qml.qjit

qjit(fn=None, *args, compiler='catalyst', **kwargs)

A decorator for just-in-time compilation of hybrid quantum programs in PennyLane.

This decorator enables both just-in-time and ahead-of-time compilation, depending on the compiler package and whether function argument type hints are provided.

Note

Currently, only two compilers are supported; the Catalyst hybrid quantum-classical compiler, which works with the JAX interface, and CUDA Quantum.

For more details on Catalyst, see the Catalyst documentation and catalyst.qjit().

Note

Catalyst only supports the JAX interface and selected devices. Supported backend devices for Catalyst include lightning.qubit, lightning.kokkos, lightning.gpu, and braket.aws.qubit, but not default.qubit.

For a full list of supported devices, please see Supported devices.

CUDA Quantum supports softwareq.qpp, nvidia.custatevec, and nvidia.cutensornet.

Parameters

• fn (Callable) – Hybrid (quantum-classical) function to compile

• compiler (str) – Name of the compiler to use for just-in-time compilation. Available options include catalyst and cuda_quantum.

• autograph (bool) – Experimental support for automatically converting Python control flow statements to Catalyst-compatible control flow. Currently supports Python if, elif, else, and for statements. Note that this feature requires an available TensorFlow installation. See the AutoGraph guide for more information.

• keep_intermediate (bool) – Whether or not to store the intermediate files throughout the compilation. The files are stored at the location where the Python script is called. If True, intermediate representations are available via the mlir, jaxpr, and qir, representing different stages in the optimization process.

• verbosity (bool) – If True, the tools and flags used by Catalyst behind the scenes are printed out.

• logfile (TextIOWrapper) – File object to write verbose messages to (default is sys.stderr)

• pipelines (List[Tuple[str, List[str]]]) – A list of pipelines to be executed. The elements of this list are named sequences of MLIR passes to be executed. A None value (the default) results in the execution of the default pipeline. This option is considered to be used by advanced users for low-level debugging purposes.

• static_argnums (int or Sequence[Int]) – an index or a sequence of indices that specifies the positions of static arguments.

• abstracted_axes (Sequence[Sequence[str]] or Dict[int, str] or Sequence[Dict[int, str]]) – An experimental option to specify dynamic tensor shapes. This option affects the compilation of the annotated function. Function arguments with abstracted_axes specified will be compiled to ranked tensors with dynamic shapes. For more details, please see the Dynamically-shaped Arrays section below.

Returns

A class that, when executed, just-in-time compiles and executes the decorated function

Return type

catalyst.QJIT

Raises

• FileExistsError – Unable to create temporary directory

• PermissionError – Problems creating temporary directory

• OSError – Problems while creating folder for intermediate files

• AutoGraphError – Raised if there was an issue converting the given the function(s).

• ImportError – Raised if AutoGraph is turned on and TensorFlow could not be found.

Example

In just-in-time (JIT) mode, the compilation is triggered at the call site the first time the quantum function is executed. For example, circuit is compiled as early as the first call.

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

@qml.qjit
@qml.qnode(dev)
def circuit(theta):
    qml.Hadamard(wires=0)
    qml.RX(theta, wires=1)
    qml.CNOT(wires=[0,1])
    return qml.expval(qml.Z(1))

>>> circuit(0.5)  # the first call, compilation occurs here
array(0.)
>>> circuit(0.5)  # the precompiled quantum function is called
array(0.)

qjit() compiled programs also support nested container types as inputs and outputs of compiled functions. This includes lists and dictionaries, as well as any data structure implementing the JAX PyTree.

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

@qml.qjit
@qml.qnode(dev)
def f(x):
    qml.RX(x["rx_param"], wires=0)
    qml.RY(x["ry_param"], wires=0)
    qml.CNOT(wires=[0, 1])
    return {
        "XY": qml.expval(qml.X(0) @ qml.Y(1)),
        "X": qml.expval(qml.X(0)),
    }

>>> x = {"rx_param": 0.5, "ry_param": 0.54}
>>> f(x)
{'X': array(-0.75271018), 'XY': array(1.)}

For more details on using the qjit() decorator and Catalyst with PennyLane, please refer to the Catalyst quickstart guide, as well as the sharp bits and debugging tips page for an overview of the differences between Catalyst and PennyLane, and how to best structure your workflows to improve performance when using Catalyst.

