Inspecting circuits

A qjit()-compiled program does not run in Python, so Python’s print statements and debuggers only run once, while the program is compiled. Instead, PennyLane provides tools that run as part of the compiled program, as well as access to the compiled program itself.

Printing values at runtime

The catalyst.debug.print() function prints values while the compiled program runs, including the outcomes of mid-circuit measurements:

from catalyst import debug

dev = qp.device("lightning.qubit", wires=2)

@qp.qjit
@qp.qnode(dev)
def circuit(x):
    debug.print("x = {x}", x=x)
    qp.RX(x, wires=0)
    m = qp.measure(0)
    debug.print("measurement outcome: {m}", m=m)
    qp.cond(m, qp.X)(wires=1)
    return qp.expval(qp.Z(1))
>>> circuit(np.pi)
x = 3.141592653589793
measurement outcome: 1
Array(-1., dtype=float64)

Running Python callbacks

The catalyst.debug.callback() decorator turns a Python function into one that is called with runtime values while the compiled program runs. This can be used for logging or checking intermediate values:

@debug.callback
def check(x):
    print(f"Received x = {x:.2f}")

@qp.qjit
@qp.qnode(dev)
def circuit(x):
    check(x)
    qp.RX(x, wires=0)
    return qp.expval(qp.Z(0))
>>> circuit(0.5)
Received x = 0.50
Array(0.87758256, dtype=float64)

Inspecting the compiled program

The intermediate representation of a compiled program is available as text. The mlir property contains the program before any compilation pass, and the mlir_opt property contains the program after all compilation passes:

print(circuit.mlir)
print(circuit.mlir_opt)

Additional resources

For more details on debugging compiled programs, see the Catalyst sharp bits and debugging tips. For more details on compilation, see compiling workflows.