import base64
import dataclasses
import io
import json
import logging
import math
import operator
import typing
from contextlib import contextmanager
from dataclasses import dataclass, field
from enum import Enum
from typing import Any, Callable, cast, Dict, Iterator, List, Optional, Union
import sympy
import torch
import torch.export.exported_program as ep
from torch._export.verifier import load_verifier
from torch._subclasses.fake_tensor import FakeTensor, FakeTensorMode
from torch.fx.experimental import symbolic_shapes
from torch.utils._pytree import treespec_dumps, treespec_loads
from torch.utils._sympy.value_ranges import ValueRanges
from .schema import ( # type: ignore[attr-defined]
_Union,
Argument,
BufferMutationSpec,
CustomObjArgument,
Device,
ExportedProgram,
GradientToParameterSpec,
GradientToUserInputSpec,
Graph,
GraphArgument,
GraphModule,
GraphSignature,
InputSpec,
InputToBufferSpec,
InputToParameterSpec,
InputToTensorConstantSpec,
Layout,
LossOutputSpec,
MemoryFormat,
ModuleCallEntry,
ModuleCallSignature,
NamedArgument,
Node,
OptionalTensorArgument,
OutputSpec,
RangeConstraint,
ScalarType,
SCHEMA_VERSION,
SymBool,
SymBoolArgument,
SymExpr,
SymExprHint,
SymInt,
SymIntArgument,
TensorArgument,
TensorMeta,
TREESPEC_VERSION,
UserInputSpec,
UserOutputSpec,
)
__all__ = [
"serialize",
"GraphModuleSerializer",
"ExportedProgramSerializer",
"GraphModuleDeserializer",
"ExportedProgramDeserializer",
]
from torch.export.exported_program import (
ConstantArgument as PyConstantArgument,
SymIntArgument as PySymIntArgument,
TensorArgument as PyTensorArgument,
)
from .upgrade import GraphModuleOpUpgrader
log = logging.getLogger(__name__)
class SerializeError(RuntimeError):
pass
def _reverse_map(d: Dict[Any, Enum]):
return {v.value: k for k, v in d.items()}
MetaType = Union[FakeTensor, int, torch.SymInt, bool, torch.SymBool]
ST_DELIMITER = ";"
_TORCH_TO_SERIALIZE_DTYPE = {
torch.uint8: ScalarType.BYTE,
torch.int8: ScalarType.CHAR,
torch.int16: ScalarType.SHORT,
torch.int32: ScalarType.INT,
torch.int64: ScalarType.LONG,
torch.float16: ScalarType.HALF,
torch.float32: ScalarType.FLOAT,
torch.float64: ScalarType.DOUBLE,
torch.complex32: ScalarType.COMPLEXHALF,
torch.complex64: ScalarType.COMPLEXFLOAT,
torch.complex128: ScalarType.COMPLEXDOUBLE,
torch.bool: ScalarType.BOOL,
torch.bfloat16: ScalarType.BFLOAT16
}
_SERIALIZE_TO_TORCH_DTYPE = _reverse_map(_TORCH_TO_SERIALIZE_DTYPE) # type: ignore[arg-type]
_TORCH_TO_SERIALIZE_LAYOUT = {
torch.sparse_coo: Layout.SparseCoo,
torch.sparse_csr: Layout.SparseCsr,
torch.sparse_csc: Layout.SparseCsc,
torch.sparse_bsr: Layout.SparseBsr,
torch.sparse_bsc: Layout.SparseBsc,
torch._mkldnn: Layout._mkldnn, # type: ignore[attr-defined]
torch.strided: Layout.Strided,
}
_SERIALIZE_TO_TORCH_LAYOUT = _reverse_map(_TORCH_TO_SERIALIZE_LAYOUT) # type: ignore[arg-type]
_TORCH_TO_SERIALIZE_MEMORY_FORMAT = {
torch.contiguous_format: MemoryFormat.ContiguousFormat,
torch.channels_last: MemoryFormat.ChannelsLast,
torch.channels_last_3d: MemoryFormat.ChannelsLast3d,
torch.preserve_format: MemoryFormat.PreserveFormat,
}
_SERIALIZE_TO_TORCH_MEMORY_FORMAT = _reverse_map(_TORCH_TO_SERIALIZE_MEMORY_FORMAT) # type: ignore[arg-type]
_SYM_INT_OPS = {
operator.mul,
operator.add,
operator.sub,
operator.floordiv,
operator.mod,
}
_SYM_BOOL_OPS = {
operator.eq,
operator.ne,
operator.le,
operator.ge,
operator.lt,
operator.gt,
}
@dataclass
class SerializedArtifact:
exported_program: Union[ExportedProgram, bytes]
state_dict: bytes
constants: bytes
def deserialize_device(d: Device) -> torch.device:
if d.index is None:
return torch.device(type=d.type) # type: ignore[call-overload]
return torch.device(type=d.type, index=d.index)
def serialize_sym_int(s: Union[int, torch.SymInt]) -> SymInt:
if isinstance(s, (torch.SymInt, int)):
if symbolic_shapes.is_concrete_int(s):
return SymInt.create(as_int=int(s))
else:
assert isinstance(s, torch.SymInt)
if s.node.hint is None:
return SymInt.create(as_expr=SymExpr(str(s)))
else:
return SymInt.create(as_expr=SymExpr(str(s), hint=SymExprHint.create(as_int=s.node.hint)))
else:
raise SerializeError(
f"SymInt should be either symbol or int, got `{s}` of type `{type(s)}`"
)
def serialize_sym_bool(s: Union[bool, torch.SymBool]) -> SymBool:
if isinstance(s, (torch.SymBool, bool)):
if symbolic_shapes.is_concrete_bool(s):
return SymBool.create(as_bool=bool(s))
else:
return SymBool.create(as_expr=SymExpr(expr_str=str(s)))
else:
raise SerializeError(
f"SymBool should be either symbol or bool, got `{s}` of type `{type(s)}`"
)
def serialize_tensor_meta(t: torch.Tensor) -> TensorMeta:
"""
Extract a TensorMeta describing `t`.
"""
return TensorMeta(
dtype=_TORCH_TO_SERIALIZE_DTYPE[t.dtype],
sizes=[serialize_sym_int(s) for s in t.shape],
requires_grad=t.requires_grad,
device=Device(type=t.device.type, index=t.device.index),
strides=[serialize_sym_int(s) for s in t.stride()],
storage_offset=0,
layout=_TORCH_TO_SERIALIZE_LAYOUT[t.layout],
)
def serialize_torch_artifact(artifact) -> bytes:
buffer = io.BytesIO()
