OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity
e2a3817dfd
pytorch/torch/distributed/_spmd/distribute.py
Aaron Gokaslan e2a3817dfd [BE] Enable C419 rule for any all shortcircuiting (#99890) 
Apparently https://github.com/pytorch/pytorch/pull/78142 made torch.JIT allow for simple generator expressions which allows us to enable rules that replace unnecessary list comprehensions with generators in any/all. This was originally part of #99280 but I split it off into this PR so that it can be easily reverted should anything break.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/99890
Approved by: https://github.com/justinchuby, https://github.com/kit1980, https://github.com/malfet
2023-04-25 15:02:13 +00:00

949 lines
35 KiB
Python
Raw Blame History

import logging
import operator
from dataclasses import dataclass
from enum import auto, Enum
from functools import partial
from typing import (
Any,
Callable,
cast,
Dict,
List,
Optional,
Sequence,
Set,
Tuple,
Union,
)
import torch
import torch.distributed._spmd.experimental_ops
import torch.fx as fx
import torch.nn as nn
from torch._functorch.aot_autograd import aot_module, make_boxed_func
from torch._guards import detect_fake_mode
from torch._subclasses.fake_tensor import FakeTensorMode
from torch.distributed._spmd.aot_function_patch import patched_aot_function
from torch.distributed._spmd.comm_tensor import _get_tracer
from torch.distributed._spmd.distributed_graph import DistributedGraph
from torch.distributed._spmd.graph_utils import OP
from torch.distributed._spmd.log_utils import get_logger
from torch.distributed._tensor import DeviceMesh, DTensor
from torch.distributed._tensor.dispatch import (
_CURRENT_DECOMPOSITION_TABLE,
_operator_dispatch,
)
from torch.distributed._tensor.op_schema import OpSchema
from torch.distributed._tensor.placement_types import (
_Partial,
Placement,
Replicate,
Shard,
)
from torch.distributed._tensor.redistribute import _redistribute_with_local_tensor
from torch.fx.experimental.proxy_tensor import (
make_fx,
maybe_disable_fake_tensor_mode,
proxy_slot,
)
from torch.utils._pytree import tree_flatten, tree_map, tree_map_only, tree_unflatten
# patch aot_function so that we can pass the full (non-sharded) input to capture the graph
# pyre-fixme
torch._functorch.aot_autograd.aot_function = patched_aot_function # type: ignore[assignment]
logger: Optional[logging.Logger] = None
aten = torch.ops.aten
class TrainingPhase(Enum):
FORWARD = auto()
BACKWARD = auto()
@dataclass
class Schema:
mesh: DeviceMesh
placements: List[Placement]
@dataclass
class DSymInt:
"""
DSymInt represents a value retrieved by a SymInt op from a DTensor. DSymInt
helps View and Factory ops to determine the placement and shape of the
output tensor, as those operators either do not have an input DTensor or
the input DTensor is insufficient to determine the output tensor's placement.
"""
global_value: int # value that the SymInt evaluates to
local_value: int # vaue that this SymInt evaluates to on the local shard
mesh: DeviceMesh # device mesh of the DTensor where this SymInt is retrieved from
def is_shard(self) -> bool:
return self.local_value != self.global_value
@classmethod
def from_node(cls, node: fx.Node, dtensor: DTensor) -> "DSymInt":
dim: int = 0
if node.target == aten.sym_size:
dim = cast(int, node.args[1])
return cls(
global_value=dtensor.size(dim),
local_value=dtensor.to_local().size(dim),
mesh=dtensor.device_mesh,
)
elif node.target == aten.sym_numel:
return cls(
global_value=dtensor.numel(),
local_value=dtensor.to_local().numel(),
mesh=dtensor.device_mesh,
)
elif node.target == aten.sym_stride:
dim = cast(int, node.args[1])
return cls(
global_value=dtensor.stride(dim),
local_value=dtensor.to_local().stride(dim),
mesh=dtensor.device_mesh,
)
else:
raise NotImplementedError(f"DSymInt does not support {node.target}")
def _is_partial_dtensor(obj: Any) -> bool:
"""check if object is 1) DTensor and 2) with any placement of _Partial"""
if not isinstance(obj, DTensor):
return False
is_partial = False
for placement in obj.placements:
if isinstance(placement, _Partial):
is_partial = True
break
return is_partial
def _dispatch_with_local_tensors(
op: torch._ops.OpOverload,
local_args: Tuple[Any, ...],
kwargs: Optional[Dict[str, Any]] = None,
specs: Optional[
Dict[
torch.Tensor,
Tuple[torch.Size, DeviceMesh, Sequence[Placement], Sequence[Placement]],
]
] = None,
) -> Any:
if kwargs is None:
kwargs = {}
if specs is None:
specs = {}
def redistribute(arg: Any) -> Any:
return (
_redistribute_with_local_tensor(arg, *specs[arg]) # type: ignore[index]
if isinstance(arg, torch.Tensor) and arg in specs # type: ignore[operator]
else arg
)
# TODO: this is broken because it won't redistributed potential tensors on the kwargs
return op(*tree_map(redistribute, local_args), **kwargs)
# Figure out how to specify a type spec for the return specs value
# without the entire structure.
# pyre-fixme
def _update_specs_for_redistribute(args, target_schema, redistribute):
# Code adapted from pack_args_kwargs_with_local_tensor
flatten_args, args_tree_spec = tree_flatten(args)
flatten_args_schema, _ = tree_flatten(target_schema.args_schema)
specs: Dict[
torch.Tensor,
Tuple[
torch.Size,
DeviceMesh,
Sequence[Placement],
Sequence[Placement],
],
] = {}
for i, arg in enumerate(flatten_args):
if isinstance(arg, DTensor):
if redistribute:
specs[arg._local_tensor] = (
arg.size(),
flatten_args_schema[i].mesh,
arg.placements,
flatten_args_schema[i].placements,
)
flatten_args_schema[i] = arg._local_tensor
unflattened_args = tree_unflatten(flatten_args_schema, args_tree_spec)
return specs, unflattened_args
# When no tensor redistribution is required, we only need to update non-tensor args
# of the node according to op_schema and avoid building a GraphModule just for the
# node.
def _update_node_from_op_schema(node: torch.fx.Node, op_schema: OpSchema) -> None:
flat_args, args_tree_spec = tree_flatten(node.args)
flat_args_schema, _ = tree_flatten(op_schema.args_schema)
def is_sym_int_or_int(arg: Union[int, torch.fx.Node]) -> bool:
if isinstance(arg, torch.fx.Node):
return arg.target in [
aten.sym_size,
aten.sym_numel,
aten.sym_stride,
]
return isinstance(arg, int)
assert len(flat_args) == len(flat_args_schema)
for i, (arg, arg_schema) in enumerate(zip(flat_args, flat_args_schema)):
if is_sym_int_or_int(arg) and isinstance(arg_schema, int):
flat_args[i] = arg_schema
args = tree_unflatten(flat_args, args_tree_spec)
for idx, arg in enumerate(args):
node.update_arg(idx, arg)
return None
def _remap_arg(node_to_obj: Dict[fx.Node, Any], arg: Any) -> Any:
if isinstance(arg, torch.fx.Node):
obj = node_to_obj[arg]
if _get_tracer():
# This is a shared arg, already has a tracer from previous
# tracing. Delete the tracer.
del cast(Dict[Any, Any], obj.__dict__)[proxy_slot]
return obj
else:
return arg
def unpack_sizes_and_dims(
sizes: List[Union[DSymInt, int]], mesh: DeviceMesh
) -> Tuple[List[int], List[Placement]]:
local_sizes: List[int] = [
s.local_value if isinstance(s, DSymInt) else s for s in sizes
]
placements: List[Placement] = [
Shard(i)
for i, a in enumerate(sizes)
if (isinstance(a, DSymInt) and a.is_shard())
] or [Replicate()]
assert len(placements) == mesh.ndim, (
f"The number of sharded dimensions ({len(placements)}) must "
f"match number of dimensions in device mesh ({mesh.ndim})."
