diff --git a/examples/2_evaluate_pretrained_policy.py b/examples/2_evaluate_pretrained_policy.py
index bf3c442a..edbbad38 100644
--- a/examples/2_evaluate_pretrained_policy.py
+++ b/examples/2_evaluate_pretrained_policy.py
@@ -44,7 +44,7 @@ pretrained_policy_path = "lerobot/diffusion_pusht"
 # OR a path to a local outputs/train folder.
 # pretrained_policy_path = Path("outputs/train/example_pusht_diffusion")
 
-policy = DiffusionPolicy.from_pretrained(pretrained_policy_path, map_location=device)
+policy = DiffusionPolicy.from_pretrained(pretrained_policy_path)
 
 # Initialize evaluation environment to render two observation types:
 # an image of the scene and state/position of the agent. The environment
diff --git a/tests/test_policies.py b/tests/test_policies.py
index b54810f3..adc55f0b 100644
--- a/tests/test_policies.py
+++ b/tests/test_policies.py
@@ -252,11 +252,12 @@ def test_save_and_load_pretrained(dummy_dataset_metadata, tmp_path, policy_name:
     policy_cfg.input_features = {
         key: ft for key, ft in features.items() if key not in policy_cfg.output_features
     }
-    policy = policy_cls(policy_cfg)  # config.device = gpu
+    policy = policy_cls(policy_cfg)
+    policy.to(policy_cfg.device)
     save_dir = tmp_path / f"test_save_and_load_pretrained_{policy_cls.__name__}"
     policy.save_pretrained(save_dir)
-    policy_ = policy_cls.from_pretrained(save_dir, config=policy_cfg)
-    assert all(torch.equal(p, p_) for p, p_ in zip(policy.parameters(), policy_.parameters(), strict=True))
+    loaded_policy = policy_cls.from_pretrained(save_dir, config=policy_cfg)
+    torch.testing.assert_close(list(policy.parameters()), list(loaded_policy.parameters()), rtol=0, atol=0)
 
 
 @pytest.mark.parametrize("insert_temporal_dim", [False, True])