Introduce ManagedDeviceMesh to integrate DeviceMesh with TorchFT

Summary:
ManagedDeviceMesh allow users to manipulate DeviceMesh with TorchFT ManagedProcessGroup.

ghstack-source-id: faf2012b63df387807ddac7e9dc30af634abc3c5
Pull Request resolved: #56

Author: fegin

torchft/fsdp_test.py

@@ -0,0 +1,70 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+import os
+from concurrent.futures import ThreadPoolExecutor
+from typing import Any, Dict, Tuple
+from unittest import skipUnless, TestCase
+from unittest.mock import Mock
+
+import torch
+import torch.distributed as dist
+from torch import nn
+from torch._C._distributed_c10d import (
+    _resolve_process_group,
+    AllgatherOptions,
+    AllreduceOptions,
+    BroadcastOptions,
+    ReduceOp,
+)
+from torch.distributed import (
+    _functional_collectives,
+    get_world_size,
+    ReduceOp,
+    TCPStore,
+    Work,
+)
+from torch.distributed._composable.fsdp import fully_shard
+from torch.distributed.device_mesh import init_device_mesh
+from torch.testing._internal.common_distributed import MultiProcessTestCase
+
+from torchft.manager import Manager
+from torchft.process_group import ft_init_device_mesh, ManagedProcessGroup
+
+
+class FSDPTest(MultiProcessTestCase):
+    @property
+    def world_size(self):
+        return 4
+
+    def setUp(self):
+        super().setUp()
+        os.environ["TORCH_NCCL_DESYNC_DEBUG"] = "0"
+        self._spawn_processes()
+
+    def test_fsdp(self) -> None:
+        group_size = self.world_size // 2
+        group = self.rank // group_size
+        group_rank = self.rank % group_size
+
+        os.environ["MASTER_ADDR"] = "127.0.0.1"
+        os.environ["MASTER_PORT"] = str(12346 + group)
+        os.environ["RANK"] = str(group_rank)
+        os.environ["WORLD_SIZE"] = str(group_size)
+
+        manager = Mock(spec=Manager)
+        device_mesh = ft_init_device_mesh(
+            device_type="cuda",
+            mesh_shape=(2, 2),
+            mesh_dim_names=("dp_replicate", "dp_shard"),
+            replicate_dim=0,
+            manager=manager,
+        )
+        manager.num_participants.return_value = 1
+        model = nn.Linear(128, 128).cuda()
+        batch = torch.randn(4, 128).cuda()
+        shard_model = fully_shard(model, mesh=device_mesh)
+        shard_model(batch).mean().backward()

torchft/process_group.py

@@ -20,7 +20,7 @@
 import threading
 from abc import ABC
 from datetime import timedelta
-from typing import TYPE_CHECKING, Dict, List, Optional, Type
+from typing import TYPE_CHECKING, Dict, List, Optional, Tuple, Type, Union
 
 import torch
 import torch.distributed as dist
@@ -38,6 +38,7 @@
     Store,
     TCPStore,
     get_rank,
+    init_device_mesh,
 )
 from torch.distributed.distributed_c10d import Work, _world
 from torch.futures import Future
@@ -130,17 +131,7 @@ def size(self) -> int:
     def getBackendName(self) -> str:
         raise NotImplementedError("not implemented")
 
-    def register(self, name: str) -> "ProcessGroup":
-        """
-        Registers the process group with the global registry. This enables usage
-        with things like functional_collectives which are compilable.
-
-        This should only be called once.
-
-        Args:
-            name: name must be a unique name for this process group
-        """
-
+    def _register(self, name: str) -> str:
         group_name = f"{self.getBackendName()}:{name}"
 
         # This is needed for DeviceMesh and functional collectives to work.
@@ -158,6 +149,21 @@ def create_pg(
             devices = ["cpu"]
         dist.Backend.register_backend(group_name, create_pg, devices=devices)
 
+        return group_name
+
+    def register(self, name: str) -> "ProcessGroup":
+        """
+        Registers the process group with the global registry. This enables usage
+        with things like functional_collectives which are compilable.
+
+        This should only be called once.
+
+        Args:
+            name: name must be a unique name for this process group
+        """
+
+        group_name = self._register(name)
+
         return dist.new_group(
             ranks=[dist.get_rank()],
             backend=group_name,
@@ -496,6 +502,9 @@ def allreduce(self, tensors: List[torch.Tensor], opts: object) -> Work:
     def size(self) -> int:
         return self._manager.num_participants()
 