)
return local_sizes, placements
def binop_sym_int_consumer_rule(node: fx.Node, args: Tuple[Any, ...]) -> DTensor:
assert len(args) == 2, f"Expect two args but got op {node.target} with args {args}"
assert isinstance(
args[0], DTensor
), f"Expect 1st argument to be DTensor but got {args[0]}"
assert isinstance(args[1], list), f"Expect 2nd argument as list but got {args[1]}"
# extract sharded dimensions in the size list, the output DTensor should
# follow these placements.
local_sizes, placements = unpack_sizes_and_dims(args[1], args[0].device_mesh)
# set node args to real int sizes.
node.args = (node.args[0], local_sizes)
op = cast(torch._ops.OpOverload, node.target)
return DTensor.from_local(
local_tensor=op(args[0]._local_tensor, local_sizes),
device_mesh=args[0].device_mesh,
placements=placements,
run_check=False,
)
def slice_backwad_sym_int_consumer_rule(
node: fx.Node, args: Tuple[Any, ...]
) -> DTensor:
grad_output, input_sizes, dim, start, end, step = args
local_sizes: List[int] = [
s.local_value if isinstance(s, DSymInt) else s for s in input_sizes
]
input_tensor = torch.zeros(
local_sizes, device=grad_output.device, dtype=grad_output.dtype
)
return DTensor.from_local(
local_tensor=torch.slice_scatter(
input_tensor, grad_output.to_local(), dim, start, end, step
),
device_mesh=grad_output.device_mesh,
placements=grad_output.placements,
run_check=False,
)
def factory_with_sizes_rule(
node: fx.Node,
args: Tuple[Any, ...],
kwargs: Dict[str, Any],
default_mesh: DeviceMesh,
) -> DTensor:
flat_args = tree_flatten(args)[0]
assert not any(isinstance(a, DTensor) for a in flat_args), (
f"Not expect DTensor argument for factory op, but got {node.target} "
f"with arguments {args}."
)
assert isinstance(args[0], list), f"Expect 2nd argument as list but got {args[1]}"
local_sizes, placements = unpack_sizes_and_dims(args[0], default_mesh)
node.args = (local_sizes, *args[1:])
op = cast(torch._ops.OpOverload, node.target)
return DTensor.from_local(
local_tensor=op(*node.args, **kwargs),
device_mesh=default_mesh,
placements=placements,
run_check=False,
)
def factory_arange_rule(
node: fx.Node,
args: Tuple[Any, ...],
kwargs: Dict[str, Any],
default_mesh: DeviceMesh,
) -> DTensor:
node.args = tree_map(lambda a: a.local_value if isinstance(a, DSymInt) else a, args)
op = cast(torch._ops.OpOverload, node.target)
return DTensor.from_local(
local_tensor=op(*node.args, **kwargs),
device_mesh=default_mesh,
placements=[Replicate()],
run_check=False,
)
def default_factory_op_rule(
node: fx.Node,
args: Tuple[Any, ...],
kwargs: Dict[str, Any],
default_mesh: DeviceMesh,
) -> DTensor:
node.args, node.kwargs = args, kwargs
op = cast(torch._ops.OpOverload, node.target)
return DTensor.from_local(
local_tensor=op(*node.args, **node.kwargs),
device_mesh=default_mesh,
placements=[Replicate()],
run_check=False,
)
