257]

Hi there,

thanks for raising this issue. The problem here is that a circuit parameter in lambeq is of type <class 'lambeq.ansatz.base.Symbol'> whereas discopy uses <class 'sympy.core.symbol.Symbol'>. Hence, lambeq models that was previously intialised with lambeq circuits won't recognise a native discopy circuit.

This is actually a bug we need to fix. I'll think about it!

[Thommy257]

This is now solved by the 0.2.8 release.

run

pip install -- upgrade lambeq