diff --git a/src/careamics/utils/torch_utils.py b/src/careamics/utils/torch_utils.py
index b2b9755fb..b46a253ff 100644
--- a/src/careamics/utils/torch_utils.py
+++ b/src/careamics/utils/torch_utils.py
@@ -4,7 +4,6 @@
 These functions are used to control certain aspects and behaviours of PyTorch.
 """
 import logging
-import os
 import sys
 
 import torch
@@ -48,46 +47,44 @@ def compile_model(model: torch.nn.Module) -> torch.nn.Module:
         return model
 
 
-def seed_everything(seed: int) -> None:
-    """
-    Seed all random number generators for reproducibility.
-
-    Parameters
-    ----------
-    seed : int
-        Seed.
-    """
-    import random
-
-    import numpy as np
-
-    random.seed(seed)
-    np.random.seed(seed)
-    torch.manual_seed(seed)
-    torch.cuda.manual_seed_all(seed)
-
-
-def setup_cudnn_reproducibility(
-    deterministic: bool = True, benchmark: bool = True
-) -> None:
-    """
-    Prepare CuDNN benchmark and sets it to be deterministic/non-deterministic mode.
-
-    https://pytorch.org/docs/stable/notes/randomness.html#cuda-convolution-benchmarking.
-
-    Parameters
-    ----------
-    deterministic : bool
-        Deterministic mode, if running CuDNN backend.
-    benchmark : bool
-        If True, uses CuDNN heuristics to figure out which algorithm will be most
-        performant for your model architecture and input. False may slow down training.
-    """
-    if torch.cuda.is_available():
-        if deterministic:
-            deterministic = os.environ.get("CUDNN_DETERMINISTIC", "True") == "True"
-        torch.backends.cudnn.deterministic = deterministic
-
-        if benchmark:
-            benchmark = os.environ.get("CUDNN_BENCHMARK", "True") == "True"
-        torch.backends.cudnn.benchmark = benchmark
+# def seed_everything(seed: int) -> None:
+#     """
+#     Seed all random number generators for reproducibility.
+
+#     Parameters
+#     ----------
+#     seed : int
+#         Seed.
+#     """
+#     import random
+
+#     import numpy as np
+
+#     random.seed(seed)
+#     np.random.seed(seed)
+#     torch.manual_seed(seed)
+#     torch.cuda.manual_seed_all(seed)
+
+
+# def setup_cudnn_reproducibility(
+#     deterministic: bool = True, benchmark: bool = True
+# ) -> None:
+#     """
+#     Prepare CuDNN benchmark and sets it to be deterministic/non-deterministic mode.
+
+#     Parameters
+#     ----------
+#     deterministic : bool
+#         Deterministic mode, if running CuDNN backend.
+#     benchmark : bool
+#         If True, uses CuDNN heuristics to figure out which algorithm will be most
+#         performant for your model architecture and input. False may slow down training
+#     """
+#     if torch.cuda.is_available():
+#         if deterministic:
+#             deterministic = os.environ.get("CUDNN_DETERMINISTIC", "True") == "True"
+#         torch.backends.cudnn.deterministic = deterministic
+
+#         if benchmark:
+#             benchmark = os.environ.get("CUDNN_BENCHMARK", "True") == "True"
+#         torch.backends.cudnn.benchmark = benchmark
diff --git a/tests/utils/test_torch_utils.py b/tests/utils/test_torch_utils.py
index 8114eadb0..947a3685e 100644
--- a/tests/utils/test_torch_utils.py
+++ b/tests/utils/test_torch_utils.py
@@ -3,7 +3,6 @@
 
 from careamics.utils.torch_utils import (
     get_device,
-    setup_cudnn_reproducibility,
 )
 
 
@@ -14,10 +13,10 @@ def test_get_device(device):
     assert device.type == "cuda" if torch.cuda.is_available() else "cpu"
 
 
-@pytest.mark.gpu
-@pytest.mark.parametrize("deterministic", [True, False])
-@pytest.mark.parametrize("benchmark", [True, False])
-def test_setup_cudnn_reproducibility(deterministic, benchmark):
-    setup_cudnn_reproducibility(deterministic=deterministic, benchmark=benchmark)
-    assert torch.backends.cudnn.deterministic == deterministic
-    assert torch.backends.cudnn.benchmark == benchmark
+# @pytest.mark.gpu
+# @pytest.mark.parametrize("deterministic", [True, False])
+# @pytest.mark.parametrize("benchmark", [True, False])
+# def test_setup_cudnn_reproducibility(deterministic, benchmark):
+#     setup_cudnn_reproducibility(deterministic=deterministic, benchmark=benchmark)
+#     assert torch.backends.cudnn.deterministic == deterministic
+#     assert torch.backends.cudnn.benchmark == benchmark