# Dispatch override for view and factory ops that consume SymInt arguments,
# where the output spec should follow dimension placement where the SymInt comes
# from.
VIEW_SYM_INT_CONSUMERS: Dict[torch._ops.OpOverload, Callable] = {
aten._unsafe_view.default: binop_sym_int_consumer_rule,
aten.expand.default: binop_sym_int_consumer_rule,
aten.slice_backward.default: slice_backwad_sym_int_consumer_rule,
aten.view.default: binop_sym_int_consumer_rule,
}
FACTORY_SYM_INT_CONSUMERS: Dict[torch._ops.OpOverload, Callable] = {
aten.full.default: factory_with_sizes_rule,
aten.arange.default: factory_arange_rule,
aten.arange.start: factory_arange_rule,
}
# Dispatch override for factory ops, as DTensor cannot propogate sharding spec
# without DTensor inputs.
FACTORY_OPS: Dict[torch._ops.OpOverload, Callable] = {
aten.scalar_tensor.default: default_factory_op_rule,
}
def _get_dtensor_dispatch_graph(
node: fx.Node,
node_to_obj: Dict[fx.Node, Any],
*,
force_make_fx: bool = False,
default_mesh: Optional[DeviceMesh] = None,
) -> Optional[fx.GraphModule]:
with torch.no_grad():
# Args should be a list of objects post remapping.
args = tree_map(partial(_remap_arg, node_to_obj), node.args)
kwargs = tree_map(partial(_remap_arg, node_to_obj), node.kwargs)
op_overload = cast(torch._ops.OpOverload, node.target)
if any(a.is_shard() for a in tree_flatten(args)[0] if isinstance(a, DSymInt)):
if op_overload in VIEW_SYM_INT_CONSUMERS:
assert len(kwargs) == 0, f"Expect empty kwargs, but got {kwargs}"
node_to_obj[node] = VIEW_SYM_INT_CONSUMERS[op_overload](node, args)
return None
elif op_overload in FACTORY_SYM_INT_CONSUMERS:
assert default_mesh is not None, "Requires default mesh for factory ops"
node_to_obj[node] = FACTORY_SYM_INT_CONSUMERS[op_overload](
node, args, kwargs, default_mesh
)
return None
else:
assert isinstance(logger, logging.Logger)
logger.warning(
"Assuming using local_value from SymInt for %s"
"is mathematically correct. Full args are %s.",
op_overload,
args,
)
if node.target == aten.view.default:
# HACK: this is a hack to get around with the fact that some
# view operations on a "global" tensor is invalid usage
# but somehow the view operation on the batch input might hit it
# so we convert the view op to reshape before calling DTensor
op_overload = aten.reshape.default
# DSymInt args are not sharded on any dimension, local value and global
# value should be the same
args = tree_map(lambda a: a.local_value if isinstance(a, DSymInt) else a, args)
kwargs = tree_map(
lambda a: a.local_value if isinstance(a, DSymInt) else a, kwargs
)
if op_overload in FACTORY_OPS:
# Don't pass factory ops to DTensor dispatch, as DTensor cannot
# propagate sharding spec without DTensor inputs.
node_to_obj[node] = FACTORY_OPS[op_overload](
node, args, kwargs, default_mesh
)
return None
# run dispatch once to get the real DTensor output.
out, op_schema, output_sharding = _operator_dispatch(
op_overload,
args,
kwargs, # kwargs in this set of tests are all constants
DTensor._propagator,
)
node_to_obj[node] = out
assert output_sharding.schema_suggestions is not None
target_schema = output_sharding.schema_suggestions[0]
redistribute = target_schema is not op_schema
# If no redistribution is needed, we don't need to replace
# the original node.
if not redistribute:
_update_node_from_op_schema(node, target_schema)
return None
# TODO: this is broken when kwargs contains tensors
# or if a non-tensor kwarg was modified by the sharding propagation
# (in order to fix, need to port over pack_args_kwargs_with_local_tensor for kwargs as well)
updated_args_spec, unflattened_args = _update_specs_for_redistribute(
args, target_schema, redistribute
)
dispatch = partial(
_dispatch_with_local_tensors,
op_overload,
kwargs=kwargs,
specs=updated_args_spec,
)
gm = make_fx(dispatch, _allow_non_fake_inputs=False)(unflattened_args)
# FIXME(@wanchaol, @mrshenli): the above seems to accidentally captured
# DeviceMesh tensor ops when handling inplace operators? The ``_to_copy`` is
# not connected to graph output. So, using DCE to get rid of it, but this
# doesn't look correct.
#
# The following operators appear in the captured graph, where the dtype is
# torch.int64.
#
# get_attr _tensor_constant0 _tensor_constant0 ()
# call_function transpose aten.transpose.int (_tensor_constant0, -1, 0)
# call_function view aten.view.default (transpose, [-1, 2])
# call_function view_1 aten.view.default (view, [2])
# call_function _to_copy aten._to_copy.default (view_1,)
gm.graph.eliminate_dead_code()
return gm
def _build_dummy_add_graph(
dt: DTensor, node_to_obj: Dict[fx.Node, Any]
) -> Tuple[fx.GraphModule, Any]:
"""
Creates a graph for a dummy add function from a partial DTensor.
