add gatherep select. (#2740)

### What this PR does / why we need it?
add gatherep select.

- vLLM version: v0.10.1.1
- vLLM main:
e599e2c65e

Signed-off-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>
Co-authored-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>

tests/ut/ops/test_ascend_forwad_context.py

@@ -0,0 +1,22 @@
+import os
+import unittest
+from unittest import mock
+
+from vllm_ascend.ascend_forward_context import get_dispatcher_name
+
+
+class TestGetDispatcherName(unittest.TestCase):
+
+    def test_get_dispatcher_name(self):
+        result = get_dispatcher_name(1, False)
+        assert result == "TokenDispatcherWithAllGather"
+        result = get_dispatcher_name(4, False)
+        assert result == "TokenDispatcherWithAll2AllV"
+        result = get_dispatcher_name(16, True)
+        assert result == "TokenDispatcherWithAll2AllV"
+        result = get_dispatcher_name(16, False)
+        assert result == "TokenDispatcherWithMC2"
+        with mock.patch.dict(os.environ,
+                             {"VLLM_ENABLE_FUSED_EXPERTS_ALLGATHER_EP": "1"}):
+            result = get_dispatcher_name(16, False)
+            assert result == "TokenDispatcherWithAllGather"

vllm_ascend/ascend_forward_context.py

@@ -45,13 +45,12 @@ def _get_fused_moe_state(ep_size: int, with_prefill: bool,
 def get_dispatcher_name(ep_size: int, with_prefill: bool) -> str:
     if ep_size == 1:
         return "TokenDispatcherWithAllGather"
-
+    elif envs_ascend.VLLM_ENABLE_FUSED_EXPERTS_ALLGATHER_EP and ep_size > 1:
-    if ep_size < 16:
+        return "TokenDispatcherWithAllGather"
+    elif ep_size < 16 or with_prefill:
         return "TokenDispatcherWithAll2AllV"
-
+    else:
-    if with_prefill:
+        return "TokenDispatcherWithMC2"
-        return "TokenDispatcherWithAll2AllV"
-    return "TokenDispatcherWithMC2"
 
 
 @contextmanager

vllm_ascend/ops/moe_dispatcher/token_dispatcher.py

@@ -28,6 +28,7 @@ import torch
 import torch_npu
 from vllm.distributed.parallel_state import get_ep_group
 
+import vllm_ascend.envs as envs_ascend
 from vllm_ascend.distributed.parallel_state import get_mc2_group
 from vllm_ascend.distributed.tensor_parallel import \
     gather_from_sequence_parallel_region
@@ -50,6 +51,9 @@ def setup_token_dispatchers(ep_size: int, **kwargs):
 
     if ep_size == 1 and "TokenDispatcherWithAllGather" not in existing_dispatchers:
         _register_token_dispatcher(TokenDispatcherWithAllGather(**kwargs))
+    elif envs_ascend.VLLM_ENABLE_FUSED_EXPERTS_ALLGATHER_EP and ep_size > 1 \
+        and "TokenDispatcherWithAllGather" not in existing_dispatchers:
+        _register_token_dispatcher(TokenDispatcherWithAllGather(**kwargs))
     elif ep_size < 16 and "TokenDispatcherWithAll2AllV" not in existing_dispatchers:
         _register_token_dispatcher(TokenDispatcherWithAll2AllV(**kwargs))
     elif ep_size >= 16: