Upgrade to new vllm commit (#3719)

### What this PR does / why we need it?
Upgrade to new vllm commit:
c9461e05a4

- Fix many imports, caused by
https://github.com/vllm-project/vllm/pull/26908
- Fix import ```sha256```, caused by
https://github.com/vllm-project/vllm/pull/27169
- Remove ```SchedulerConfig.send_delta_data```, caused by
https://github.com/vllm-project/vllm/pull/27142
- Fix ```FusedMoE``` because of dual stream execution, caused by
https://github.com/vllm-project/vllm/pull/26440

### Does this PR introduce _any_ user-facing change?
N/A

### How was this patch tested?
CI passed with new added/existing test.


- vLLM version: v0.11.0rc3
- vLLM main:
17c540a993

---------

Signed-off-by: MengqingCao <cmq0113@163.com>
Signed-off-by: Icey <1790571317@qq.com>
Co-authored-by: MengqingCao <cmq0113@163.com>

vllm_ascend/ops/fused_moe/fused_moe.py

@@ -435,10 +435,12 @@ class AscendSharedFusedMoE(SharedFusedMoE, AscendFusedMoE):
     def __init__(
         self,
         shared_experts: torch.nn.Module,
+        gate: Optional[torch.nn.Module] = None,
         use_overlapped: bool = True,
         **kwargs,
     ):
         AscendFusedMoE.__init__(self, **kwargs)
+
         self._shared_experts = shared_experts
         self.use_overlapped = use_overlapped
         self.shared_expert_stream = None
@@ -449,6 +451,16 @@ class AscendSharedFusedMoE(SharedFusedMoE, AscendFusedMoE):
                 "Sequence parallelism is enabled, shared experts are replicated for best performance."
             )
 
+        self._gate = gate
+
+    @property
+    def gate(self) -> Optional[torch.nn.Module]:
+        return self._gate if self.use_overlapped else None
+
+    @property
+    def is_internal_router(self) -> bool:
+        return False
+
     def forward(
         self,
         hidden_states: torch.Tensor,

vllm_ascend/ops/register_custom_ops.py

@@ -7,12 +7,17 @@ from vllm.distributed import (get_dp_group, get_ep_group,
                               tensor_model_parallel_all_reduce,
                               tensor_model_parallel_reduce_scatter)
 from vllm.forward_context import get_forward_context
-from vllm.utils import direct_register_custom_op
 
 import vllm_ascend.envs as envs_ascend
 from vllm_ascend.ascend_forward_context import MoECommType
 from vllm_ascend.ops.weight_prefetch import maybe_npu_prefetch
-from vllm_ascend.utils import npu_stream_switch, prefetch_stream
+from vllm_ascend.utils import (npu_stream_switch, prefetch_stream,
+                               vllm_version_is)
+
+if vllm_version_is("0.11.0"):
+    from vllm.utils import direct_register_custom_op
+else:
+    from vllm.utils.torch_utils import direct_register_custom_op
 
 
 def _maybe_all_gather_and_maybe_unpad_impl(