This dummy add is used for triggering all_reduce on a Partial DTensor
during the DTensor expansion of the traced graph.
Also returns the actual DTensor after resharding.
"""
def dummy_add(grad: torch.Tensor, zero: torch.Tensor) -> torch.Tensor:
return grad + zero
grad: torch.Tensor = dt._local_tensor
zero: torch.Tensor = torch.zeros_like(dt._local_tensor)
traced_add = make_fx(dummy_add)(grad, zero)
placeholders = [n for n in traced_add.graph.nodes if n.op == OP.PLACEHOLDER]
call_functions = [n for n in traced_add.graph.nodes if n.op == OP.CALL_FUNCTION]
assert len(placeholders) == 2
assert len(call_functions) == 1
node_to_obj[placeholders[0]] = dt
node_to_obj[placeholders[1]] = DTensor.from_local(
zero, dt.device_mesh, [Replicate()], run_check=False
)
traced_dispatch = _get_dtensor_dispatch_graph(
call_functions[0], node_to_obj, force_make_fx=True
)
assert traced_dispatch is not None
# TODO(anj): This depends on the call function node -> actual DTensor output
# mapping that we want to avoid for SPMD expansion
return traced_dispatch, node_to_obj[call_functions[0]]
def _convert_output(
gm: fx.GraphModule,
node: fx.Node,
node_to_obj: Dict[fx.Node, Any],
) -> fx.Node:
new_args = []
has_partial = False
for argument in node.args[0]: # type: ignore[union-attr]
if not isinstance(argument, fx.Node):
new_args.append(argument)
continue
obj = node_to_obj[argument]
if not _is_partial_dtensor(obj):
new_args.append(argument)
continue
has_partial = True
# we know it's a dtensor from is partial DT check...
dt = cast(DTensor, obj)
traced_dispatch, result_obj = _build_dummy_add_graph(dt, node_to_obj)
wait = [
n
for n in traced_dispatch.graph.nodes
if n.name == "wait_comm" or n.name == "wait_tensor"
]
add = [n for n in traced_dispatch.graph.nodes if n.name == "add"]
assert len(wait) == 1 and len(add) == 1
# remove add node and replace it with wait node
add[0].replace_all_uses_with(wait[0])
traced_dispatch.graph.eliminate_dead_code()
# also update the actual DTensor corresponding to the node
# TODO(anj): We require mapping of the final DTensor output to the wait
# comm node.
node_to_obj[wait[0]] = result_obj
value_remap: Dict[fx.Node, fx.Node] = {}
for dtn in traced_dispatch.graph.nodes:
if dtn.op == OP.PLACEHOLDER:
# do nothing, ignore placeholders, as it has
# already been prepared in value_remap
value_remap[dtn] = argument
elif dtn.op == OP.OUTPUT:
assert (
len(dtn.args) == 1 and len(dtn.args[0]) == 1
), f"Expecting single output, but got {dtn.args} {len(dtn.args)}"
new_args.append(value_remap[dtn.args[0][0]])
# the concrete DTensor value of output was added when creating the
# inner graph (in _build_dummy_add_graph). Just add it to the final
# output node so that we can report the final output specs correctly.
# TODO(anj): We are depending on the concrete DTensor output of the dummy add.
node_to_obj[value_remap[dtn.args[0][0]]] = node_to_obj[dtn.args[0][0]]
else:
if dtn.op == OP.GET_ATTR:
setattr(
gm,
dtn.target,
getattr(traced_dispatch, dtn.target),
)
with gm.graph.inserting_before(node):
value_remap[dtn] = gm.graph.node_copy(dtn, lambda n: value_remap[n])
if has_partial:
gm.graph.erase_node(node)
return gm.graph.output(new_args)
else:
return node
def _rebuild_graph(
gm: fx.GraphModule,
node_replacements: Dict[torch.fx.Node, torch.fx.GraphModule],
) -> None:
# replace nodes in local traced graph with DTensor's dispatch graph
for node in gm.graph.nodes:
if node not in node_replacements:
continue
traced_dispatch = node_replacements[node]
