[CI][Bugfix] Quickfix for DPMetaData (#3234)

### What this PR does / why we need it?
Fix `dpmetadata` and `Qwen3MoeSparseMoeBlock` break introduced by
26a7a33b88 (diff-c1550d0a38469d039370567d8981969530cbfffc7302cd1778e7c2c8a9322dea)

NOTE: we maintain a different sp in vllm-ascend with vllm, thus we can
just use `cu_tokens_across_sp(1)` as `cu_tokens_across_dp_cpu`

close https://github.com/vllm-project/vllm-ascend/issues/3236,
https://github.com/vllm-project/vllm-ascend/issues/3239
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?


- vLLM version: v0.10.2
- vLLM main:
https://github.com/vllm-project/vllm/commit/releases/v0.11.0

---------

Signed-off-by: MengqingCao <cmq0113@163.com>

tests/ut/ops/test_fused_moe_prepare_and_finalize.py

@@ -8,6 +8,7 @@ from vllm_ascend.ops.moe.fused_moe_prepare_and_finalize import (
     FusedMoEPrepareAndFinalizeWithAll2All,
     FusedMoEPrepareAndFinalizeWithAllGather, FusedMoEPrepareAndFinalizeWithMC2,
     FusedMoEPrepareAndFinalizeWithNaiveMulticast)
+from vllm_ascend.utils import vllm_version_is
 
 
 class TestFusedMoEPrepareAndFinalize(unittest.TestCase):
@@ -230,8 +231,12 @@ class TestFusedMoEPrepareAndFinalize(unittest.TestCase):
                                               mock_get_dp_group):
         # Mock forward context with DP metadata
         mock_context = MagicMock()
-        mock_context.dp_metadata.cu_tokens_across_dp_cpu = torch.tensor(
-            [2, 5, 7])
+        if vllm_version_is("0.10.2"):
+            mock_context.dp_metadata.cu_tokens_across_dp_cpu = torch.tensor(
+                [2, 5, 7])
+        else:
+            mock_context.dp_metadata.cu_tokens_across_sp.return_value = torch.tensor(
+                [2, 5, 7])
         mock_get_forward_context.return_value = mock_context
 
         # Setup DP group mock

tests/ut/ops/test_fused_ops.py

@@ -28,7 +28,7 @@ from vllm_ascend.ops.fused_moe import (AscendFusedMoE,
                                        AscendUnquantizedFusedMoEMethod)
 from vllm_ascend.ops.moe.experts_selector import select_experts
 from vllm_ascend.ops.moe.moe_mlp import cumsum_group_list, unified_apply_mlp
-from vllm_ascend.utils import AscendSocVersion, adapt_patch
+from vllm_ascend.utils import AscendSocVersion, adapt_patch, vllm_version_is
 
 adapt_patch(True)
 
@@ -93,14 +93,18 @@ def mock_dist_env(mocker: MockerFixture):
 
     mock_moe_comm_method.finalize.side_effect = mock_finalize
 
-    mock_forward_context_obj = MagicMock(
-        moe_comm_method=mock_moe_comm_method,
-        moe_comm_type=MoECommType.MC2,
-        max_tokens_across_dp=10,
-        dp_metadata=MagicMock(cu_tokens_across_dp_cpu=[5, 10]),
-        mc2_mask=torch.zeros(16, dtype=torch.bool),
-        padded_num_tokens=16,
-        with_quant=False)
+    if vllm_version_is("0.10.2"):
+        dp_metadata = MagicMock(cu_tokens_across_dp_cpu=[5, 10])
+    else:
+        dp_metadata = MagicMock(num_tokens_across_dp_cpu=[5, 5])
+    mock_forward_context_obj = MagicMock(moe_comm_method=mock_moe_comm_method,
+                                         moe_comm_type=MoECommType.MC2,
+                                         max_tokens_across_dp=10,
+                                         dp_metadata=dp_metadata,
+                                         mc2_mask=torch.zeros(
+                                             16, dtype=torch.bool),
+                                         padded_num_tokens=16,
+                                         with_quant=False)
 
     with patch('torch.distributed.get_rank', return_value=0), \
         patch('torch.distributed.get_world_size', return_value=4), \

tests/ut/torchair/ops/test_torchair_fused_moe.py

@@ -26,7 +26,8 @@ from vllm_ascend.ascend_forward_context import _get_fused_moe_state
 from vllm_ascend.quantization.quant_config import AscendFusedMoEMethod
 from vllm_ascend.torchair.ops.torchair_fused_moe import (
     TorchairAscendFusedMoE, TorchairAscendUnquantizedFusedMoEMethod)
-from vllm_ascend.utils import AscendSocVersion, adapt_patch  # noqa E402
+from vllm_ascend.utils import adapt_patch  # noqa E402
+from vllm_ascend.utils import AscendSocVersion, vllm_version_is
 
 adapt_patch(True)
 
@@ -53,6 +54,10 @@ def mock_dp_and_tp_group(mocker):
 @pytest.fixture
 def mock_dist_env(mocker: MockerFixture):
     # init dist env patch
+    if vllm_version_is("0.10.2"):
+        dp_metadata = MagicMock(cu_tokens_across_dp_cpu=[5, 10])
+    else:
+        dp_metadata = MagicMock(num_tokens_across_dp_cpu=[5, 5])
 
     with patch('torch.distributed.get_rank', return_value=0), \
          patch('torch.distributed.get_world_size', return_value=4), \
@@ -80,7 +85,7 @@ def mock_dist_env(mocker: MockerFixture):
          patch('vllm_ascend.torchair.ops.torchair_fused_moe.get_forward_context',
                return_value=MagicMock(
                    max_tokens_across_dp=10,
-                   dp_metadata=MagicMock(cu_tokens_across_dp_cpu=[5, 10])
+                   dp_metadata=dp_metadata,
                )), \
         patch('vllm_ascend.torchair.ops.torchair_fused_moe.get_current_vllm_config',
                return_value=MagicMock(