# This is a workaround for backend's tensor deserialization problem:
# unpickleTensor() always create a tensor on the device where it was originally saved
# This behavior is bad for multi-gpu training, as we wish to directly load the tensor
# on the designated device.
# For now, we simply move the tensor to cpu before saving.
# TODO: this should be fixed by deserialization instead.
torch.save(artifact, buffer)
return buffer.getvalue()
def deserialize_torch_artifact(serialized: bytes):
if len(serialized) == 0:
return {}
buffer = io.BytesIO(serialized)
buffer.seek(0)
return torch.load(buffer)
def _sympy_int_to_int(val: sympy.Expr):
# Convert simple sympy Integers into concrete int
if val == sympy.oo:
return math.inf
if val == -sympy.oo:
return -math.inf
if isinstance(val, sympy.Integer):
return int(val)
raise RuntimeError(
"Export constraints cannot be non-integer expressions"
)
def _int_to_sympy_int(val) -> sympy.Expr:
# Convert concrete int into simple sympy Integers
if val == math.inf:
return sympy.oo
if val == -math.inf:
return -sympy.oo
return sympy.Integer(val)
def serialize_range_constraints(
range_constraints: Dict[sympy.Symbol, ValueRanges]
) -> Dict[str, RangeConstraint]:
return {
str(k): RangeConstraint(
_sympy_int_to_int(v.lower), # type: ignore[arg-type]
_sympy_int_to_int(v.upper), # type: ignore[arg-type]
)
for k, v in range_constraints.items()
}
def _is_single_tensor_return(target: torch._ops.OpOverload) -> bool:
returns = target._schema.returns
return len(returns) == 1 and isinstance(returns[0].real_type, torch.TensorType)
def _is_single_tensor_list_return(target: torch._ops.OpOverload) -> bool:
returns = target._schema.returns
if len(returns) != 1:
return False
return_type = returns[0].real_type
return isinstance(return_type, torch.ListType) and isinstance(
return_type.getElementType(), torch.TensorType
)
@dataclass
class GraphState:
inputs: List[Argument] = field(default_factory=list)
outputs: List[Argument] = field(default_factory=list)
nodes: List[Node] = field(default_factory=list)
tensor_values: Dict[str, TensorMeta] = field(default_factory=dict)
sym_int_values: Dict[str, SymInt] = field(default_factory=dict)
sym_bool_values: Dict[str, SymBool] = field(default_factory=dict)
is_single_tensor_return: bool = False
class GraphModuleSerializer:
def __init__(
self,
graph_signature: ep.ExportGraphSignature,
module_call_graph: List[ep.ModuleCallEntry]
):
self.graph_state = GraphState()
self.graph_signature = graph_signature
self.module_call_graph = module_call_graph
self.custom_objs: Dict[str, torch._C.ScriptObject] = {}
@contextmanager
def save_graph_state(self):
saved = self.graph_state
self.graph_state = GraphState()
try:
yield
finally:
self.graph_state = saved
def handle_placeholder(self, node: torch.fx.Node):
assert node.op == "placeholder"
if isinstance(node.meta['val'], torch.Tensor):
graph_input = Argument.create(as_tensor=TensorArgument(name=node.name))
self.graph_state.tensor_values[node.name] = serialize_tensor_meta(node.meta["val"])
elif isinstance(node.meta['val'], torch.SymInt):
raise AssertionError("SymInt graph input is not implemented yet.")
elif isinstance(node.meta['val'], (int, bool, str, float, type(None))):
graph_input = self.serialize_input(node.meta['val'])
else:
raise AssertionError(f"Unimplemented graph input type: {node.meta['val']}")
self.graph_state.inputs.append(graph_input)
def handle_output(self, node: torch.fx.Node):
assert node.op == "output"
assert len(node.args) == 1, "FX.Node's args should have one arg"
node_args = node.args[0]
if isinstance(node_args, torch.fx.Node):
# For singleton tensor returns
self.graph_state.is_single_tensor_return = True
self.graph_state.outputs = [self.serialize_input(node_args)]
else:
assert isinstance(node_args, (tuple, list))
self.graph_state.outputs = [self.serialize_input(arg) for arg in node_args]
def serialize_operator(self, target) -> str:
if isinstance(target, str):
return target
elif target.__module__.startswith("torch._ops"):
# TODO(zhxchen17) Maybe provide a function name helper in FX.
# From torch.fx.node._get_qualified_name
module = target.__module__.replace("torch._ops", "torch.ops")
return f"{module}.{target.__name__}"
else: # TODO(zhxchen17) Don't catch all here.
return f"{target.__module__}.{target.__name__}"
def handle_call_function(self, node: torch.fx.Node):
assert node.op == "call_function"
# getitem has been handled in the producer node, skip it here
if node.target is operator.getitem:
return
if node.target in _SYM_INT_OPS:
assert len(node.kwargs) == 0
meta_val = node.meta["val"]
ex_node = Node(
target=self.serialize_operator(node.target),
inputs=self.serialize_sym_op_inputs(node.args),
outputs=[Argument.create(as_sym_int=self.serialize_sym_int_output(node.name, meta_val))],
metadata=self.serialize_metadata(node),
)
elif node.target in _SYM_BOOL_OPS:
assert len(node.kwargs) == 0
meta_val = node.meta["val"]
ex_node = Node(
target=self.serialize_operator(node.target),
inputs=self.serialize_sym_op_inputs(node.args),
outputs=[Argument.create(as_sym_bool=self.serialize_sym_bool_output(node.name, meta_val))],
metadata=self.serialize_metadata(node),
)
elif isinstance(node.target, torch._ops.OpOverload):
ex_node = Node(
target=self.serialize_operator(node.target),
inputs=self.serialize_inputs(node.target, node.args, node.kwargs),
outputs=self.serialize_outputs(node),
# TODO: create a new tensor_values here, meta might have faketensor info
metadata=self.serialize_metadata(node),
)
elif isinstance(node.target, torch._ops.HigherOrderOperator):
inputs = [
NamedArgument(
name="", # TODO(zhxchen17) This is sad, should be improved when HOO has schema arg names.
arg=self.serialize_input(a),
) for a in node.args
]
meta_val = node.meta["val"]
if isinstance(meta_val, torch.Tensor):
outputs = [Argument.create(as_tensor=self.serialize_tensor_output(node.name, meta_val))]
elif isinstance(meta_val, (list, tuple)) and all(isinstance(v, torch.Tensor) for v in meta_val):
arg_list = self._handle_getitem_users(node)
outputs = [Argument.create(as_tensors=arg_list)]
else:
raise SerializeError(
"Only single tensor output or list of tensor output "
"is supported for HigherOrderOperator serialization"
)
ex_node = Node(
target=self.serialize_operator(node.target),
inputs=inputs,
outputs=outputs,
metadata=self.serialize_metadata(node),
)
else:
raise SerializeError(f"Serializing {node.target} is not supported")
self.graph_state.nodes.append(ex_node)
def handle_get_attr(self, node):
pass
def serialize_metadata(self, node: torch.fx.Node) -> Dict[str, str]:
ret = {}
if stack_trace := node.meta.get("stack_trace"):
ret["stack_trace"] = stack_trace
if nn_module_stack := node.meta.get("nn_module_stack"):
def export_nn_module_stack(val):
assert isinstance(val, tuple) and len(val) == 2
path, ty = val
assert isinstance(path, str)
normalized_ty = ty.__module__ + "." + ty.__qualname__
return path + "," + normalized_ty
# Serialize to "key,orig_path,type_str"
nn_module_list = [
f"{k},{export_nn_module_stack(v)}"
for k, v in nn_module_stack.items()
]
ret["nn_module_stack"] = ST_DELIMITER.join(nn_module_list)
if source_fn_st := node.meta.get("source_fn_stack"):
source_fn_list = [f"{source_fn[0]},{self.serialize_operator(source_fn[1])}" for source_fn in source_fn_st]
ret["source_fn_stack"] = ST_DELIMITER.join(source_fn_list)
return ret
def serialize_sym_op_inputs(self, args) -> List[NamedArgument]:
serialized_args = []
args_names = ["a", "b"]
for args_name, arg in zip(args_names, args):
serialized_args.append(
NamedArgument(name=args_name, arg=self.serialize_input(arg))
)
return serialized_args
def serialize_inputs(
self, target: torch._ops.OpOverload, args, kwargs=None
) -> List[NamedArgument]:
assert isinstance(target, torch._ops.OpOverload)
kwargs = kwargs or {}
serialized_args = []
for i, schema_arg in enumerate(target._schema.arguments):
if schema_arg.name in kwargs:
serialized_args.append(
NamedArgument(
name=schema_arg.name,
arg=self.serialize_input(kwargs[schema_arg.name]),
)
)
elif not schema_arg.kwarg_only and i < len(args):
serialized_args.append(
NamedArgument(
name=schema_arg.name,
arg=self.serialize_input(args[i]),
)
)
else:
# We intentionally don't serialize the missing arguments
# with default values
pass
return serialized_args
def is_sym_int_arg(self, arg) -> bool:
return isinstance(arg, int) or (
isinstance(arg, torch.fx.Node) and arg.name in self.graph_state.sym_int_values
)
def is_sym_bool_arg(self, arg) -> bool:
return isinstance(arg, bool) or (
isinstance(arg, torch.fx.Node) and arg.name in self.graph_state.sym_bool_values
)
def serialize_input(self, arg) -> Argument:
import torch._inductor.ir as inductor_ir
inductor_tensor_buffers = (
inductor_ir.Buffer,
inductor_ir.ReinterpretView,
)
if isinstance(arg, torch.fx.Node):
if arg.op == "get_attr":
assert isinstance(arg.target, str)
attr = getattr(arg.graph.owning_module, arg.target)
if isinstance(attr, torch.Tensor):
raise SerializeError("getattr nodes containing tensors should not appear in the graph")
elif isinstance(attr, torch.fx.GraphModule):
with self.save_graph_state():
graph = self.serialize_graph(attr)
return Argument.create(as_graph=GraphArgument(name=arg.target, graph=graph))
else:
raise SerializeError(f"Unsupported getattr attribute {arg.target} with type: {type(attr)}")
elif self.is_sym_int_arg(arg):
return Argument.create(as_sym_int=SymIntArgument.create(as_name=arg.name))
elif self.is_sym_bool_arg(arg):
return Argument.create(as_sym_bool=SymBoolArgument.create(as_name=arg.name))
else:
return Argument.create(as_tensor=TensorArgument(name=arg.name))
elif isinstance(arg, inductor_tensor_buffers):
