Add FLAML experiment runner.

experiment.sh

@@ -15,7 +15,15 @@ run_experiment() {
     automl="ag"
   fi
 
-  source env/bin/activate
+  if [[ "$*" == *"flaml"* ]]; then
+    automl="flaml"
+  fi
+
+  if [[ "$automl" == "imba" ]]; then
+    source env/bin/activate
+  else
+    source devenv/bin/activate
+  fi
 
   "$VIRTUAL_ENV"/bin/python -m experiment.main --automl="$automl" --out="$out" --preset="$preset"
 }

experiment/flaml_automl.py

@@ -0,0 +1,29 @@
+from typing import Union
+
+import numpy as np
+import pandas as pd
+from sklearn.exceptions import NotFittedError
+from sklearn.metrics import f1_score
+
+from experiment.runner import AutoMLRunner
+from flaml import AutoML
+
+
+class FLAMLExperimentRunner(AutoMLRunner):
+    def fit(self, X_train: Union[np.ndarray, pd.DataFrame], y_train: Union[np.ndarray, pd.Series], target_label: str,
+            dataset_name: str):
+        flaml = AutoML()
+        flaml.fit(X_train, y_train, task='classification', time_budget=-1, metric='f1')
+
+        best_loss = flaml.best_loss
+        best_model = flaml.best_estimator
+        self._log_val_loss_alongside_model_class({best_model: best_loss})
+
+        self._fitted_model = flaml
+
+    def predict(self, X_test: Union[np.ndarray, pd.DataFrame]) -> np.ndarray:
+        if self._fitted_model is None:
+            raise NotFittedError()
+
+        predictions = self._fitted_model.predict(X_test)
+        return predictions

experiment/main.py

@@ -67,6 +67,10 @@ def main():
             from experiment.imba import ImbaExperimentRunner
 
             automl_runner = ImbaExperimentRunner()
+        elif automl == 'flaml':
+            from experiment.flaml_automl import FLAMLExperimentRunner
+
+            automl_runner = FLAMLExperimentRunner()
         else:
             raise ValueError("Invalid --automl option. Options available: ['imba', 'ag'].")
 

experiment/runner.py

@@ -118,6 +118,8 @@ def run(self, n_evals: Optional[int] = None):
 
             iterator_of_class_belongings = iter(sorted(class_belongings))
             *_, positive_class_label = iterator_of_class_belongings
+            logger.info(f"Pos class label: {positive_class_label}")
+
             number_of_positives = class_belongings.get(positive_class_label, None)
 
             if number_of_positives is None: