quairkit.application.comb.comb¶

The source file of the PQCombNet class.

class quairkit.application.comb.comb.PQCombNet(target_function, num_slots=1, ancilla=0, slot_dim=2, train_unitary_info=1000, test_unitary_info=0, train_mode='process', task_name=None, is_ctrl_U=False, seed=None)¶

Parameterized Quantum Comb Net.

Parameters:¶

target_function : Callable[[Tensor], Tensor]¶: The function to apply to each unitary in the dataset.
num_slots : int¶: The number of unitaries to be queried.
ancilla : List[int] | int¶: The ancilla dimension or dimension list.
slot_dim : int¶: The slot dimension for the unitaries to be queried.
train_unitary_info : int | Tensor¶: The number of unitaries or the unitary dataset to be used for training or the training unitary set.
test_unitary_info : int | Tensor¶: The number of unitaries or the unitary dataset to be used for testing or the testing unitary set.
train_mode : str¶: The training mode, which can be “process”, “comb”, or “swap”, default is “process”.
task_name : str¶: Optional name for the task, useful for data logging and storage.
is_ctrl_U : bool¶: Flag to indicate if a controlled-U operation is used in the training process.
seed : int | None¶: Optional seed for random number generation, enhancing reproducibility.

update_V_circuit(index, new_V)¶

Update the V circuit at the specified index with a new circuit.

Parameters:¶

index : int¶: The index of the V circuit to update.
new_V : Circuit | ParamOracle | Tensor | Tuple[Tensor, List[int]]¶: The new V circuit, which can be a ParamOracle, Circuit, torch.Tensor or Tuple[torch.Tensor, List[int]].

Raises:¶

ValueError – If the index is out of range or if the dimension of the provided Circuit does not match the dimension of the existing Circuit.
TypeError – If the new_V is not a Circuit, ParamOracle, torch.Tensor, or Tuple[torch.Tensor, List[int]].

train(projector=None, base_lr=0.1, max_epochs=10000, is_save_data=False, is_auto_stop=True)¶

Train the PQCombNet model.

Parameters:¶

projector : Tensor | None¶: The projector to apply to the ancilla system of the output state.
base_lr : float¶: The base learning rate for the optimizer.
max_epochs : int¶: The maximum number of epochs to train for.
is_save_data : bool¶: A flag to indicate whether to save the training data.
is_auto_stop : bool¶: A flag to indicate whether to stop training early if the learning rate is too low.

extract_highest_fidelity()¶

Call the _extract_highest_fidelity function to generate the fidelity tables. If the file does not exist, prompt the user to set is_save_data to True.