[Communicator] Add monkey patch (#30)

vllm_ascend/communicator.py

@@ -17,12 +17,62 @@
 
 import torch
 import torch.distributed as dist
-from vllm.distributed.device_communicators.base_communicator import \
-    CommunicatorBase
 
 
-class NPUCommunicator(CommunicatorBase):
+class NPUCommunicator:
+
+    def __init__(self, group, unique_name=""):
+        self.group = group
+        self.unique_name = unique_name
+        self.rank = dist.get_rank(group)
+        self.world_size = dist.get_world_size(self.group)
+        self.ranks = dist.get_process_group_ranks(self.group)
+        global_rank = dist.get_rank()
+        self.rank_in_group = dist.get_group_rank(self.group, global_rank)
 
     def all_reduce(self, x: torch.Tensor) -> torch.Tensor:
         dist.all_reduce(x, group=self.group)
         return x
+
+    def gather(self, input_: torch.Tensor, dst: int = 0, dim: int = -1):
+        # NOTE: We assume that the input tensor is on the same device across
+        # all the ranks.
+        # NOTE: `dst` is the local rank of the destination rank.
+        # Allocate output tensor.
+        if self.rank_in_group == dst:
+            gather_list = [
+                torch.empty_like(input_) for _ in range(self.world_size)
+            ]
+        else:
+            gather_list = None
+        # Gather.
+        dist.gather(input_, gather_list, dst=self.ranks[dst], group=self.group)
+        if self.rank_in_group == dst:
+            output_tensor = torch.cat(gather_list, dim=dim)
+        else:
+            output_tensor = None
+        return output_tensor
+
+    def all_gather(self, input_: torch.Tensor, dim: int = -1) -> torch.Tensor:
+        if dim < 0:
+            # Convert negative dim to positive.
+            dim += input_.dim()
+        input_size = input_.size()
+        # NOTE: we have to use concat-style all-gather here,
+        # stack-style all-gather has compatibility issues with
+        # torch.compile . see https://github.com/pytorch/pytorch/issues/138795
+        output_size = (input_size[0] * self.world_size, ) + input_size[1:]
+        # Allocate output tensor.
+        output_tensor = torch.empty(output_size,
+                                    dtype=input_.dtype,
+                                    device=input_.device)
+        # All-gather.
+        dist.all_gather_into_tensor(output_tensor, input_, group=self.group)
+        # Reshape
+        output_tensor = output_tensor.reshape((self.world_size, ) + input_size)
+        output_tensor = output_tensor.movedim(0, dim)
+        output_tensor = output_tensor.reshape(input_size[:dim] +
+                                              (self.world_size *
+                                               input_size[dim], ) +
+                                              input_size[dim + 1:])
+        return output_tensor

vllm_ascend/patch/__init__.py

@@ -0,0 +1,18 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+# This file is a part of the vllm-ascend project.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from vllm_ascend.patch import patch_commnicator  # noqa

vllm_ascend/patch/patch_commnicator.py

@@ -0,0 +1,69 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+# This file is a part of the vllm-ascend project.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# This file is used to monkey patch communicator in vllm to support ascend.
+# Remove this file when vllm support by
+# https://github.com/vllm-project/vllm/pull/11324.
+
+import torch
+from vllm.distributed.parallel_state import GroupCoordinator
+from vllm.utils import resolve_obj_by_qualname
+
+
+class GroupCoordinatorPatch(GroupCoordinator):
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.device = torch.device(f"npu:{self.local_rank}")
+
+        from vllm.platforms import current_platform
+        device_comm_cls = resolve_obj_by_qualname(
+            current_platform.get_device_communicator_cls())
+        # we have checked and ensure that reusing tpu tag here is fine.
+        use_custom_device = kwargs.get("use_tpu_communicator", False)
+        if use_custom_device and self.world_size > 1:
+            self.communicator = device_comm_cls(group=self.device_group,
+                                                unique_name=self.unique_name)
+
+    def all_reduce(self, input_):
+        # Bypass the function if we are using only 1 device.
+        if self.world_size == 1:
+            return input_
+
+        return self.communicator.all_reduce(input_)
+
+    def gather(self, input_, dst=0, dim=-1):
+        # Bypass the function if we are using only 1 device.
+        if self.world_size == 1:
+            return input_
+        assert -input_.dim() <= dim < input_.dim(), (
+            f"Invalid dim ({dim}) for input tensor with shape {input_.size()}")
+        if dim < 0:
+            # Convert negative dim to positive.
+            dim += input_.dim()
+
+        return self.communicator.gather(input_, dst, dim)
+
+    def all_gather(self, input_, dim=-1):
+        # Bypass the function if we are using only 1 device.
+        if self.world_size == 1:
+            return input_
+        assert -input_.dim() <= dim < input_.dim(), (
+            f"Invalid dim ({dim}) for input tensor with shape {input_.size()}")
+        return self.communicator.all_gather(input_, dim)
+
+
+GroupCoordinator = GroupCoordinatorPatch

vllm_ascend/platform.py

@@ -96,8 +96,9 @@ def mem_get_info(cls) -> Tuple[int, int]:
 
     @classmethod
     def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
-        # Register ops when setup.
+        # Register ops and patch when setup.
         from vllm_ascend import ops  # noqa: F401
+        from vllm_ascend import patch  # noqa: F401
 
         parallel_config = vllm_config.parallel_config
         if parallel_config.worker_cls == "auto":