+    def getBackendName(self) -> str:
+        return self._manager._pg.getBackendName()
+
 
 class _BabyWork(Work):
     def __init__(
@@ -689,7 +698,6 @@ def _future_handler(self, future_queue: mp.Queue) -> None:
             logger.exception(f"got unexpected error in future handler: {e}")
 
     def _get_future(self, op_id: int) -> Future[object]:
-
         with self._futures_lock:
             fut = Future()  # pyre-fixme[29]: is not a function
             self._futures[op_id] = fut
@@ -797,3 +805,201 @@ def extend_device_mesh(
         mesh=mesh.mesh.unsqueeze(dim),
         mesh_dim_names=tuple(mesh_dim_names),
     )
+
+
+class _ManagedDeviceMesh(DeviceMesh):
+    def __init__(
+        self,
+        mesh: Optional[DeviceMesh],
+        mesh_dim_names: Tuple[str],
+        replicate_pg: ManagedProcessGroup,
+        replicate_dim: int,
+        parent: Optional["_ManagedDeviceMesh"],
+    ):
+        if mesh is None and parent is not None:
+            raise ValueError(
+                "_ManagedDeviceMesh doesn't support both mesh and parent are None."
+            )
+        self.mesh = mesh
+        self.mesh_dim_names = mesh_dim_names
+        self.replicate_pg = replicate_pg
+        self.replicate_dim = replicate_dim
+        self.replicate_dim_name = mesh_dim_names[replicate_dim]
+        self.parent = parent
+        self.flatten_meshes = {}
+        self.device_type = mesh.device_type if mesh is not None else parent.device_type
+        self._flatten_mesh_list = tuple()
+        self._thread_id = None
+
+    def __getitem__(self, mesh_dim_names: Union[str, Tuple[str, ...]]) -> DeviceMesh:
+        if isinstance(mesh_dim_names, str):
+            if mesh_dim_names == self.replicate_dim_name:
+                return _ManagedDeviceMesh(
+                    mesh=None,
+                    mesh_dim_names=(mesh_dim_names,),
+                    replicate_pg=self.replicate_pg,
+                    replicate_dim=0,
+                    parent=self,
+                )
+            elif mesh_dim_names in self.flatten_meshes:
+                return self.flatten_meshes[mesh_dim_names]
+            else:
+                return self.mesh[mesh_dim_names]
+        else:
+            assert isinstance(mesh_dim_names, tuple)
+            if self.replicate_dim_name in mesh_dim_names:
+                return self.mesh[mesh_dim_names]
+            else:
+                return _ManagedDeviceMesh(
+                    self.mesh[mesh_dim_names],
+                    mesh_dim_names,
+                    self.replicate_pg,
+                    mesh_dim_name.index(self.replicate_dim_name),
+                    parent=self,
+                )
+
+    def _real_mesh_dim(self, mesh_dim: int) -> int:
+        return mesh_dim - 1 if mesh_dim > self.replicate_dim else mesh_dim
+
+    def get_group(self, mesh_dim: Optional[str] = None) -> BaseProcessGroup:
+        if mesh_dim is None:
+            assert self.mesh is None
+            return self.replicate_pg
+        elif mesh_dim == self.replicate_dim_name:
+            return self.replicate_pg
+        else:
+            return self.mesh.get_group(self._real_mesh_dim(mesh_dim))
+
+    def _flatten(self, mesh_dim_name: str) -> "DeviceMesh":
+        flatten_mesh = _FlattenDeviceMesh(self)
+        if self.parent is None:
+            self.flatten_meshes[mesh_dim_name] = flatten_mesh
+        else:
+            self.parent.flatten_meshes[mesh_dim_name] = flatten_mesh
+        return flatten_mesh
+
+    def size(self, mesh_dim: Optional[int] = None) -> int:
+        if mesh_dim is None:
+            if self.mesh is None:
+                return self.replicate_pg.size()
+            else:
+                return self.mesh.size() * self.replicate_pg.size()
+        elif mesh_dim == self.replicate_dim:
+            return self.replicate_pg.size()