# Other branches are for arguments in fx node.
# This is a special branch for handling buffers (representing tensor arguments)
# for inductor's ExternalFallbackNode
# export_extern_kernel_node() is using this function to serialize arguments
arg_name = arg.get_name()
assert arg_name is not None, "Buffer must have valid name"
return Argument.create(as_tensor=TensorArgument(name=arg_name))
elif isinstance(arg, torch.SymInt):
# This is a special branch for handling SymInt args in inductor's
# ExternalFallbackNode.
# For regular FX graph, SymInt arg should be a fx.Node with
# self.is_sym_int_arg(arg) being true
return Argument.create(as_sym_int=SymIntArgument.create(as_name=str(arg)))
elif isinstance(arg, bool):
return Argument.create(as_bool=arg)
elif isinstance(arg, str):
return Argument.create(as_string=arg)
elif isinstance(arg, int):
return Argument.create(as_int=arg)
elif isinstance(arg, float):
return Argument.create(as_float=arg)
elif arg is None:
return Argument.create(as_none=())
elif isinstance(arg, (list, tuple)):
# Must check bool first, as bool is also treated as int
if all(isinstance(a, bool) for a in arg):
return Argument.create(as_bools=list(arg))
elif all(isinstance(a, int) for a in arg):
return Argument.create(as_ints=list(arg))
elif all(isinstance(a, float) for a in arg):
return Argument.create(as_floats=list(arg))
elif all(isinstance(a, str) for a in arg):
return Argument.create(as_strings=list(arg))
elif all(isinstance(a, torch.SymInt) for a in arg):
# This is a special branch for handling SymInt args in inductor's
# ExternalFallbackNode.
# For regular FX graph, SymInt arg should be a fx.Node with
# self.is_sym_int_arg(arg) being true
return Argument.create(
as_sym_ints=[SymIntArgument.create(as_name=str(a)) for a in arg]
)
elif all(self.is_sym_int_arg(a) for a in arg):
# list of sym_ints
values = []
for a in arg:
if isinstance(a, torch.fx.Node):
values.append(SymIntArgument.create(as_name=a.name))
elif isinstance(a, int):
values.append(SymIntArgument.create(as_int=a))
return Argument.create(as_sym_ints=values)
elif all(self.is_sym_bool_arg(a) for a in arg):
# list of sym_bools
values = []
for a in arg:
if isinstance(a, torch.fx.Node):
values.append(SymBoolArgument.create(as_name=a.name))
elif isinstance(a, bool):
values.append(SymBoolArgument.create(as_bool=a))
return Argument.create(as_sym_bools=values)
elif all(isinstance(a, torch.fx.Node) for a in arg):
# list of tensors
arguments = []
for a in arg:
if a.op == "get_attr":
raise SerializeError("getattr nodes containing tensors should not appear in the graph")
arguments.append(TensorArgument(name=a.name))
return Argument.create(as_tensors=arguments)
elif all(isinstance(a, (torch.fx.Node, type(None))) for a in arg):
# list of optional tensors
def serialize_optional_tensor_args(a):
if a is None:
return OptionalTensorArgument.create(as_none=())
elif isinstance(a, torch.fx.Node):
return OptionalTensorArgument.create(as_tensor=a.name)
else:
raise SerializeError(f"Unsupported list/tuple argument: {a}")
return Argument.create(
as_optional_tensors=list(map(serialize_optional_tensor_args, arg))
)
elif all(isinstance(a, inductor_tensor_buffers) for a in arg):
# list of inductor buffers
return Argument.create(
as_tensors=[TensorArgument(name=a.get_name()) for a in arg],
)
elif all(isinstance(a, (*inductor_tensor_buffers, type(None))) for a in arg):
# list of inductor buffers as optional tensors
def serialize_optional_tensor_args(a):
if a is None:
return OptionalTensorArgument.create(as_none=())
elif isinstance(a, inductor_tensor_buffers):
return OptionalTensorArgument.create(as_tensor=a.get_name())
else:
raise SerializeError(f"Unsupported list/tuple argument: {a}")
return Argument.create(
as_optional_tensors=list(map(serialize_optional_tensor_args, arg))
)
else:
raise SerializeError(f"Unsupported list/tuple argument type: {[type(a) for a in arg]}")
elif isinstance(arg, torch.dtype):
return Argument.create(as_scalar_type=_TORCH_TO_SERIALIZE_DTYPE[arg])
elif isinstance(arg, torch.device):
return Argument.create(as_device=Device(type=arg.type, index=arg.index))
elif isinstance(arg, torch.memory_format):
return Argument.create(as_memory_format=_TORCH_TO_SERIALIZE_MEMORY_FORMAT[arg])
elif isinstance(arg, torch.layout):
return Argument.create(as_layout=_TORCH_TO_SERIALIZE_LAYOUT[arg])
elif isinstance(arg, torch._C.ScriptObject):
if not (
arg._has_method("__getstate__") and # type: ignore[attr-defined]
arg._has_method("__setstate__") # type: ignore[attr-defined]
):
raise SerializeError(
f"Unable to serialize custom class {arg}. Please define "
"serialization methods via def_pickle()."
)
