diff --git a/.github/verify_labels.py b/.github/verify_labels.py
index 6b555eb55..08ba8aa4a 100644
--- a/.github/verify_labels.py
+++ b/.github/verify_labels.py
@@ -51,7 +51,11 @@
 
 
 def query_repo(cmd: str, *, accept) -> Any:
-    response = requests.get(f"https://api.github.com/repos/{GH_ORG}/{GH_REPO}/{cmd}", headers=dict(Accept=accept))
+    response = requests.get(
+        f"https://api.github.com/repos/{GH_ORG}/{GH_REPO}/{cmd}",
+        headers={"Accept": accept},
+        timeout=5,
+    )
     return response.json()
 
 
diff --git a/api/tests/conftest.py b/api/tests/conftest.py
index aa7980b77..c9729203a 100644
--- a/api/tests/conftest.py
+++ b/api/tests/conftest.py
@@ -9,7 +9,7 @@
 @pytest.fixture(scope="session")
 def mock_classification_image(tmpdir_factory):
     url = "https://m.media-amazon.com/images/I/517Nh08xqkL._AC_SX425_.jpg"
-    return requests.get(url).content
+    return requests.get(url, timeout=5).content
 
 
 @pytest_asyncio.fixture(scope="function")
diff --git a/holocron/optim/wrapper.py b/holocron/optim/wrapper.py
index 777654015..bc137ed21 100644
--- a/holocron/optim/wrapper.py
+++ b/holocron/optim/wrapper.py
@@ -67,8 +67,8 @@ def load_state_dict(self, state_dict: Dict[str, Any]) -> None:
         # Update last key of class dict
         self.__setstate__({"base_state_dict": self.base_optimizer.state_dict()})
 
-    def zero_grad(self):
-        self.base_optimizer.zero_grad()
+    def zero_grad(self, set_to_none: bool = True) -> None:
+        self.base_optimizer.zero_grad(set_to_none)
 
     def step(self, closure: Optional[Callable[[], float]] = None) -> Optional[float]:
         """Performs a single optimization step.
@@ -200,8 +200,8 @@ def load_state_dict(self, state_dict: Dict[str, Any]) -> None:
         # Update last key of class dict
         self.__setstate__({"base_state_dict": self.base_optimizer.state_dict()})
 
-    def zero_grad(self):
-        self.base_optimizer.zero_grad()
+    def zero_grad(self, set_to_none: bool = True) -> None:
+        self.base_optimizer.zero_grad(set_to_none)
 
     def step(self, closure: Optional[Callable[[], float]] = None) -> Optional[float]:
         """Performs a single optimization step.
diff --git a/pyproject.toml b/pyproject.toml
index f14629daa..9c2468672 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -29,8 +29,8 @@ classifiers = [
 ]
 dynamic = ["version"]
 dependencies = [
-    "torch>=1.9.1,<2.0.0",
-    "torchvision>=0.10.1,<1.0.0",
+    "torch>=2.0.0,<3.0.0",
+    "torchvision>=0.15.0,<1.0.0",
     "tqdm>=4.1.0",
     "numpy>=1.17.2,<2.0.0",
     "fastprogress>=1.0.0,<2.0.0",
@@ -48,6 +48,7 @@ test = [
     # cf. https://github.com/frgfm/Holocron/security/dependabot/4
     "pytest>=7.2.0",
     "coverage[toml]>=4.5.4",
+    "onnx>=1.13.0,<2.0.0",
 ]
 training = [
     "wandb>=0.10.31,<1.0.0",
diff --git a/tests/test_trainer.py b/tests/test_trainer.py
index 2b3b61c04..93eb82a62 100644
--- a/tests/test_trainer.py
+++ b/tests/test_trainer.py
@@ -132,20 +132,20 @@ def _test_trainer(
     learner.gradient_acc = 2
     learner._reset_opt(lr)
     train_iter = iter(learner.train_loader)
-    assert all(torch.all(p.grad == 0) for p in learner.model.parameters() if p.requires_grad)
+    assert all(torch.all(p.grad == 0) for p in learner.model.parameters() if p.requires_grad and p.grad is not None)
     x, target = next(train_iter)
     x, target = learner.to_cuda(x, target)
     loss = learner._get_loss(x, target)
     learner._backprop_step(loss)
     assert torch.equal(learner.model.state_dict()[ref_param], model_w)
-    assert all(torch.any(p.grad != 0) for p in learner.model.parameters() if p.requires_grad)
+    assert all(torch.any(p.grad != 0) for p in learner.model.parameters() if p.requires_grad and p.grad is not None)
     # With accumulation of 2, the update step is performed every 2 batches
     x, target = next(train_iter)
     x, target = learner.to_cuda(x, target)
     loss = learner._get_loss(x, target)
     learner._backprop_step(loss)
     assert not torch.equal(learner.model.state_dict()[ref_param], model_w)
-    assert all(torch.all(p.grad == 0) for p in learner.model.parameters() if p.requires_grad)
+    assert all(torch.all(p.grad == 0) for p in learner.model.parameters() if p.requires_grad and p.grad is not None)
 
 
 def test_classification_trainer(tmpdir_factory):