# Map DT's dispatch graph input placeholder nodes to the ones in
# local traced graph. It uses index-based accessing, which is
# brittle, just for testing purpose.
flatten_args, _ = tree_flatten(node.args)
i, value_remap = 0, {}
for dtn in traced_dispatch.graph.nodes:
if dtn.op == OP.PLACEHOLDER:
value_remap[dtn] = flatten_args[i]
i += 1
# insert DT's dispatch graph to traced local graph.
with gm.graph.inserting_before(node):
for dtn in traced_dispatch.graph.nodes:
if dtn.op == OP.PLACEHOLDER:
# do nothing, ignore placeholders, as it has already
# been prepared in value_remap
pass
elif dtn.op == OP.OUTPUT:
assert (
len(dtn.args) == 1
), f"Expecting single output, but got {dtn.args} {len(dtn.args[0])}"
outputs = dtn.args[0]
# we currently support two very specific types of output
# 1. single output
# 2. multiple outputs resulting from getitem of all elements of tuple
if len(outputs) == 1:
# for single output, we replace the node with the single node
output = outputs[0]
else:
# for multiple outputs, we check that these outputs correspond
# to all elements of a tuple. In that case, we replace
# uses of the output directly with the original tuple
source = None
for i, out in enumerate(outputs):
# we allow None outputs for certain items in the tuple
if out is None:
continue
assert out.op == "call_function"
assert out.target.__module__ == "_operator"
assert out.target.__name__ == "getitem"
assert source is None or source == out.args[0]
source = out.args[0]
assert out.args[1] == i
assert source is not None
output = source
new_node = value_remap[output]
node.replace_all_uses_with(new_node)
else:
value_remap[dtn] = gm.graph.node_copy(dtn, lambda n: value_remap[n])
if all(
isinstance(n.target, torch._ops.OpOverload)
and n.target._schema.name.startswith(
("aten::_foreach", "aten::_fused_adam")
)
for n in [dtn, node]
):
# FIXME(@mrshenli): This is a temporary solution enable
# foreach ops. The problem is that foreach ops returns
# List[Tensor], but make_fx will flatten that before
# passing those tensors to output node, which will
# introduce additional getitem nodes. These redundant
# getitem nodes breaks graph correctness as we cannot do
# getitem(getitem(foreach_out, 0), 0). This temporary
# solution skips getitem nodes in DTensor expanded
# subgraphs.
node.replace_all_uses_with(value_remap[dtn])
break
# explicitly erase node instead of relying on DCE, as DCE does not
# remove inplace copy_ correctly.
gm.graph.erase_node(node)
gm.graph.eliminate_dead_code()
gm.recompile()
def _get_last_consumer_to_nodes(
graph: fx.Graph,
) -> Dict[fx.Node, List[fx.Node]]:
# Run through reverse nodes and record the first instance of a use
# of a given node. This represents the *last* use of the node in the
# execution order of the program, which we will use to free unused
# values
node_to_last_consumer: Dict[fx.Node, fx.Node] = {}
last_consumer_to_nodes: Dict[fx.Node, List[fx.Node]] = {}
def _register_final_consumer(arg_node: fx.Node, consumer: fx.Node) -> None:
if arg_node not in node_to_last_consumer:
node_to_last_consumer[arg_node] = consumer
last_consumer_to_nodes.setdefault(consumer, []).append(arg_node)
for node in reversed(graph.nodes):
fx.node.map_arg(
node.args, lambda arg_node: _register_final_consumer(arg_node, node)
)
fx.node.map_arg(
node.kwargs,
lambda kwarg_node: _register_final_consumer(kwarg_node, node),
)
return last_consumer_to_nodes
def _convert_to_distributed(
gm: fx.GraphModule,
inps: List[torch.Tensor],
schemas: List[Schema],
default_mesh: Optional[DeviceMesh] = None,
_allow_partial: bool = False,
) -> Tuple[fx.GraphModule, Dict[str, Schema]]:
"""
Returns:
- transformed graph module
- map from output name to DTensorSpec
"""
global logger
logger = get_logger("spmd_exp")