# Custom objects through torchind are serializable with pickle,
# through implementing the .def_pickle function. This should result
# in the object containing a __getstate__ and __setstate__
# serialize/deserialize function.
custom_obj_name = f"_custom_obj_{len(self.custom_objs)}"
self.custom_objs[custom_obj_name] = arg
return Argument.create(as_custom_obj=CustomObjArgument(custom_obj_name))
else:
raise SerializeError(f"Unsupported argument type: {type(arg)}")
def serialize_tensor_output(self, name, meta_val) -> TensorArgument:
assert name not in self.graph_state.tensor_values
self.graph_state.tensor_values[name] = serialize_tensor_meta(meta_val)
return TensorArgument(name=name)
def serialize_sym_int_output(self, name, meta_val) -> SymIntArgument:
assert name not in self.graph_state.sym_int_values
self.graph_state.sym_int_values[name] = serialize_sym_int(meta_val)
return SymIntArgument.create(as_name=name)
def serialize_sym_bool_output(self, name, meta_val) -> SymIntArgument:
assert name not in self.graph_state.sym_bool_values
self.graph_state.sym_bool_values[name] = serialize_sym_bool(meta_val)
return SymBoolArgument.create(as_name=name)
def serialize_input_spec(self, spec: ep.InputSpec) -> InputSpec:
if spec.kind == ep.InputKind.USER_INPUT:
return InputSpec.create(
user_input=UserInputSpec(
arg=self.serialize_argument_spec(spec.arg)
)
)
elif spec.kind == ep.InputKind.PARAMETER:
assert spec.target is not None
assert isinstance(spec.arg, ep.TensorArgument)
return InputSpec.create(
parameter=InputToParameterSpec(
arg=TensorArgument(name=spec.arg.name),
parameter_name=spec.target,
)
)
elif spec.kind == ep.InputKind.BUFFER:
assert spec.target is not None
assert isinstance(spec.arg, ep.TensorArgument)
return InputSpec.create(
buffer=InputToBufferSpec(
arg=TensorArgument(name=spec.arg.name),
buffer_name=spec.target,
)
)
elif spec.kind == ep.InputKind.CONSTANT_TENSOR:
assert spec.target is not None
assert isinstance(spec.arg, ep.TensorArgument)
return InputSpec.create(
tensor_constant=InputToTensorConstantSpec(
arg=TensorArgument(name=spec.arg.name),
tensor_constant_name=spec.target,
)
)
else:
raise AssertionError(f"Unknown argument kind: {spec}")
def serialize_output_spec(self, spec: ep.OutputSpec) -> OutputSpec:
if spec.kind == ep.OutputKind.USER_OUTPUT:
return OutputSpec.create(
user_output=UserOutputSpec(
arg=self.serialize_argument_spec(spec.arg)
)
)
elif spec.kind == ep.OutputKind.LOSS_OUTPUT:
assert isinstance(spec.arg, ep.TensorArgument)
return OutputSpec.create(
loss_output=LossOutputSpec(
arg=TensorArgument(name=spec.arg.name)
)
)
elif spec.kind == ep.OutputKind.BUFFER_MUTATION:
assert spec.target is not None
assert isinstance(spec.arg, PyTensorArgument)
return OutputSpec.create(
buffer_mutation=BufferMutationSpec(
arg=TensorArgument(name=spec.arg.name),
buffer_name=spec.target,
)
)
elif spec.kind == ep.OutputKind.GRADIENT_TO_PARAMETER:
assert spec.target is not None
assert isinstance(spec.arg, PyTensorArgument)
return OutputSpec.create(
gradient_to_parameter=GradientToParameterSpec(
arg=TensorArgument(name=spec.arg.name),
parameter_name=spec.target,
)
)
elif spec.kind == ep.OutputKind.GRADIENT_TO_USER_INPUT:
assert spec.target is not None
assert isinstance(spec.arg, PyTensorArgument)
return OutputSpec.create(
gradient_to_user_input=GradientToUserInputSpec(
arg=TensorArgument(name=spec.arg.name),
user_input_name=spec.target,
)
)
else:
raise AssertionError(f"Unknown argument kind: {spec}")
def serialize_signature(self, sig: ep.ExportGraphSignature) -> GraphSignature:
return GraphSignature(
input_specs=[self.serialize_input_spec(s) for s in sig.input_specs],
output_specs=[self.serialize_output_spec(s) for s in sig.output_specs],
)
def serialize_argument_spec(self, x: ep.ArgumentSpec) -> Argument:
if isinstance(x, PyTensorArgument):
return Argument.create(as_tensor=TensorArgument(name=x.name))
elif isinstance(x, PySymIntArgument):
return Argument.create(as_sym_int=SymIntArgument.create(as_name=x.name))
elif isinstance(x, PyConstantArgument):
return self.serialize_input(x.value)
else:
raise AssertionError("TODO")
def serialize_module_call_signature(self, module_call_signature: ep.ModuleCallSignature) -> ModuleCallSignature:
return ModuleCallSignature(
inputs=[self.serialize_argument_spec(x) for x in module_call_signature.inputs],
outputs=[self.serialize_argument_spec(x) for x in module_call_signature.outputs],
in_spec=treespec_dumps(module_call_signature.in_spec, TREESPEC_VERSION),
out_spec=treespec_dumps(module_call_signature.out_spec, TREESPEC_VERSION),
)
def serialize_module_call_graph(self, module_call_graph: List[ep.ModuleCallEntry]) -> List[ModuleCallEntry]:
return [
ModuleCallEntry(
fqn=entry.fqn,
signature=self.serialize_module_call_signature(entry.signature) if entry.signature else None,
) for entry in module_call_graph
]
def serialize_outputs(self, node: torch.fx.Node) -> List[Argument]:
"""For a given node, return the dataclass representing its output values.
[NOTE: Multiple outputs] We handle aggregates differently than FX. For
FX, it looks like:
x = call_function("multiple_return", ...)
element0 = call_function(getitem, x, 0)
foo = call_function("use_output", element0)
We do not want the intermediate `getitem` call, so our serialized thing looks like:
element0, element1, element2 = call_function("multiple_return", ...)
foo = call_function("use_output", element0)
We want names to be consistent across these two schemes, so that we can
mostly reuse the names coming from FX. This function computes a mapping from
the FX representation to our representation, preserving the names.
"""
assert node.op == "call_function" and isinstance(node.target, torch._ops.OpOverload)
assert isinstance(node.target, torch._ops.OpOverload)
returns = node.target._schema.returns
if len(returns) == 0:
return []
meta_val = node.meta["val"]
def output_node_at_index(node, index):
for user in node.users:
assert user.target is operator.getitem, f"{user} is not a getitem node"
if index == user.args[1]:
return user
return None
# Check single value return
if _is_single_tensor_return(node.target):
# e.g "-> Tensor"
return [Argument.create(as_tensor=self.serialize_tensor_output(node.name, meta_val))]
elif len(returns) == 1 and isinstance(meta_val, torch.SymInt):
# e.g "-> SymInt"
return [Argument.create(as_sym_int=self.serialize_sym_int_output(node.name, meta_val))]
elif len(returns) == 1 and isinstance(meta_val, torch.SymBool):
# e.g "-> SymBool"
return [Argument.create(as_sym_bool=self.serialize_sym_bool_output(node.name, meta_val))]
elif _is_single_tensor_list_return(node.target):
# e.g "-> Tensor[]"
tensor_args = []
for idx, meta in enumerate(meta_val):
user_node = output_node_at_index(node, idx)
name = (
user_node.name
if user_node is not None
else f"{node.name}_unused_{idx}"
)
tensor_args.append(self.serialize_tensor_output(name, meta))
return [Argument.create(as_tensors=tensor_args)]
# There are a two possibilities at this point:
# - This operator returns a tuple of Tensors, e.g. "-> (Tensor, Tensor)"
# - This operator returns a tuple of mixed of Tensor and Tensors, e.g. "-> (Tensor, Tensor[])"
#
# Either way, start by gathering a list of TensorArguments with the correct names.
# For consistent naming with FX, consult the downstream `getitem` node and
# make sure our outputs have the same name.
output_arguments = []
for idx, (meta, return_schema) in enumerate(zip(meta_val, returns)):
if meta is None:
assert isinstance(return_schema.real_type, torch.OptionalType)
output_arguments.append(Argument.create(as_none=()))
elif isinstance(meta, torch._subclasses.fake_tensor.FakeTensor):
assert isinstance(return_schema.real_type, torch.TensorType)
user_node = output_node_at_index(node, idx)
name = (
user_node.name
if user_node is not None
else f"{node.name}_unused_{idx}"
)
output_arguments.append(
Argument.create(as_tensor=self.serialize_tensor_output(name, meta))
)
elif isinstance(meta, list):
# for List[Tensor] return type
assert isinstance(
return_schema.real_type, torch.ListType
) and isinstance(
return_schema.real_type.getElementType(), torch.TensorType
)
user_node = output_node_at_index(node, idx)
assert user_node is not None
args = []
for i, m in enumerate(meta):
if m is None:
continue
sub_user_node = output_node_at_index(user_node, i)
assert sub_user_node is not None, f"No user found at index {i}"
args.append(self.serialize_tensor_output(sub_user_node.name, m))
output_arguments.append(Argument.create(as_tensors=args))
return output_arguments
def _handle_getitem_users(self, node: torch.fx.Node) -> List[TensorArgument]:
meta_val = node.meta["val"]
idx_to_name = {}
for user in node.users:
assert user.target is operator.getitem, f"User node {user} of {node} is incorrect"
idx_to_name[user.args[1]] = user.name
for idx, _ in enumerate(meta_val):
