♻️ Refactor RL actions by SDK

.gitignore

@@ -53,3 +53,6 @@ venv.bak/
 .mypy_cache/
 
 src/**/_version.py
+
+# RL training checkpoints and logs
+events.out.*

CHANGELOG.md

@@ -11,6 +11,7 @@ This project adheres to [Semantic Versioning], with the exception that minor rel
 
 ### Changed
 
+- ♻️ Restructure existing RL passes into SDK-level action modules ([#680]) ([**@flowerthrower**])
 - 🎨 Improve the RL state machine logic ([#677]) ([**@flowerthrower**])
 - 🐛 Support BQSKit conversion of IQM's native `r` gate ([#679]) ([**@flowerthrower**])
 - 🔧 Replace `mypy` with `ty` ([#572]) ([**@denialhaag**])
@@ -48,6 +49,7 @@ _📚 Refer to the [GitHub Release Notes](https://github.com/munich-quantum-tool
 
 <!-- PR links -->
 
+[#680]: https://github.com/munich-quantum-toolkit/predictor/pull/680
 [#677]: https://github.com/munich-quantum-toolkit/predictor/pull/677
 [#679]: https://github.com/munich-quantum-toolkit/predictor/pull/679
 [#572]: https://github.com/munich-quantum-toolkit/predictor/pull/572

src/mqt/predictor/hellinger/utils.py

@@ -39,7 +39,7 @@ def calc_device_specific_features(
 ) -> NDArray[np.float64]:
     """Creates and returns a device-specific feature vector for a given quantum circuit and device.
 
-    Arguments:
+    Args:
         qc: The quantum circuit for which the features are calculated.
         device: The device for which the features are calculated.
         ignore_gates: A list of gates to be ignored when calculating the features. Defaults to ["barrier", "id", "measure", "if_else"].

src/mqt/predictor/ml/helper.py

@@ -112,7 +112,7 @@ def dict_to_featurevector(gate_dict: dict[str, int]) -> dict[str, int]:
 def create_feature_vector(qc: QuantumCircuit) -> list[int | float]:
     """Creates and returns a feature dictionary for a given quantum circuit.
 
-    Arguments:
+    Args:
         qc: The quantum circuit to be compiled.
 
     Returns:

src/mqt/predictor/ml/predictor.py

@@ -66,7 +66,7 @@ def setup_device_predictor(
 ) -> bool:
     """Sets up the device predictor for the given figure of merit.
 
-    Arguments:
+    Args:
         devices: The devices to be used for training.
         figure_of_merit: The figure of merit to be used for training. Defaults to "expected_fidelity".
         path_uncompiled_circuits: The path to the directory containing the circuits to be compiled. Defaults to None.
@@ -127,7 +127,7 @@ def __init__(
     ) -> None:
         """Initializes the Predictor class.
 
-        Arguments:
+        Args:
             figure_of_merit: The figure of merit to be used for training.
             devices: The devices to be used for training.
             logger_level: The level of the logger. Defaults to logging.INFO.
@@ -151,7 +151,7 @@ def _compile_all_circuits_devicewise(
     ) -> None:
         """Compiles all circuits in the given directory with the given timeout and saves them in the given directory.
 
-        Arguments:
+        Args:
             device: The device to be used for compilation.
             timeout: The timeout in seconds for the compilation of a single circuit.
             path_uncompiled_circuits: The path to the directory containing the circuits to be compiled. Defaults to None.
@@ -204,7 +204,7 @@ def compile_training_circuits(
     ) -> None:
         """Compiles all circuits in the given directory with the given timeout and saves them in the given directory.
 
-        Arguments:
+        Args:
             path_uncompiled_circuits: The path to the directory containing the circuits to be compiled. Defaults to None.
             path_compiled_circuits: The path to the directory where the compiled circuits should be saved. Defaults to None.
             timeout: The timeout in seconds for the compilation of a single circuit. Defaults to 600.
@@ -237,7 +237,7 @@ def generate_training_data(
     ) -> None:
         """Creates and saves training data from all generated training samples.
 
-        Arguments:
+        Args:
             path_uncompiled_circuits: The path to the directory containing the uncompiled circuits. Defaults to None.
             path_compiled_circuits: The path to the directory containing the compiled circuits. Defaults to None.
             path_training_data: The path to the directory where the generated training data should be saved. Defaults to None.
@@ -295,7 +295,7 @@ def _generate_training_sample(
     ) -> tuple[tuple[list[Any], Any], str, list[float]]:
         """Handles to create a training sample from a given file.
 
-        Arguments:
+        Args:
             file: The name of the file to be used for training.
             path_uncompiled_circuit: The path to the directory containing the uncompiled circuits. Defaults to None.
             path_compiled_circuits: The path to the directory containing the compiled circuits. Defaults to None.
@@ -365,7 +365,7 @@ def train_random_forest_model(
     ) -> RandomForestRegressor | RandomForestClassifier:
         """Trains a random forest model for the given figure of merit.
 
-        Arguments:
+        Args:
             training_data: The training data to be used for training the model. If None, the training data is loaded from the pre-prepared training data files.
 
         Returns:
@@ -453,7 +453,7 @@ def predict_device_for_figure_of_merit(
 ) -> Target:
     """Returns the probabilities for all supported quantum devices to be the most suitable one for the given quantum circuit.
 
-    Arguments:
+    Args:
         qc: The QuantumCircuit or Path to the respective qasm file.
         figure_of_merit: The figure of merit to be used for compilation.
 

src/mqt/predictor/qcompile.py

@@ -27,7 +27,7 @@ def qcompile(
 ) -> tuple[QuantumCircuit, list[str], str]:
     """Compiles a given quantum circuit to a device with the highest predicted figure of merit.
 
-    Arguments:
+    Args:
         qc: The quantum circuit to be compiled.
         figure_of_merit: The figure of merit to be used for compilation. Defaults to "expected_fidelity".
 

src/mqt/predictor/reward.py

@@ -46,7 +46,7 @@ def crit_depth(qc: QuantumCircuit, precision: int = 10) -> float:
 def expected_fidelity(qc: QuantumCircuit, device: Target, precision: int = 10) -> float:
     """Calculates the expected fidelity of a given quantum circuit on a given device.
 
-    Arguments:
+    Args:
         qc: The quantum circuit to be compiled.
         device: The device to be used for compilation.
         precision: The precision of the returned value. Defaults to 10.
@@ -83,7 +83,7 @@ def estimated_success_probability(qc: QuantumCircuit, device: Target, precision:
     It is calculated by multiplying the expected fidelity with a min(T1,T2)-dependent
     decay factor during qubit idle times. To this end, the circuit is scheduled using ASAP scheduling.
 
-    Arguments:
+    Args:
         qc: The quantum circuit to be compiled.
         device: The device to be used for compilation.
         precision: The precision of the returned value. Defaults to 10.
@@ -210,7 +210,7 @@ def estimated_success_probability(qc: QuantumCircuit, device: Target, precision:
 def esp_data_available(device: Target) -> bool:
     """Check if calibration data to calculate ESP is available for the device.
 
-    Arguments:
+    Args:
         device: The device to be checked for calibration data.
 
     Returns:
@@ -310,7 +310,7 @@ def estimated_hellinger_distance(
 ) -> float:
     """Calculates the estimated Hellinger distance of a given quantum circuit on a given device.
 
-    Arguments:
+    Args:
         qc: The quantum circuit to be compiled.
         device: The device to be used for compilation.
         model: The pre-trained model to use for prediction (optional). If not provided, the model will try to be loaded from files.