diff --git a/src/facebias/__init__.py b/src/facebias/__init__.py
index 2a94dfd..2159cd1 100644
--- a/src/facebias/__init__.py
+++ b/src/facebias/__init__.py
@@ -10,10 +10,7 @@ from typing import Any, Callable, Optional
 import cv2
 import numpy as np
 
-logging.basicConfig(
-    level=logging.DEBUG,  # Set the minimum logging level
-)
-
+logging.basicConfig(level=logging.DEBUG)
 logger = logging.getLogger(__name__)
 
 
@@ -83,15 +80,6 @@ def load_dataset(
             im = cv2.cvtColor(cv2.imread(p), cv2.COLOR_BGR2RGB)
         except cv2.error:
             logger.info(f'File "{p}" is not an image. Skipping.')
-            # if meta_path is None:
-            #     logger.info(f'Trying to read "{p}" as metadata.')
-            #     if p.suffix == ".csv":
-            #         metadata = load_metadata(p)
-            #         logger.info("Metadata read successfully.")
-            #     else:
-            #         logger.error(f'Failed to read "{p}" as metadata. Skipping.')
-            # elif not metadata:
-            #     logger.critical("Logic error: Metadata should have been read already.")
         else:
             proc_imname = imname_proc_fn(p.name) if imname_proc_fn is not None else p.name
             ims[proc_imname] = im
diff --git a/src/facebias/estimators/fairface.py b/src/facebias/estimators/fairface.py
index 687094e..3580c36 100644
--- a/src/facebias/estimators/fairface.py
+++ b/src/facebias/estimators/fairface.py
@@ -14,9 +14,31 @@ from torchvision import transforms
 
 
 class FairFace(BaseEstimator):
+    """FairFace trained model for age-group, sex and ethnicity estimation.
+
+    Note that the ethnicity estimation is disabled, since each model estimates
+    a completely different set of classes. Additionally, there is an argument
+    that ethnicity is extremelly difficult to estimate given a single picture
+    of a person, thus it is better to estimate skin color, which this model
+    does not do.
+
+    Parameters
+    ----------
+    checkpoint_path: Optional[Path]
+        Path to the model weights stored as a PyTorch file.
+
+    device: Optional[torch.device]
+        The device to load the model into. By default is `"cpu"`
+
+    Reference
+    ---------
+    FairFace repository: https://github.com/dchen236/FairFace
     """
-    """
-    def __init__(self, checkpoint_path: Optional[Path]=None, device=torch.device("cpu")):
+    def __init__(
+            self,
+            checkpoint_path: Optional[Path]=None,
+            device: Optional[torch.device]=torch.device("cpu")
+    ):
         self.T = transforms.Compose([
             transforms.ToPILImage(),
             transforms.Resize((224, 224)),
@@ -30,12 +52,12 @@ class FairFace(BaseEstimator):
         self.model = self.model.eval().to(device)
         if checkpoint_path is not None:
             self.load_state_dict(checkpoint_path)
-            self.model.to(device)
 
     def load_state_dict(self, checkpoint_path: Path) -> "FairFace":
         self.model.load_state_dict(
             torch.load(checkpoint_path, map_location=torch.device("cpu"))
         )
+        self.model.to(self.device)
         return self
 
     def predict(self, ims: dict[Path, np.ndarray]) -> dict[Path, dict[Capability, Any]]:
diff --git a/src/facebias/estimators/mivolov1.py b/src/facebias/estimators/mivolov1.py
index eebc36f..6baf019 100644
--- a/src/facebias/estimators/mivolov1.py
+++ b/src/facebias/estimators/mivolov1.py
@@ -26,6 +26,10 @@ class MiVOLOv1(BaseEstimator):
 
     device: Optional[torch.device]
         Device to load the checkpoints into. By default is "cpu".
+
+    Reference
+    ---------
+    MiVOLO repository: https://github.com/wildchlamydia/mivolo
     """
     def __init__(
             self,