# FX does not emit a getitem node for any outputs that are unused.
# However, we need a name for them so that the number of outputs will
# correctly match the schema. Just assign a dummy name.
if idx not in idx_to_name:
idx_to_name[idx] = f"{node.name}_unused_{idx}"
arg_list = []
for i, element_meta_val in enumerate(meta_val):
arg_list.append(
self.serialize_tensor_output(idx_to_name[i], element_meta_val)
)
return arg_list
def serialize_graph(self, graph_module: torch.fx.GraphModule) -> Graph:
assert isinstance(graph_module, torch.fx.GraphModule)
for node in graph_module.graph.nodes:
try:
getattr(self, f"handle_{node.op}")(node)
except Exception as e:
raise SerializeError(f"Failed serializing node {node} in graph: {node.format_node()}") from e
return Graph(
inputs=self.graph_state.inputs,
nodes=self.graph_state.nodes,
tensor_values=self.graph_state.tensor_values,
sym_int_values=self.graph_state.sym_int_values,
sym_bool_values=self.graph_state.sym_bool_values,
outputs=self.graph_state.outputs,
is_single_tensor_return=self.graph_state.is_single_tensor_return,
)
def serialize(self, graph_module: torch.fx.GraphModule) -> GraphModule:
graph = self.serialize_graph(graph_module)
return GraphModule(
graph=graph,
signature=self.serialize_signature(self.graph_signature),
module_call_graph=self.serialize_module_call_graph(self.module_call_graph),
)
class ExportedProgramSerializer:
def __init__(self, opset_version: Optional[Dict[str, int]] = None):
self.opset_version: Dict[str, int] = {}
if opset_version:
self.opset_version.update(opset_version)
if "aten" not in self.opset_version:
self.opset_version["aten"] = torch._C._get_max_operator_version()
def serialize(self, exported_program: ep.ExportedProgram) -> SerializedArtifact:
"""
Args:
exported_program: Exported Program to serialize
"""
gm_serializer = GraphModuleSerializer(
exported_program.graph_signature,
exported_program.module_call_graph
)
serialized_graph_module = gm_serializer.serialize(exported_program.graph_module)
serialized_range_constraints = serialize_range_constraints(exported_program.range_constraints)
# TODO: Directly serialize exported_program.constants once
# CustomClassHolders get stored in the ExportedProgram rather than in
# the graph
constants = {}
for n, c in gm_serializer.custom_objs.items():
constants[n] = c
for n, t in exported_program.tensor_constants.items():
assert n not in constants
constants[n] = t
return SerializedArtifact(
ExportedProgram(
graph_module=serialized_graph_module,
opset_version=self.opset_version,
range_constraints=serialized_range_constraints,
schema_version=SCHEMA_VERSION,
dialect=exported_program.dialect,
),
serialize_torch_artifact(exported_program.state_dict),
serialize_torch_artifact(constants),
)
class GraphModuleDeserializer:
@dataclasses.dataclass
class Result:
graph_module: torch.fx.GraphModule
signature: ep.ExportGraphSignature
module_call_graph: List[ep.ModuleCallEntry]
names_to_symbols: Dict[str, sympy.Symbol]
def __init__(self):
self.serialized_name_to_node: Dict[str, torch.fx.Node] = {}
self.serialized_name_to_meta: Dict[str, MetaType] = {}
self.graph = torch.fx.Graph()
self.module = torch.nn.Module()
@contextmanager
def save_graph_module(self) -> Iterator[None]:
saved = self.graph, self.module, self.serialized_name_to_node, self.serialized_name_to_meta
self.graph = torch.fx.Graph()
self.module = torch.nn.Module()
self.serialized_name_to_node = {}
self.serialized_name_to_meta = {}
try:
yield
finally:
self.graph, self.module, self.serialized_name_to_node, self.serialized_name_to_meta = saved
def deserialize_operator(self, serialized_target: str):
if serialized_target.startswith("_operator"): # TODO(zhxchen17) Follow up on this.
module = operator
serialized_target_names = serialized_target.split(".")[1:]
elif serialized_target.startswith("torch.ops"):
module = torch.ops
serialized_target_names = serialized_target.split(".")[2:]
else: # TODO(zhxchen17) Don't catch all here.
return serialized_target
target = module
for name in serialized_target_names:
if not hasattr(target, name):
return serialized_target
else:
target = getattr(target, name)
return target
def deserialize_sym_int(self, s: SymInt) -> Union[int, torch.SymInt]:
val = s.value
if s.type == "as_expr":
if val.expr_str in self.symbol_name_to_symbol:
sym = self.symbol_name_to_symbol[val.expr_str]
else:
sym = sympy.sympify(val.expr_str, locals=self.symbol_name_to_symbol)
if isinstance(sym, sympy.Symbol):
self.symbol_name_to_symbol[val.expr_str] = sym
if vr := self.symbol_name_to_range.get(val.expr_str):
symbolic_shapes._constrain_symbol_range(
self.shape_env,
sym,
compiler_min=vr.lower, # type: ignore[arg-type]
compiler_max=vr.upper, # type: ignore[arg-type]
runtime_min=vr.lower, # type: ignore[arg-type]
runtime_max=vr.upper # type: ignore[arg-type]
)
if val.hint is None:
hint = None
else:
assert val.hint.type == "as_int"
hint = val.hint.value
return self.shape_env.create_symintnode(sym, hint=hint)
elif s.type == "as_int":
assert isinstance(val, int)
return val
else:
raise SerializeError(
f"SymInt has invalid field type {s.type} with value {s.value}"
)
def deserialize_sym_bool(self, s: SymBool) -> Union[bool, torch.SymBool]:
val = s.value
if s.type == "as_expr":
expr = sympy.sympify(val.expr_str, locals=self.symbol_name_to_symbol)
return self.shape_env.create_symboolnode(expr)
elif s.type == "as_bool":
assert isinstance(val, bool)
return val
else:
raise SerializeError(
f"SymBool has invalid field type {s.type} with value {s.value}"
)
def deserialize_tensor_meta(
self,
tensor_meta: TensorMeta,
fake_tensor_mode: FakeTensorMode,
) -> FakeTensor:
with fake_tensor_mode:
return cast(
FakeTensor,
torch.empty_strided(
tuple(self.deserialize_sym_int(val) for val in tensor_meta.sizes), # type: ignore[misc]
tuple(self.deserialize_sym_int(val) for val in tensor_meta.strides), # type: ignore[misc]
device=deserialize_device(tensor_meta.device),
dtype=_SERIALIZE_TO_TORCH_DTYPE[tensor_meta.dtype],
),
)
def deserialize_graph_output(self, output) -> torch.fx.Node:
if isinstance(output.value, TensorArgument):
return self.serialized_name_to_node[output.value.name]
elif isinstance(output.value, (SymIntArgument, SymBoolArgument)):
return self.serialized_name_to_node[output.value.as_name]
else:
raise SerializeError(f"Unable to deserialize output node {output}")
def deserialize_graph(self, serialized_graph: Graph) -> torch.fx.Graph:
# Handle the tensor metas.
for name, tensor_value in serialized_graph.tensor_values.items():
meta_val = self.deserialize_tensor_meta(tensor_value, self.fake_tensor_mode)
self.serialized_name_to_meta[name] = meta_val
for name, sym_int_value in serialized_graph.sym_int_values.items():
self.serialized_name_to_meta[name] = self.deserialize_sym_int(sym_int_value)
for name, sym_bool_value in serialized_graph.sym_bool_values.items():
self.serialized_name_to_meta[name] = self.deserialize_sym_bool(sym_bool_value)
# Inputs: convert to placeholder nodes in FX.
for input in serialized_graph.inputs:
placeholder_node = self.graph.placeholder(input.as_tensor.name)
self.sync_fx_node(input.as_tensor.name, placeholder_node)
# Nodes: convert to call_function nodes.
for serialized_node in serialized_graph.nodes:
try:
target = self.deserialize_operator(serialized_node.target)
self.deserialize_node(serialized_node, target)
except Exception as e:
raise SerializeError(f"Failed deserializing node {serialized_node}") from e
# Outputs: convert to a single `output` node.
outputs = []
for output in serialized_graph.outputs:
outputs.append(self.deserialize_graph_output(output))
if serialized_graph.is_single_tensor_return:
assert len(outputs) == 1
outputs = outputs[0] # type: ignore[assignment]
else:
outputs = tuple(outputs) # type: ignore[assignment]
output_node = self.graph.output(outputs)
if serialized_graph.is_single_tensor_return:
output_node.meta["val"] = output_node.args[0].meta["val"]
else:
output_node.meta["val"] = tuple(
arg.meta["val"] for arg in output_node.args[0]
)
return self.graph
def deserialize_node(self, serialized_node: Node, target: Callable) -> None:
if target.__module__ == "_operator": # TODO(zhxchen17) Follow up on this.
name = serialized_node.outputs[0].value.as_name
args = self.deserialize_sym_op_inputs(serialized_node.inputs)
fx_node = self.graph.create_node("call_function", target, args, {}, name)
self.deserialize_sym_op_outputs(serialized_node, fx_node)
elif isinstance(target, torch._ops.HigherOrderOperator):
assert (
len(serialized_node.outputs) == 1
and serialized_node.outputs[0].type in ("as_tensors", "as_tensor")
), "Only single tensor output or list of tensor output is supported for higher order operators."
output = serialized_node.outputs[0]
name = (
output.value.name
if output.type == "as_tensor"
else None # FX will generate a name for us.
)
args = tuple(self.deserialize_input(input.arg) for input in serialized_node.inputs)
fx_node = self.graph.create_node("call_function", target, args, {}, name)
if output.type == "as_tensor":
self.sync_fx_node(name, fx_node)
if output.type == "as_tensors":
self.deserialize_multiple_outputs(serialized_node, fx_node)
elif isinstance(target, torch._ops.OpOverload):
# For convenience: if this node returns a single tensor, name the
# newly-created node after it. This ensures that these tensor values
# have names that are consistent with serialized.
name = (
serialized_node.outputs[0].value.name
if _is_single_tensor_return(target)
else None # FX will generate a name for us.
)
args, kwargs = self.deserialize_inputs(target, serialized_node)
fx_node = self.graph.create_node("call_function", target, args, kwargs, name)
self.deserialize_outputs(serialized_node, fx_node)
else:
raise SerializeError(f"Unsupported target type for node {serialized_node}: {target}")
fx_node.meta.update(self.deserialize_metadata(serialized_node.metadata))
def deserialize_input_spec(self, i: InputSpec) -> ep.InputSpec:
if i.user_input is not None:
return ep.InputSpec(
kind=ep.InputKind.USER_INPUT,
arg=self.deserialize_argument_spec(i.user_input.arg),
target=None
)
elif i.parameter is not None:
return ep.InputSpec(
kind=ep.InputKind.PARAMETER,
arg=PyTensorArgument(name=i.parameter.arg.name),
target=i.parameter.parameter_name,
)
elif i.buffer is not None:
return ep.InputSpec(
kind=ep.InputKind.BUFFER,
arg=PyTensorArgument(name=i.buffer.arg.name),
target=i.buffer.buffer_name,
)
elif i.tensor_constant is not None:
return ep.InputSpec(
kind=ep.InputKind.CONSTANT_TENSOR,
arg=PyTensorArgument(name=i.tensor_constant.arg.name),
target=i.tensor_constant.tensor_constant_name,
)
else:
raise AssertionError(f"Unkown input spec {i}")
def deserialize_output_spec(self, o: OutputSpec) -> ep.OutputSpec:
if o.user_output is not None:
return ep.OutputSpec(
kind=ep.OutputKind.USER_OUTPUT,
arg=self.deserialize_argument_spec(o.user_output.arg),
target=None,
)
elif o.loss_output is not None:
return ep.OutputSpec(
kind=ep.OutputKind.LOSS_OUTPUT,
arg=PyTensorArgument(name=o.loss_output.arg.name),
target=None,
)
elif o.buffer_mutation is not None:
return ep.OutputSpec(
kind=ep.OutputKind.BUFFER_MUTATION,
arg=PyTensorArgument(name=o.buffer_mutation.arg.name),
target=o.buffer_mutation.buffer_name
)
elif o.gradient_to_parameter is not None:
return ep.OutputSpec(
kind=ep.OutputKind.GRADIENT_TO_PARAMETER,
arg=PyTensorArgument(name=o.gradient_to_parameter.arg.name),
target=o.gradient_to_parameter.parameter_name
)
elif o.gradient_to_user_input is not None:
return ep.OutputSpec(
kind=ep.OutputKind.GRADIENT_TO_USER_INPUT,
arg=PyTensorArgument(name=o.gradient_to_user_input.arg.name),
target=o.gradient_to_user_input.user_input_name
)
else:
raise AssertionError(f"Unknown output spec {o}")
def deserialize_signature(self, sig: GraphSignature) -> ep.ExportGraphSignature:
return ep.ExportGraphSignature(
input_specs=[self.deserialize_input_spec(i) for i in sig.input_specs],
output_specs=[self.deserialize_output_spec(o) for o in sig.output_specs]
)
def deserialize(
self,
serialized_graph_module: GraphModule,
symbol_name_to_range: Optional[Dict[str, symbolic_shapes.ValueRanges]] = None,
constants: Optional[Dict[str, Any]] = None,
) -> Result:
self.shape_env = symbolic_shapes.ShapeEnv(assume_static_by_default=True)
self.fake_tensor_mode = FakeTensorMode(
allow_fallback_kernels=False,
allow_non_fake_inputs=True,
shape_env=self.shape_env,
)
self.symbol_name_to_symbol: Dict[str, sympy.Symbol] = {}
self.symbol_name_to_range = {} if symbol_name_to_range is None else symbol_name_to_range
self.constants = {} if constants is None else constants
self.deserialize_graph(serialized_graph_module.graph)
sig = self.deserialize_signature(serialized_graph_module.signature)
module_call_graph = self.deserialize_module_call_graph(serialized_graph_module.module_call_graph)
return GraphModuleDeserializer.Result(
graph_module=torch._export.exported_program._create_graph_module_for_export(self.module, self.graph),
signature=sig,
module_call_graph=module_call_graph,
names_to_symbols=self.symbol_name_to_symbol,
)
def sync_fx_node(self, name: str, fx_node: torch.fx.Node):
if name in self.serialized_name_to_node:
raise SerializeError(f"Node {name} has already been deserialized before.")