operators = {getattr(operator, name) for name in operator.__all__}
node_to_obj: Dict[fx.Node, Any] = {}
# map local op node in traced_f to its corresponding subgraph of
# DTensor ops.
node_replacements: Dict[torch.fx.Node, torch.fx.GraphModule] = {}
last_consumer_to_nodes = _get_last_consumer_to_nodes(gm.graph)
output_schemas: Dict[str, Schema] = {}
for i, node in enumerate(gm.graph.nodes):
assert logger is not None
logger.info("node%s: op=%s target=%s", i, node.op, node.target)
if node.op == OP.PLACEHOLDER:
assert i < len(
inps
), f"got more placeholder nodes ({i + 1}) than inputs ({len(inps)})"
# our example inputs are local shards. Create DTensors from them.
node_to_obj[node] = DTensor.from_local(
inps[i].clone(), # use clone to avoid modifications from inplace ops
schemas[i].mesh,
schemas[i].placements,
# prevent running this collective in backwards pass
run_check=False,
)
elif isinstance(node.target, torch._ops.OpOverloadPacket):
dtensor = cast(DTensor, node_to_obj[node.args[0]])
node_to_obj[node] = DSymInt.from_node(node, dtensor)
elif isinstance(node.target, torch._ops.OpOverload):
replacement = _get_dtensor_dispatch_graph(
node, node_to_obj, default_mesh=default_mesh
)
if replacement is not None:
node_replacements[node] = replacement
elif node.op == OP.OUTPUT:
if not _allow_partial:
# Returns an expanded dummy add node that ensures
# that the partial output tensor has been converted
# to a replicated tensor.
node = _convert_output(gm, node, node_to_obj)
# Save output sharding for the inputs to backward pass.
# TODO(anj): Pipe the output schema for the BW pass
# instead of requiring the full output DTensor to be
# materialized.
for inp_arg in node.args[0]:
if isinstance(inp_arg, fx.Node):
obj = node_to_obj[inp_arg]
if isinstance(obj, DTensor):
output_schemas[inp_arg.name] = Schema(
obj.device_mesh, obj.placements # type: ignore[arg-type]
)
elif node.op == OP.CALL_FUNCTION:
args = tree_map(partial(_remap_arg, node_to_obj), node.args)
kwargs = tree_map(partial(_remap_arg, node_to_obj), node.kwargs)
dsymints = list(
filter(lambda a: isinstance(a, DSymInt), args + tuple(kwargs.values()))
)
if node.target in operators and len(dsymints) > 0:
assert all(
dsymints[0].mesh == d.mesh for d in dsymints
), "all DSymInts must have the same mesh. "
local_args = tree_map_only(DSymInt, lambda a: a.local_value, args)
local_kwargs = tree_map_only(DSymInt, lambda a: a.local_value, kwargs)
global_args = tree_map_only(DSymInt, lambda a: a.global_value, args)
global_kwargs = tree_map_only(DSymInt, lambda a: a.global_value, kwargs)
node.args = local_args
node.kwargs = local_kwargs
node_to_obj[node] = DSymInt(
local_value=node.target(*local_args, **local_kwargs),
global_value=node.target(*global_args, **global_kwargs),
mesh=dsymints[0].mesh,
)
else:
assert len(dsymints) == 0, (
"SPMD expansion does not support SymInt in non-operator "
f"nodes, got {node.target}."
)
node_to_obj[node] = node.target(*args, **kwargs)
else:
raise ValueError(f"Unrecognized node.op type {node.op}")
if node in last_consumer_to_nodes:
# Save memory by deleting objs that wont be used anymore.
for arg_node in last_consumer_to_nodes[node]:
del node_to_obj[arg_node]
_rebuild_graph(gm, node_replacements)
return gm, output_schemas
class _SPMD:
def __init__(
self,
dist_graph: DistributedGraph,
param_schema: Schema,
input_schemas: Sequence[Placement],
) -> None:
self._dist_graph = dist_graph
self._param_schema = param_schema
# Override the default sharding of input to the model.
self._input_schemas = input_schemas
# used to propagate sharding from the output of the forward pass to
# the input of backward pass
self._known_specs_by_node_name: Dict[str, Schema] = {}
def _is_param(self, t: torch.Tensor) -> bool:
# N.B.: id(t) and id(param) does not match
orig_module = cast(nn.Module, self._dist_graph.orig_module)
return t.data_ptr() in (p.data_ptr() for p in orig_module.parameters())
def _compile_wrapper(
self,
training_phase: TrainingPhase,
original_inputs: List[List[torch.Tensor]],
gm: fx.GraphModule,
inps: List[torch.Tensor],
) -> fx.GraphModule:
with maybe_disable_fake_tensor_mode():
return self._compile(training_phase, gm, original_inputs[0])
def _compile(
self,
training_phase: TrainingPhase,
gm: fx.GraphModule,
inps: List[torch.Tensor],
) -> fx.GraphModule:
shard_schema: Schema = Schema(
mesh=self._param_schema.mesh, placements=[Shard(0)]
)
schemas: List[Schema] = []
inp_schema_count = 0
nparams = 0
# iterate through inputs (and initial nodes of the graph that should
# correspond 1:1 to those inputs)
for inp, placeholder_node in zip(inps, gm.graph.nodes):
# This is a no-op but we want the order of schemas
# to match the order of inputs when we iterate through
# the graph. Usually the non-tensor inputs are at the
# end of the list so we could drop the schemas for it.
assert placeholder_node.op == "placeholder", (
"Expected initial nodes of the GraphModule to be input placeholders. "
"Got {placeholder_node.op}"
)
known_schema = self._known_specs_by_node_name.get(placeholder_node.name)
if known_schema is not None:
schemas.append(known_schema)
elif not isinstance(inp, torch.Tensor):
schemas.append(
Schema(mesh=self._param_schema.mesh, placements=[Replicate()])
)
else:
if self._is_param(inp):
schemas.append(self._param_schema)
nparams += 1
elif self._input_schemas:
schemas.append(self._input_schemas[inp_schema_count]) # type: ignore[arg-type]
inp_schema_count += 1
else:
schemas.append(shard_schema)
parallelized_gm, output_specs = _convert_to_distributed(
gm,
inps,
schemas,
_allow_partial=False,
)
self._known_specs_by_node_name.update(output_specs)
if training_phase == TrainingPhase.FORWARD:
self._dist_graph.fwd_graph_modules.append(parallelized_gm)
elif training_phase == TrainingPhase.BACKWARD:
self._dist_graph.bwd_graph_modules.append(parallelized_gm)
return make_boxed_func(parallelized_gm)
def distribute(
dist_graph: DistributedGraph,
param_schema: Schema,
input_schemas: Sequence[Placement],
*args: Tuple[Any, ...],
**kwargs: Dict[str, Any],
) -> nn.Module:
flat_args, _ = tree_flatten(args)
flat_kwargs, _ = tree_flatten(kwargs)
input_set: Set[Any] = set(flat_args + flat_kwargs)
fake_mode: FakeTensorMode = detect_fake_mode(input_set)
if fake_mode is None:
fake_mode = FakeTensorMode()
# will update this to the original forward inputs
original_inputs: List[Optional[Sequence[Any]]] = [None]
def input_to_fake(input: Any) -> Any:
if not isinstance(input, torch.Tensor):
return input
y = fake_mode.from_tensor(input)
if input in input_set:
# "unshard" our fake tensor
# (considers that inputs are sharded)
y = y.repeat(param_schema.mesh.size(0), *((1,) * (y.ndim - 1)))
# TODO assume non-inputs (params, etc) are replicated for now.
return y
def gather_inputs_for_compilation(
inps: Tuple[Any, ...],
) -> Tuple[Any, ...]:
original_inputs[0] = inps
return tuple(input_to_fake(x) for x in inps)
spmd = _SPMD(dist_graph, param_schema, input_schemas)
compiled_m = aot_module(
cast(nn.Module, dist_graph.orig_module),
partial(spmd._compile_wrapper, TrainingPhase.FORWARD, original_inputs),
partial(spmd._compile, TrainingPhase.BACKWARD),
pre_compile_fn=gather_inputs_for_compilation,
decompositions=_CURRENT_DECOMPOSITION_TABLE,
)
return compiled_m
Reference in New Issue View Git Blame Copy Permalink