[torch.compile] improve allreduce registration (vllm-project#9061)

Signed-off-by: Alvant <[email protected]>

Author: youkaichao

vllm/distributed/device_communicators/custom_all_reduce.py

@@ -265,24 +265,21 @@ def all_reduce_unreg(self, inp: torch.Tensor, out: torch.Tensor = None):
 
     def custom_all_reduce(self, input: torch.Tensor) -> Optional[torch.Tensor]:
         # when custom allreduce is disabled, this will be None
-        if self.disabled:
+        if self.disabled or not self.should_custom_ar(input):
             return None
         if self._IS_CAPTURING:
             if torch.cuda.is_current_stream_capturing():
-                if self.should_custom_ar(input):
-                    return self.all_reduce_reg(input)
+                return self.all_reduce_reg(input)
             else:
-                if self.should_custom_ar(input):
-                    # if warm up, mimic the allocation pattern
-                    # since custom allreduce is out-of-place
-                    return torch.empty_like(input)
+                # if warm up, mimic the allocation pattern
+                # since custom allreduce is out-of-place
+                return torch.empty_like(input)
         else:
             # note: outside of cuda graph context,
             # custom allreduce incurs a cost of cudaMemcpy, which should
             # be small(<=1% of overall latency) compared to the performance
             # gains of using custom kernels
-            if self.should_custom_ar(input):
-                return self.all_reduce_unreg(input)
+            return self.all_reduce_unreg(input)
 
         return None
 

vllm/distributed/parallel_state.py

@@ -105,7 +105,7 @@ def inplace_all_reduce(tensor: torch.Tensor, group_name: str) -> None:
         group = _groups[group_name]()
         if group is None:
             raise ValueError(f"Group {group_name} is destroyed.")
-        group._all_reduce(tensor)
+        group._all_reduce_in_place(tensor)
 
     @inplace_all_reduce.register_fake
     def _(tensor: torch.Tensor, group_name: str) -> None:
@@ -118,7 +118,7 @@ def outplace_all_reduce(tensor: torch.Tensor,
         group = _groups[group_name]()
         if group is None:
             raise ValueError(f"Group {group_name} is destroyed.")
-        return group._all_reduce(tensor)
+        return group._all_reduce_out_place(tensor)
 
     @outplace_all_reduce.register_fake
     def _(tensor: torch.Tensor, group_name: str) -> torch.Tensor:
@@ -338,40 +338,33 @@ def all_reduce(self, input_: torch.Tensor) -> torch.Tensor:
             return input_
 
         if not supports_custom_op():
-            return self._all_reduce(input_)
+            self._all_reduce_in_place(input_)
+            return input_
 
         if self.tpu_communicator is not None and \
             not self.tpu_communicator.disabled:
             # TPU handles Dynamo with its own logic.
-            return self._all_reduce(input_)
+            return self.tpu_communicator.all_reduce(input_)
 
-        if self.ca_comm is not None and self.ca_comm.should_custom_ar(input_):
+        if self.ca_comm is not None and \
+            not self.ca_comm.disabled and \
+                self.ca_comm.should_custom_ar(input_):
             return torch.ops.vllm.outplace_all_reduce(
                 input_, group_name=self.unique_name)
         else:
             torch.ops.vllm.inplace_all_reduce(input_,
                                               group_name=self.unique_name)
             return input_
 
-    def _all_reduce(self, input_: torch.Tensor) -> torch.Tensor:
-        """
-        The actual all-reduce implementation.
-
-        NOTE: This operation will be applied in-place or out-of-place. 
-        Always assume this function modifies its input, but use the return
-        value as the output.
-        """
+    def _all_reduce_out_place(self, input_: torch.Tensor) -> torch.Tensor:
         ca_comm = self.ca_comm
+        assert ca_comm is not None
+        assert not ca_comm.disabled
+        out = ca_comm.custom_all_reduce(input_)
+        assert out is not None
+        return out
 
-        # For TPUs, use TPU communicator.
-        tpu_comm = self.tpu_communicator
-        if tpu_comm is not None and not tpu_comm.disabled:
-            return tpu_comm.all_reduce(input_)
-
-        if ca_comm is not None:
-            out = ca_comm.custom_all_reduce(input_)
-            if out is not None:
-                return out
+    def _all_reduce_in_place(self, input_: torch.Tensor) -> None:
         pynccl_comm = self.pynccl_comm
         if (pynccl_comm is not None and not pynccl_comm.disabled):
             pynccl_comm.all_reduce(input_)
@@ -380,7 +373,6 @@ def _all_reduce(self, input_: torch.Tensor) -> torch.Tensor:
             ipex.distributed.all_reduce(input_, group=self.device_group)
         else:
             torch.distributed.all_reduce(input_, group=self.device_group)
-        return input_
 
     def all_gather(self, input_: torch.Tensor, dim: int = -1) -> torch.Tensor:
         world_size = self.world_size