Static arguments

static_argnums defines which elements should be treated as static. If it takes an integer, it means the argument whose index is equal to the integer is static. If it takes an iterable of integers, arguments whose index is contained in the iterable are static. Changing static arguments will trigger re-compilation.

A valid static argument must be hashable and its __hash__ method must be able to reflect any changes of its attributes.

@dataclass
class MyClass:
    val: int

    def __hash__(self):
        return hash(str(self))

@qjit(static_argnums=1)
def f(
    x: int,
    y: MyClass,
):
    return x + y.val

f(1, MyClass(5))
f(1, MyClass(6)) # re-compilation
f(2, MyClass(5)) # no re-compilation

In the example above, y is static. Note that the second function call triggers re-compilation since the input object is different from the previous one. However, the third function call directly uses the previous compiled one and does not introduce re-compilation.

@dataclass
class MyClass:
    val: int

    def __hash__(self):
        return hash(str(self))

@qjit(static_argnums=(1, 2))
def f(
    x: int,
    y: MyClass,
    z: MyClass,
):
    return x + y.val + z.val

my_obj_1 = MyClass(5)
my_obj_2 = MyClass(6)
f(1, my_obj_1, my_obj_2)
my_obj_1.val = 7
f(1, my_obj_1, my_obj_2) # re-compilation

In the example above, y and z are static. The second function will cause function f to re-compile because my_obj_1 is changed. This requires that the mutation is properly reflected in the hash value.

Note that when static_argnums is used in conjunction with type hinting, ahead-of-time compilation will not be possible since the static argument values are not yet available. Instead, compilation will be just-in-time.

Dynamically-shaped arrays

There are three ways to use abstracted_axes; by passing a sequence of tuples, a dictionary, or a sequence of dictionaries. Passing a sequence of tuples:

abstracted_axes=((), ('n',), ('m', 'n'))

Each tuple in the sequence corresponds to one of the arguments in the annotated function. Empty tuples can be used and correspond to parameters with statically known shapes. Non-empty tuples correspond to parameters with dynamically known shapes.

In this example above,

• the first argument will have a statically known shape,

• the second argument will have dynamic shape n for the zeroth axis, and

• the third argument will have dynamic shape m for its zeroth axis and dynamic shape n for its first axis.

Passing a dictionary:

abstracted_axes={0: 'n'}

This approach allows a concise expression of the relationships between axes for different function arguments. In this example, it specifies that for all function arguments, the zeroth axis will have dynamic shape n.

Passing a sequence of dictionaries:

abstracted_axes=({}, {0: 'n'}, {1: 'm', 0: 'n'})

The example here is a more verbose version of the tuple example. This convention allows axes to be omitted from the list of abstracted axes.

Using abstracted_axes can help avoid the cost of recompilation. By using abstracted_axes, a more general version of the compiled function will be generated. This more general version is parametrized over the abstracted axes and allows results to be computed over tensors independently of their axes lengths.

For example:

@qjit
def sum(arr):
    return jnp.sum(arr)

sum(jnp.array([1]))     # Compilation happens here.
sum(jnp.array([1, 1]))  # And here!

The sum function would recompile each time an array of different size is passed as an argument.

@qjit(abstracted_axes={0: "n"})
def sum_abstracted(arr):
    return jnp.sum(arr)

sum(jnp.array([1]))     # Compilation happens here.
sum(jnp.array([1, 1]))  # No need to recompile.

The sum_abstracted function would only compile once and its definition would be reused for subsequent function calls.

code/api/pennylane.qjit

 

Download Python script

 

Download Notebook

 

View on GitHub