[Perf]Boogu-Image Skip dense attention masks (#6871)

Signed-off-by: zhuhh97 <zhuhonghan@huawei.com>

tests/diffusion/models/boogu_image/test_boogu_image_attention_mask.py

@@ -0,0 +1,30 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+
+import pytest
+import torch
+
+from vllm_omni.diffusion.models.utils import make_attention_mask
+
+pytestmark = [pytest.mark.core_model, pytest.mark.cpu, pytest.mark.diffusion]
+
+
+@pytest.mark.parametrize("seq_lengths", [[16], [16, 16]])
+def test_attention_mask_is_skipped_for_dense_batch(seq_lengths):
+    hidden_states = torch.empty(len(seq_lengths), 16, 64)
+
+    assert make_attention_mask(hidden_states, seq_lengths) is None
+
+
+def test_attention_mask_preserves_variable_length_padding():
+    hidden_states = torch.empty(2, 16, 64)
+
+    attention_mask = make_attention_mask(hidden_states, [16, 12])
+
+    expected = torch.tensor(
+        [
+            [True] * 16,
+            [True] * 12 + [False] * 4,
+        ]
+    )
+    torch.testing.assert_close(attention_mask, expected)