+        else:
+            return self.mesh.size(self._real_mesh_dim(mesh_dim))
+
+    @property
+    def ndim(self) -> int:
+        return self.mesh.ndim + 1
+
+    @property
+    def shape(self) -> Tuple[int, ...]:
+        ret = list(self.mesh.shape)
+        ret.insert(self.replicate_dim, self.replicate_pg.size())
+
+    def get_rank(self) -> int:
+        return self.mesh.get_rank()
+
+    def get_local_rank(self, mesh_dim: Optional[Union[int, str]] = None) -> int:
+        if mesh_dim is None:
+            if self.mesh is None:
+                return get_rank(self.replicate_pg)
+
+            assert self.replicate_dim == 0, "replicate_dim must be the first one"
+            other_dim_size = self.mesh.size()
+            other_dim_rank = self.mesh.get_local_rank()
+            replicate_pg_rank = get_rank(self.replicate_pg)
+            return other_dim_size * replicate_pg_rank + other_dim_rank
+        elif mesh_dim in (self.replicate_dim, self.replicate_dim_name):
+            return get_rank(self.replicate_pg)
+        else:
+            return self.mesh.get_local_rank(self._real_mesh_dim(mesh_dim))
+
+    def get_coordinate(self) -> Optional[List[int]]:
+        """
+        Return the relative indices of this rank relative to all
+        dimensions of the mesh. If this rank is not part of the mesh, return None.
+        """
+        return self.mesh._coordinate_on_dim if self.mesh._coordinate_on_dim else None
+
+    def get_all_groups(self) -> List[ProcessGroup]:
+        raise NotImplementedError
+
+
+class _FlattenDeviceMesh(DeviceMesh):
+    def __init__(self, managed_mesh: _ManagedDeviceMesh):
+        self.managed_mesh = managed_mesh
+
+    def __getitem__(self, mesh_dim_names: Union[str, Tuple[str, ...]]) -> DeviceMesh:
+        raise NotImplementedError
+
+    def get_group(self, mesh_dim: Optional[str] = None) -> BaseProcessGroup:
+        raise NotImplementedError
+
+    def _flatten(self, mesh_dim_name: str) -> "DeviceMesh":
+        raise NotImplementedError
+
+    def size(self, mesh_dim: Optional[int] = None) -> int:
+        assert mesh_dim is None
+        return self.managed_mesh.size()
+
+    @property
+    def ndim(self) -> int:
+        raise NotImplementedError
+
+    @property
+    def shape(self) -> Tuple[int, ...]:
+        raise NotImplementedError
+
+    def get_rank(self) -> int:
+        raise NotImplementedError
+
+    def get_local_rank(self, mesh_dim: Optional[Union[int, str]] = None) -> int:
+        assert mesh_dim is None
+        return self.managed_mesh.get_local_rank()
+
+    def get_all_groups(self) -> List[ProcessGroup]:
+        raise NotImplementedError
+
+
+def ft_init_device_mesh(
+    *,
+    device_type: str,
+    mesh_shape: Tuple[int, ...],
+    mesh_dim_names: Tuple[str, ...],
+    replicate_dim: int,
+    manager: "Manager",
+):
+    # We need to mislead DeviceMesh into thinking that replicate_dim has only
+    # 1 rank.
+    _mesh_shape = list(mesh_shape)
+    _mesh_shape.pop(replicate_dim)
+    _mesh_dim_names = list(mesh_dim_names)
+    _mesh_dim_names.pop(replicate_dim)
+    mesh = init_device_mesh(
+        device_type,
+        mesh_shape=tuple(_mesh_shape),
+        mesh_dim_names=tuple(_mesh_dim_names),
+    )
+
+    if device_type == "cpu":
+        pg = ProcessGroupGloo()
+    elif device_type == "cuda":
+        pg = ProcessGroupNCCL()
+    else:
+        raise ValueError()
+
+    manager._pg = pg
+    replicate_pg = ManagedProcessGroup(manager)
+    # We have to use MultiProcessTestCase, otherwise c10d will complain
+    # the same backend has been registered.
+    replicate_pg.register(mesh_dim_names[replicate_dim])
+
+    return _ManagedDeviceMesh(
+        mesh=mesh,
+        mesh_dim_names=mesh_dim_names,
+        replicate_pg=replicate_pg,
+        replicate_dim=replicate_dim,
+        parent=None,
+    )