fix: test from_raggedtensor on CUDA (#3298)

Use a consistent back-end for comparison.
scikit-hep · Nov 8, 2024 · 6c47e1d · 6c47e1d
1 parent e0f5feb
commit 6c47e1d
Showing 1 changed file with 4 additions and 0 deletions.
diff --git a/tests/test_3210_to_raggedtensor_from_raggedtensor.py b/tests/test_3210_to_raggedtensor_from_raggedtensor.py
@@ -111,6 +111,10 @@ def test_convert_from_raggedtensor():
         result1.content.data == ak.to_backend(np_array1, result1.backend).layout.data
     ).all()
     assert (result1.offsets.data == [0, 2, 3, 3, 5]).all()
+    assert (
+        result1.offsets.data
+        == ak.to_backend([0, 2, 3, 3, 5], result1.backend).layout.data
+    ).all()
     assert from_raggedtensor(tf_array1).to_list() == ak_array1.to_list()
 
     tf_array2 = tf.RaggedTensor.from_nested_row_splits(