self.serialized_name_to_node[name] = fx_node
assert "val" not in fx_node.meta
fx_node.meta["val"] = self.serialized_name_to_meta[name]
def deserialize_sym_op_inputs(self, inputs):
return tuple(self.deserialize_input(input.arg) for input in inputs)
def deserialize_inputs(self, target: torch._ops.OpOverload, serialized_node: Node):
schema_args = target._schema.arguments
actual_args = {
input.name: self.deserialize_input(input.arg) for input in serialized_node.inputs
}
args = []
kwargs = {}
for schema_arg in schema_args:
is_positional = not schema_arg.has_default_value() and not schema_arg.kwarg_only
if is_positional:
args.append(actual_args[schema_arg.name])
else:
if schema_arg.name in actual_args:
kwargs[schema_arg.name] = actual_args[schema_arg.name]
return tuple(args), kwargs
def deserialize_input(self, inp: Argument) -> Any:
value = inp.value
typ_ = inp.type
if typ_ == "as_none":
# None should converted as None, but is encoded as bool in serialized
# Convert serialized object to torch equivalent
return None
elif typ_ == "as_scalar_type":
return _SERIALIZE_TO_TORCH_DTYPE[value]
elif typ_ == "as_memory_format":
return _SERIALIZE_TO_TORCH_MEMORY_FORMAT[value]
elif typ_ == "as_layout":
return _SERIALIZE_TO_TORCH_LAYOUT[value]
elif typ_ == "as_graph":
assert isinstance(value, GraphArgument)
with self.save_graph_module():
self.deserialize_graph(value.graph)
submodule = torch._export.exported_program._create_graph_module_for_export(self.module, self.graph)
self.module.register_module(value.name, submodule)
return self.graph.create_node(
"get_attr",
value.name,
name=value.name,
)
elif isinstance(value, Device):
return deserialize_device(value)
elif isinstance(value, TensorArgument):
return self.serialized_name_to_node[value.name]
elif isinstance(value, (int, float, bool)):
return value
elif isinstance(value, str):
return str(value)
elif isinstance(value, (SymIntArgument, SymBoolArgument)):
return self.deserialize_sym_argument(value)
elif isinstance(value, list):
if len(value) == 0:
return []
elif isinstance(value[0], TensorArgument):
result = []
for arg in value:
result.append(self.serialized_name_to_node[arg.name])
return result
elif isinstance(value[0], (int, float, bool)):
# convert from serialized.python.types.List to python list
return list(value)
elif isinstance(value[0], (SymIntArgument, SymBoolArgument)):
return [self.deserialize_sym_argument(arg) for arg in value]
elif isinstance(value[0], OptionalTensorArgument):
def deserialize_optional_tensor_args(a):
if a.type == "as_none":
return None
elif a.type == "as_tensor":
return self.serialized_name_to_node[a.value]
else:
raise SerializeError(f"Unhandled argument {inp}")
return list(map(deserialize_optional_tensor_args, value))
else:
raise SerializeError(f"Unhandled argument {inp}")
elif isinstance(value, CustomObjArgument):
return self.constants[value.name]
else:
raise SerializeError(f"Unhandled argument {inp}")
def deserialize_sym_argument(self, sym_int_arg):
if sym_int_arg.type == "as_int":
return sym_int_arg.as_int
else:
assert sym_int_arg.type == "as_name"
return self.serialized_name_to_node[sym_int_arg.as_name]
def deserialize_sym_op_outputs(self, serialized_node: Node, fx_node: torch.fx.Node):
self.sync_fx_node(serialized_node.outputs[0].value.as_name, fx_node)
def deserialize_outputs(self, serialized_node: Node, fx_node: torch.fx.Node):
# Simple case for single tensor return.
assert isinstance(fx_node.target, torch._ops.OpOverload)
returns = fx_node.target._schema.returns
# Check single value return
if len(returns) == 0:
return
if _is_single_tensor_return(fx_node.target):
self.sync_fx_node(serialized_node.outputs[0].as_tensor.name, fx_node)
return
elif len(returns) == 1 and isinstance(serialized_node.outputs[0].value, (SymIntArgument, SymBoolArgument)):
self.sync_fx_node(serialized_node.outputs[0].value.as_name, fx_node)
return
self.deserialize_multiple_outputs(serialized_node, fx_node)
def deserialize_multiple_outputs(self, serialized_node: Node, fx_node: torch.fx.Node) -> None:
deserialized_metadata = self.deserialize_metadata(serialized_node.metadata)
def generate_getitem(meta_val, fx_node: torch.fx.Node, arg: TensorArgument, idx: int):
name = arg.name
individual_output = self.graph.create_node(
"call_function",
operator.getitem,
(fx_node, idx),
name=name,
)
self.sync_fx_node(name, individual_output)
meta_val.append(self.serialized_name_to_meta[name])
# The derived `getitem` nodes should have the same stacktrace as the
# original `fx_node`
individual_output.meta.update(deserialized_metadata)
def generate_getitems(meta_val, fx_node: torch.fx.Node, args):
for idx, arg in enumerate(args):
if isinstance(arg, Argument):
arg = arg.value
if isinstance(arg, TensorArgument):
generate_getitem(meta_val, fx_node, arg, idx)
elif isinstance(arg, (list, tuple)):
list_output = self.graph.create_node(
"call_function",
operator.getitem,
(fx_node, idx),
)
meta_val.append([])
generate_getitems(meta_val[-1], list_output, arg)
list_output.meta.update(deserialized_metadata)
list_output.meta['val'] = meta_val[-1]
else:
raise NotImplementedError(f"Unimplemented node output type: {arg}")
# Convert multiple return types to FX format.
# In FX, each node only returns one value. So in order to represent
# multiple return values, we have to emit a `getitem` node for each
# return value.
# This performs the inverse mapping of the `serialize_outputs` call in
# serialization, see [NOTE: Multiple outputs]
meta_val: List[Any] = []
if len(serialized_node.outputs) == 1:
assert isinstance(serialized_node.outputs[0].value, list)
assert isinstance(serialized_node.outputs[0].value[0], TensorArgument)
generate_getitems(meta_val, fx_node, serialized_node.outputs[0].as_tensors)
else:
generate_getitems(meta_val, fx_node, serialized_node.outputs)
# also update the metaval for `fx_node` to be a list(meta)
fx_node.meta["val"] = tuple(meta_val)
self.serialized_name_to_node[fx_node.name] = fx_node
def deserialize_metadata(self, metadata: Dict[str, str]) -> Dict[str, Any]:
ret: Dict[str, Any] = {}
if stack_trace := metadata.get("stack_trace"):
ret["stack_trace"] = stack_trace
def deserialize_meta_func(serialized_target: str):
module = None
if serialized_target.startswith("torch.nn"):
module = torch.nn
serialized_target_names = serialized_target.split(".")[2:]
elif serialized_target.startswith("torch"):
module = torch
serialized_target_names = serialized_target.split(".")[1:]
else:
return self.deserialize_operator(serialized_target)
target = module
for name in serialized_target_names:
if not hasattr(target, name):
return serialized_target
else:
target = getattr(target, name)
return target
if nn_module_stack_str := metadata.get("nn_module_stack"):
# Originally serialized to "key,orig_path,type_str"
def import_nn_module_stack(key, path, ty):
return key, (path, ty)
nn_module_stack = dict(
import_nn_module_stack(*item.split(","))
for item in nn_module_stack_str.split(ST_DELIMITER)
)
ret["nn_module_stack"] = nn_module_stack
if source_fn_st_str := metadata.get("source_fn_stack"):
# Originally serializes to "fx_node_name,op_str"
source_fn_st = []
for source_fn_str in source_fn_st_str.split(ST_DELIMITER):
name, target_str = source_fn_str.split(",")
source_fn_st.append((name, deserialize_meta_func(target_str)))
ret["source_fn_stack"] = source_fn_st
return ret
def deserialize_argument_spec(self, x: Argument) -> ep.ArgumentSpec:
if x.as_tensor is not None:
return PyTensorArgument(name=x.as_tensor.name)
elif x.as_sym_int is not None:
return PySymIntArgument(name=x.as_sym_int.as_name)
else:
return PyConstantArgument(value=self.deserialize_input(x))
def deserialize_module_call_signature(self, module_call_signature: ModuleCallSignature) -> ep.ModuleCallSignature:
return ep.ModuleCallSignature(
inputs=[self.deserialize_argument_spec(x) for x in module_call_signature.inputs],
outputs=[self.deserialize_argument_spec(x) for x in module_call_signature.outputs],
in_spec=treespec_loads(module_call_signature.in_spec),
out_spec=treespec_loads(module_call_signature.out_spec),
)
def deserialize_module_call_graph(self, module_call_graph: List[ModuleCallEntry]) -> List[ep.ModuleCallEntry]:
return [
ep.ModuleCallEntry(
fqn=entry.fqn,
signature=self.deserialize_module_call_signature(entry.signature) if entry.signature else None,
) for entry in module_call_graph
]
class ExportedProgramDeserializer:
def __init__(self, expected_opset_version: Optional[Dict[str, int]] = None):
self.expected_opset_version: Dict[str, int] = {}
if expected_opset_version:
self.expected_opset_version.update(expected_opset_version)
if "aten" not in self.expected_opset_version:
self.expected_opset_version["aten"] = torch._C._get_max_operator_version()
def deserialize_range_constraints(
self,
symbol_name_to_range: Dict[str, symbolic_shapes.ValueRanges],
symbol_name_to_symbol: Dict[str, sympy.Symbol],
) -> Dict[sympy.Symbol, ValueRanges]:
range_constraints = {}
for k, v in symbol_name_to_range.items():
if symbol := symbol_name_to_symbol.get(k):
range_constraints[symbol] = v # type: ignore[arg-type]
else:
log.warning(f"Symbol {k} did not appear in the graph that was deserialized") # noqa: G004
return range_constraints
def deserialize(
self, serialized_artifact: SerializedArtifact
) -> ep.ExportedProgram:
assert isinstance(serialized_artifact.exported_program, ExportedProgram)
if serialized_artifact.exported_program.schema_version != SCHEMA_VERSION:
raise SerializeError(
f"Serialized schema version {serialized_artifact.exported_program.schema_version} "
f"does not match our current schema version {SCHEMA_VERSION}."
