diff --git a/src/graphnet/data/dataset/sqlite/sqlite_dataset_perturbed.py b/src/graphnet/data/dataset/sqlite/sqlite_dataset_perturbed.py
index 755d96b82..b951e6916 100644
--- a/src/graphnet/data/dataset/sqlite/sqlite_dataset_perturbed.py
+++ b/src/graphnet/data/dataset/sqlite/sqlite_dataset_perturbed.py
@@ -3,6 +3,7 @@
 from typing import Dict, List, Optional, Tuple, Union
 
 import numpy as np
+from numpy.random import default_rng, Generator
 import torch
 from torch_geometric.data import Data
 
@@ -36,6 +37,7 @@ def __init__(
         loss_weight_table: Optional[str] = None,
         loss_weight_column: Optional[str] = None,
         loss_weight_default_value: Optional[float] = None,
+        seed: Optional[Union[int, Generator]] = None,
     ):
         """Construct SQLiteDatasetPerturbed.
 
@@ -78,6 +80,7 @@ def __init__(
                 in this case to events with no value in the corresponding
                 table/column. That is, if no per-event loss weight table/column
                 is provided, this value is ignored. Defaults to None.
+            seed: Optional seed for random number generation. Defaults to None.
         """
         # Base class constructor
         super().__init__(
@@ -108,6 +111,18 @@ def __init__(
             self._features.index(key) for key in self._perturbation_dict.keys()
         ]
 
+        if seed is not None:
+            if isinstance(seed, int):
+                self.rng = default_rng(seed)
+            elif isinstance(seed, Generator):
+                self.rng = seed
+            else:
+                raise ValueError(
+                    "Invalid seed. Must be an int or a numpy Generator."
+                )
+        else:
+            self.rng = default_rng()
+
     def __getitem__(self, sequential_index: int) -> Data:
         """Return graph `Data` object at `index`."""
         if not (0 <= sequential_index < len(self)):
@@ -127,7 +142,7 @@ def _perturb_features(
         self, features: List[Tuple[float, ...]]
     ) -> List[Tuple[float, ...]]:
         features_array = np.array(features)
-        perturbed_features = np.random.normal(
+        perturbed_features = self.rng.normal(
             loc=features_array[:, self._perturbation_cols],
             scale=np.array(
                 list(self._perturbation_dict.values()), dtype=np.float
diff --git a/src/graphnet/training/utils.py b/src/graphnet/training/utils.py
index 2578ff9a6..f7d5249f9 100644
--- a/src/graphnet/training/utils.py
+++ b/src/graphnet/training/utils.py
@@ -115,8 +115,7 @@ def make_train_validation_dataloader(
 ) -> Tuple[DataLoader, DataLoader]:
     """Construct train and test `DataLoader` instances."""
     # Reproducibility
-    rng = np.random.RandomState(seed=seed)
-
+    rng = np.random.default_rng(seed=seed)
     # Checks(s)
     if isinstance(pulsemaps, str):
         pulsemaps = [pulsemaps]
@@ -159,13 +158,13 @@ def make_train_validation_dataloader(
             frac=1, replace=False, random_state=rng
         )
         training_df, validation_df = train_test_split(
-            shuffled_df, test_size=test_size, random_state=rng
+            shuffled_df, test_size=test_size, random_state=seed
         )
         training_selection = training_df.values.tolist()
         validation_selection = validation_df.values.tolist()
     else:
         training_selection, validation_selection = train_test_split(
-            selection, test_size=test_size, random_state=rng
+            selection, test_size=test_size, random_state=seed
         )
 
     # Create DataLoaders