diff --git a/miplearn/features/sample.py b/miplearn/features/sample.py
index e75f42e..5c19172 100644
--- a/miplearn/features/sample.py
+++ b/miplearn/features/sample.py
@@ -167,8 +167,8 @@ class Hdf5Sample(Sample):
     are actually accessed, and therefore it is more scalable.
     """
 
-    def __init__(self, filename: str) -> None:
-        self.file = h5py.File(filename, "r+")
+    def __init__(self, filename: str, mode: str = "r+") -> None:
+        self.file = h5py.File(filename, mode)
 
     @overrides
     def get_bytes(self, key: str) -> Optional[bytes]:
diff --git a/miplearn/instance/file.py b/miplearn/instance/file.py
index f0e5664..14f9fdf 100644
--- a/miplearn/instance/file.py
+++ b/miplearn/instance/file.py
@@ -119,7 +119,7 @@ class FileInstance(Instance):
 
     @classmethod
     def save(cls, instance: Instance, filename: str) -> None:
-        h5 = Hdf5Sample(filename)
+        h5 = Hdf5Sample(filename, mode="w")
         instance_pkl = pickle.dumps(instance)
         h5.put_bytes("pickled", instance_pkl)
 
diff --git a/tests/instance/test_file.py b/tests/instance/test_file.py
index 14c7c73..6ff9767 100644
--- a/tests/instance/test_file.py
+++ b/tests/instance/test_file.py
@@ -15,17 +15,17 @@ def test_usage() -> None:
     original = GurobiSolver().build_test_instance_knapsack()
 
     # Save instance to disk
-    file = tempfile.NamedTemporaryFile()
-    FileInstance.save(original, file.name)
-    sample = Hdf5Sample(file.name)
+    filename = tempfile.mktemp()
+    FileInstance.save(original, filename)
+    sample = Hdf5Sample(filename)
     assert len(sample.get_bytes("pickled")) > 0
 
     # Solve instance from disk
     solver = LearningSolver(solver=GurobiSolver())
-    solver.solve(FileInstance(file.name))
+    solver.solve(FileInstance(filename))
 
     # Assert HDF5 contains training data
-    sample = FileInstance(file.name).get_samples()[0]
+    sample = FileInstance(filename).get_samples()[0]
     assert sample.get_scalar("mip_lower_bound") == 1183.0
     assert sample.get_scalar("mip_upper_bound") == 1183.0
     assert len(sample.get_vector("lp_var_values")) == 5