[CI] Upgrade vllm version - 0925 (#3167)

Upgrade vLLM to newest commit.

1. Remove the useless func get_state_cls, it has been removed from vLLM
already.
e6750d0b18
2. Fix ut broken by
6160ba4151


- vLLM version: v0.10.2
- vLLM main:
b1068903fd

---------

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

tests/ut/attention/test_attention_v1.py

@@ -7,8 +7,7 @@ from vllm_ascend.attention.attention_v1 import (AscendAttentionBackend,
                                                 AscendAttentionBackendImpl,
                                                 AscendAttentionMetadataBuilder,
                                                 AscendAttentionState,
-                                                AscendMetadata,
-                                                CommonAttentionState)
+                                                AscendMetadata)
 from vllm_ascend.attention.utils import AscendCommonAttentionMetadata
 
 
@@ -25,10 +24,6 @@ class TestAscendAttentionBackend(TestBase):
         self.assertEqual(AscendAttentionBackend.get_metadata_cls(),
                          AscendMetadata)
 
-    def test_get_state_cls(self):
-        self.assertEqual(AscendAttentionBackend.get_state_cls(),
-                         CommonAttentionState)
-
     def test_get_builder_cls(self):
         self.assertEqual(AscendAttentionBackend.get_builder_cls(),
                          AscendAttentionMetadataBuilder)

tests/ut/ops/test_fused_ops.py

@@ -201,6 +201,8 @@ def default_moe_config():
 def moe_method(mock_dist_env):
     moe = MagicMock()
     moe.moe_parallel_config.return_value = MagicMock(ep_size=4)
+    moe.moe_parallel_config.use_ep = False
+    moe.moe_parallel_config.dp_size = 1
     return AscendUnquantizedFusedMoEMethod(moe)
 
 

tests/ut/torchair/ops/test_torchair_fused_moe.py

@@ -153,6 +153,8 @@ def default_moe_config():
 def moe_method(mock_dist_env):
     moe = MagicMock()
     moe.moe_parallel_config.return_value = MagicMock(ep_size=4)
+    moe.moe_parallel_config.use_ep = False
+    moe.moe_parallel_config.dp_size = 1
     return TorchairAscendUnquantizedFusedMoEMethod(moe)