# Copyright (c) Meta Platforms, Inc. and affiliates from functools import lru_cache from typing import cast, Dict, List, NamedTuple, Tuple import torch import torch.distributed._functional_collectives as funcol import torch.distributed._tensor.api as dtensor from torch.distributed._tensor.device_mesh import DeviceMesh from torch.distributed._tensor.placement_types import ( _Partial, DTensorSpec, Placement, Replicate, Shard, ) class _TransformInfo(NamedTuple): mesh_dim: int src_dst_placements: Tuple[Placement, Placement] # logical_shape on this mesh dimension logical_shape: List[int] def _replicate_then_shard(val: _TransformInfo) -> int: """ This is a helper function to allow reordering _TransformInfo list. The high level idea is that we want to reorder the sharding redistributions so that the DTensor redistribution is consistent with its full tensor. This is built on top of two simple assumptions: 1. Replication happens from inner to outer dimension. i.e. Shard -> Replicate 2. Sharding happens from outer to inner dimension, i.e. Replicate -> Shard So we always put the replication first and put sharding later. """ mesh_dim = val.mesh_dim src, dst = val.src_dst_placements if (dst.is_replicate() or dst.is_partial()) and src.is_shard(): return -mesh_dim elif (src.is_replicate() or src.is_partial()) and dst.is_shard(): return mesh_dim else: return 0 @lru_cache(maxsize=None) def _gen_transform_infos( src_spec: DTensorSpec, dst_spec: DTensorSpec, ) -> List[_TransformInfo]: """ Generate the transform infos from the source placements to the target placements. To transform from source to target placement it might have multiple steps, i.e. it might decompose Si -> Sj into Si -> R -> Sj. This would detects if there're mis-aligned shardings between src/dst placements. i.e. (Shard(0), Shard(0)) -> (Replicate(), Shard(0)), in this case Shard(0) -> Shard(0) for mesh dimension 1 actually needs reshard, because in the first case it's a sub-sharding of an already tensor dimension 0, and in the second case, it's the first sharding on tensor dimension 0. """ src_dim_counts: Dict[int, int] = {} dst_dim_counts: Dict[int, int] = {} transform_infos: List[_TransformInfo] = [] src_placements = src_spec.placements dst_placements = dst_spec.placements device_mesh = src_spec.device_mesh my_coordinate = device_mesh.get_coordinate() assert my_coordinate is not None # logical shape records the logic tensor shape on the mesh dimension # this is useful to ensure uneven sharding gets correct output shape initial_logical_shape = list(src_spec.shape) mesh_dims_to_logical_shape = [initial_logical_shape] mesh_ndim = len(src_placements) for i, (src, dst) in enumerate(zip(src_placements, dst_placements)): # detect mis-aligned sharding and build logical shapes current_logical_shape = mesh_dims_to_logical_shape[i] if isinstance(src, Shard): src_dim_counts[src.dim] = src_dim_counts.get(src.dim, 0) + 1 if i < mesh_ndim - 1: # calculate and save the logical shape for this sharding mesh_dim_size = device_mesh.size(mesh_dim=i) local_shard_size, _ = src._local_shard_size_on_dim( current_logical_shape[src.dim], mesh_dim_size, my_coordinate[i], ) new_logical_shape = list(current_logical_shape) new_logical_shape[src.dim] = local_shard_size mesh_dims_to_logical_shape.append(new_logical_shape) else: mesh_dims_to_logical_shape.append(current_logical_shape) if isinstance(dst, Shard): dst_dim_counts[dst.dim] = dst_dim_counts.get(dst.dim, 0) + 1 if ( isinstance(src, Shard) and isinstance(dst, Shard) and (mesh_ndim > 1 or src_dim_counts[src.dim] != dst_dim_counts[dst.dim]) ): # for the case when mesh ndim > 1 or shard dim counts are different # TODO: see if we can optimize the mesh_ndim > 1 case # decompose Shard(i) -> Shard(j) into Shard(i) -> Replicate() -> Shard(j) transform_infos.append( _TransformInfo( mesh_dim=i, src_dst_placements=(src, Replicate()), logical_shape=mesh_dims_to_logical_shape[i], ) ) transform_infos.append( _TransformInfo( mesh_dim=i, src_dst_placements=(Replicate(), dst), logical_shape=mesh_dims_to_logical_shape[i], ) ) else: transform_infos.append( _TransformInfo( mesh_dim=i, src_dst_placements=(src, dst), logical_shape=mesh_dims_to_logical_shape[i], ) ) # sort the pairs by first perform replication then sharding transform_infos.sort(key=_replicate_then_shard) return transform_infos def redistribute_local_tensor( local_tensor: torch.Tensor, current_spec: DTensorSpec, target_spec: DTensorSpec, *, async_op: bool = False, is_backward: bool = False, ) -> torch.Tensor: """ This redistribute the local tensor (torch.Tensor) from the current DTensorSpec to the target DTensorSpec, which involves the necessary collective calls to transform the local shard of the DTensor from its current spec to the target spec. """ if current_spec.mesh != target_spec.mesh: # TODO: alltoall/permute reshuffling to change device_mesh if they are not the same raise NotImplementedError("Cross device mesh comm not supported yet!") new_local_tensor = None device_mesh = current_spec.mesh my_coordinate = device_mesh.get_coordinate() if my_coordinate is None: # if rank is not part of mesh, we skip redistribute and simply return local_tensor, # which should be an empty tensor return local_tensor transform_infos = _gen_transform_infos(current_spec, target_spec) for transform_info in transform_infos: i = transform_info.mesh_dim current, target = transform_info.src_dst_placements num_chunks = device_mesh.size(mesh_dim=i) if current == target: # short cut, just use the original local tensor new_local_tensor = local_tensor continue if target.is_replicate(): # Case 1: target is Replicate if current.is_partial(): partial_spec = cast(_Partial, current) new_local_tensor = partial_spec._reduce_value( local_tensor, device_mesh, i ) elif current.is_shard(): current_placement = cast(Shard, current) new_local_tensor = current_placement._to_replicate_tensor( local_tensor, device_mesh, i, transform_info.logical_shape ) else: raise RuntimeError( f"redistribute from {current} to {target} not supported yet" ) elif target.is_shard(): # Case 2: target is Shard target_placement = cast(Shard, target) target_dim = target_placement.dim if current.is_partial(): partial_spec = cast(_Partial, current) new_local_tensor = partial_spec._reduce_shard_value( local_tensor, device_mesh, i, target_placement ) elif current.is_replicate(): # split the tensor and return the corresponding cloned local shard new_local_tensor = target_placement._replicate_to_shard( local_tensor, device_mesh, i, my_coordinate[i] ) else: assert ( current.is_shard() ), f"Current placement should be shard but found {current}" shard_spec = cast(Shard, current) if shard_spec.dim != target_placement.dim: new_local_tensor = shard_spec._to_new_shard_dim( local_tensor, device_mesh, i, transform_info.logical_shape, target_placement.dim, ) elif target.is_partial(): if current.is_replicate(): partial_spec = cast(_Partial, target) # skip the replicate to partial transformation when we are in backward pass # In this case we keep the grad as replicate, this is because we don't # want to convert the replicated gradients back to partial, although # that's logically conform with the same layout, converting the gradients # back to partial is actually useless as you would have to do reduce later # which would be more expensive than keeping it replicate! For this reason, # we keep the replicate grad here. new_local_tensor = ( partial_spec._partition_value(local_tensor, device_mesh, i) if not is_backward else local_tensor ) elif current.is_shard(): if not is_backward: raise RuntimeError( f"redistribute from {current} to {target} not supported yet" ) # for backward shard -> partial, we just need to convert the shard to replicate current_placement = cast(Shard, current) new_local_tensor = current_placement._to_replicate_tensor( local_tensor, device_mesh, i, transform_info.logical_shape ) else: # partial -> partial no op, should never hit new_local_tensor = local_tensor assert new_local_tensor is not None local_tensor = new_local_tensor assert new_local_tensor is not None, "redistribute failed!" if not async_op and isinstance(new_local_tensor, funcol.AsyncCollectiveTensor): new_local_tensor = new_local_tensor.wait() return new_local_tensor class Redistribute(torch.autograd.Function): @staticmethod def forward( # type: ignore[override] # pyre-fixme[2]: Parameter must be annotated. ctx, input: "dtensor.DTensor", device_mesh: DeviceMesh, placements: Tuple[Placement, ...], async_op: bool = False, ): current_spec = input._spec ctx.current_spec = current_spec ctx.async_op = async_op if current_spec.placements != placements: target_spec = DTensorSpec( device_mesh, placements, tensor_meta=input._spec.tensor_meta ) local_tensor = input._local_tensor output = redistribute_local_tensor( local_tensor, current_spec, target_spec, async_op=async_op ) else: # use the same local tensor if placements are the same. output = input._local_tensor return dtensor.DTensor( output, device_mesh, placements, shape=input.shape, dtype=input.dtype, requires_grad=input.requires_grad, stride=input.stride(), ) @staticmethod def backward(ctx, grad_output: "dtensor.DTensor"): # type: ignore[override] previous_spec = ctx.current_spec current_spec = grad_output._spec async_op = ctx.async_op local_tensor = grad_output._local_tensor output = redistribute_local_tensor( local_tensor, current_spec, previous_spec, async_op=async_op, is_backward=True, ) # normalize the target placement to replicate if it is partial normalized_placements: List[Placement] = [] for previous_placement in previous_spec.placements: if previous_placement.is_partial(): # keep target placement to replicate instead of partial in this case normalized_placements.append(Replicate()) else: normalized_placements.append(previous_placement) output_dtensor = dtensor.DTensor( output, previous_spec.mesh, tuple(normalized_placements), shape=grad_output.shape, dtype=grad_output.dtype, requires_grad=grad_output.requires_grad, stride=grad_output.stride(), ) return ( output_dtensor, None, None, None, )