)
symbol_name_to_range = {
k: symbolic_shapes.ValueRanges(_int_to_sympy_int(v.min_val), _int_to_sympy_int(v.max_val))
for k, v in serialized_artifact.exported_program.range_constraints.items()
}
constants = deserialize_torch_artifact(serialized_artifact.constants)
# TODO: No need to do this once CustomClassHolders are lifted to the ExportedProgram
tensor_constants = {
k: v for k, v in constants.items() if isinstance(v, torch.Tensor)
}
res = (
GraphModuleDeserializer()
.deserialize(
serialized_artifact.exported_program.graph_module,
symbol_name_to_range,
constants,
)
)
range_constraints = self.deserialize_range_constraints(
symbol_name_to_range, res.names_to_symbols,
)
model_opset_version: Optional[Dict[str, int]] = serialized_artifact.exported_program.opset_version
self._validate_model_opset_version(model_opset_version)
upgrader = GraphModuleOpUpgrader(self.expected_opset_version, model_opset_version)
state_dict = deserialize_torch_artifact(serialized_artifact.state_dict)
exported_program = ep.ExportedProgram(
res.graph_module,
res.graph_module.graph,
res.signature,
state_dict, # type: ignore[arg-type]
range_constraints,
[],
res.module_call_graph,
None,
load_verifier(serialized_artifact.exported_program.dialect),
tensor_constants=tensor_constants,
)
return upgrader.upgrade(exported_program)
def _validate_model_opset_version(self, model_opset_version: Optional[Dict[str, int]]):
"""Compare model_opset_version with expected_opset_version and raise error if we can't resolve the version
difference.
E.g., model_opset_version = {"aten": 3, "custom": 4}
expected_opset_version = {"aten": 4, "custom": 4}
This means we can use an upgrader for ATen to reconcile the deserialized model.
The logic of this method:
For common op namespaces:
1. if model version < expected version, this case can be handled by upgraders.
2. if model version > expected version, we need downgraders but not implemented yet.
3. if model version == expected version, we don't need extra handling.
For op namespace only in model_opset_version, we should give a warning because it is missing from
expected_opset_version.
"""
if not model_opset_version:
raise RuntimeError("Serialized model should have opset version.")
common_namespaces = {key for key in model_opset_version if key in self.expected_opset_version}
for namespace in common_namespaces:
assert (
isinstance(model_version := model_opset_version[namespace], int)
), f"model_opset_version value should be int, got {model_opset_version[namespace]}"
assert (
isinstance(compiler_version := self.expected_opset_version[namespace], int)
), f"expected_opset_version value should be int, got {self.expected_opset_version[namespace]}"
# TODO(larryliu0820): Add support for upgrader & downgrader
if model_version != compiler_version:
raise NotImplementedError(
f"Model opset version {model_opset_version} doesn't match to compiler opset version "
f"{self.expected_opset_version}! Upgrader/downgrader is not implemented yet."
)
for namespace in model_opset_version:
if namespace in common_namespaces:
continue
log.warning("Compiler doesn't have a version table for op namespace: {ns}. ", extra={"ns": namespace})
class EnumEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj, Enum):
return obj.value
if isinstance(obj, bytes):
return base64.b64encode(obj).decode('utf-8')
return super().default(obj)
def serialize(
exported_program: ep.ExportedProgram,
opset_version: Optional[Dict[str, int]] = None,
) -> SerializedArtifact:
exported_program._validate()
serialized_artifact = (
ExportedProgramSerializer(opset_version).serialize(exported_program)
)
assert isinstance(serialized_artifact.exported_program, ExportedProgram)
json_program = json.dumps(
dataclasses.asdict(serialized_artifact.exported_program), cls=EnumEncoder
)
json_bytes = json_program.encode('utf-8')
artifact = SerializedArtifact(
json_bytes,
serialized_artifact.state_dict,
serialized_artifact.constants
)
return artifact
def _dict_to_dataclass(cls, data):
assert not isinstance(cls, str), f"Unresolved class type: '{cls}'."
if typing.get_origin(cls) == typing.Union and type(None) in typing.get_args(cls):
if data is None:
return None
ty_args = typing.get_args(cls)
assert len(ty_args) == 2
return _dict_to_dataclass(ty_args[0], data)
elif isinstance(cls, type) and issubclass(cls, _Union):
obj = cls(**data)
field_type = cls.__annotations__[obj.type]
setattr(obj, obj.type, _dict_to_dataclass(field_type, obj.value))
return obj
elif dataclasses.is_dataclass(cls):
obj = cls(**data) # type: ignore[assignment]
type_hints = typing.get_type_hints(cls)
for f in dataclasses.fields(cls):
name = f.name
new_field_obj = _dict_to_dataclass(type_hints[name], getattr(obj, name))
setattr(obj, name, new_field_obj)
return obj
elif isinstance(data, list):
if len(data) == 0:
return data
d_type = typing.get_args(cls)[0]
return [
_dict_to_dataclass(d_type, d)
for d in data
]
elif isinstance(data, dict):
v_type = typing.get_args(cls)[1]
return {
k: _dict_to_dataclass(v_type, v)
for k, v in data.items()
}
return data
def deserialize(
artifact: SerializedArtifact,
expected_opset_version: Optional[Dict[str, int]] = None,
) -> ep.ExportedProgram:
assert isinstance(artifact.exported_program, bytes)
exported_program_str = artifact.exported_program.decode('utf-8')
exported_program_dict = json.loads(exported_program_str)
serialized_exported_program = _dict_to_dataclass(ExportedProgram, exported_program_dict)
return (
ExportedProgramDeserializer(expected_opset_version)
.deserialize(
SerializedArtifact(
serialized_exported_program,
artifact.state_dict,
artifact.constants
)
)
)