# mypy: allow-untyped-defs
import contextlib
import ctypes
import importlib
import inspect
import sys
import types
from typing import Any, Callable, Dict, List, Set, Type, Union
import torch._C
import torch.utils._pytree as pytree
from torch import _utils_internal
from torch._functorch.pyfunctorch import dispatch_functorch
from torch.utils._python_dispatch import TorchDispatchMode
# Query `hasattr` only once.
_SET_GLOBAL_FLAGS = hasattr(sys, "getdlopenflags") and hasattr(sys, "setdlopenflags")
@contextlib.contextmanager
def dl_open_guard():
"""
Context manager to set the RTLD_GLOBAL dynamic linker flag while we open a
shared library to load custom operators.
"""
if not _SET_GLOBAL_FLAGS:
yield
return
old_flags = sys.getdlopenflags()
sys.setdlopenflags(old_flags | ctypes.RTLD_GLOBAL)
try:
yield
finally:
sys.setdlopenflags(old_flags)
class OperatorBase:
"""
Base class for OpOverload (which represents C++ ATen operators) and HigherOrderOperator
(which represents Python-only operators that are unrepresentable in TorchScript).
"""
def __init__(self):
# The dispatch cache precomputes a mapping of dispatch key that the
# dispatcher wants to dispatch to, to an actual implementation of the
# dispatch key. Confusingly, the actual implementation could *also* be a
# dispatch key, but in this case, this refers to the C++ kernel that
# was registered to some dispatch key. Aliases are permitted in the
# latter but not the former; for example, you might lookup the
# entry for AutogradCPU, and this maps you to the Autograd key for
# the generic autograd kernel that works for all devices. Since this
# is the Python dispatcher, you can also put an arbitrary Python
# callable to call instead. This handler gets precisely the
# args/kwargs that the operator was __call__'ed with.
# NB: This name is hard-coded in torch/csrc/autograd/python_variable.cpp
# for use with OpOverload; cache lookup is done entirely from C++
# for speed.
# TODO: The cache is NOT currently used by HigherOrderOperator, but it should!
self._dispatch_cache: Dict[
torch._C.DispatchKey, Union[torch._C.DispatchKey, Callable[..., Any]]
] = {}
# This table allows you to override the behavior of a particular
# dispatch key to call a custom Python function, rather than the
# ordinary C++ configured behavior. This is the raison d'etre of
# Python dispatcher: to let you program the dispatcher from Python
# in case you need something unusual, and don't want to clobber
# the existing registrations using the Python operator registration
# API.
self.py_kernels: Dict[torch._C.DispatchKey, Callable[..., Any]] = {}
# This table allows you to override the behavior of a particular
# operator for a particular TorchDispatchMode. In practice,
# we are using this mostly for ProxyTensorMode. Modes can be
# thought of as an open world extension of dispatch keys, so it
# makes sense that you should be able to register them, the same
# way you can register dispatch keys.
self.python_key_mode_table: Dict[
Type[TorchDispatchMode], Callable[..., Any]
] = {}
# This table allows you to override the behavior of functorch
# transformations. NB: this currently only does something for
# HigherOrderOperator
self.functorch_table = {}
def __call__(self, *args, **kwargs):
raise NotImplementedError
def has_kernel_for_dispatch_key(self, k):
return k in self.py_kernels
def has_kernel_for_any_dispatch_key(self, ks):
for k in self.py_kernels:
if not torch._C._dispatch_is_alias_key(k) and ks.has(k):
return True
return False
def py_impl(self, k):
def inner(fn):
if inspect.isclass(k) and issubclass(k, TorchDispatchMode):
assert k not in self.python_key_mode_table
# TODO(voz): Should we replace setting torch._C.DispatchKey.Python entirely with setting mode keys?
self.python_key_mode_table[k] = fn
self._dispatch_cache.clear()
return fn
if isinstance(k, torch._C._functorch.TransformType):
assert k not in self.functorch_table
self.functorch_table[k] = fn
return fn
assert isinstance(k, torch._C.DispatchKey)
assert (
k != torch._C.DispatchKey.Python
), "Please register a mode for the torch._C.DispatchKey.Python key instead."
if k in self.py_kernels:
raise RuntimeError(
f"Trying to override a python impl for {k} on operator {self.name()}"
)
self.py_kernels[k] = fn
self._dispatch_cache.clear()
return fn
return inner
# Registers an implementation to all **3** variants of functionalization that we have:
# - DispatchKey.Functionalize
# - functorch.TransformType.Functionalize
# - FunctionalTensorMode
# Example:
# @py_functionalize_impl
# def functionalize_rule(ctx, inner_f, *args):
# args_unwrapped = ctx.unwrap_tensors(args)
# with ctx.redispatch_to_next():
# out = ctx.functionalize(inner_f)(*args_unwrapped)
# return ctx.wrap_tensors(out)
def py_functionalize_impl(self, fn):
from torch._subclasses.functional_tensor import (
CppFunctionalizeAPI as _CppFunctionalizeAPI,
FunctorchFunctionalizeAPI as _FunctorchFunctionalizeAPI,
PythonFunctionalizeAPI as _PythonFunctionalizeAPI,
)
# Construct our three flavors of functionalization,
# each of which have slightly different wrap/unwrap/redispatch policies
def functionalize_dk_fn(*args, **kwargs):
return fn(_CppFunctionalizeAPI(), *args, **kwargs)
def functionalize_dispatch_mode_fn(mode, *args, **kwargs):
return fn(_PythonFunctionalizeAPI(mode), *args, **kwargs)
def functionalize_functorch_fn(interpreter, *args, **kwargs):
return fn(_FunctorchFunctionalizeAPI(interpreter), *args, **kwargs)
self.py_impl(torch._C.DispatchKey.Functionalize)(functionalize_dk_fn)
self.py_impl(torch._subclasses.functional_tensor.FunctionalTensorMode)(
functionalize_dispatch_mode_fn
)
self.py_impl(torch._C._functorch.TransformType.Functionalize)(
functionalize_functorch_fn
)
return fn
def name(self):
raise NotImplementedError
is_included_in_alias = torch._C._dispatch_is_included_in_alias
DispatchKey = torch._C.DispatchKey
# Equivalent to computeDispatchTableEntryWithDebug
def resolve_key(op: OperatorBase, k: DispatchKey): # type: ignore[valid-type]
# 1. (Direct) operator registration
if op.has_kernel_for_dispatch_key(k):
return k
# 2.1 Use CompositeExplicitAutogradNonFunctional kernel if available
cand = DispatchKey.CompositeExplicitAutogradNonFunctional
if (
k == DispatchKey.Undefined or is_included_in_alias(k, cand)
) and op.has_kernel_for_dispatch_key(cand):
return cand
# 2.2 Use CompositeExplicitAutograd kernel if available
cand = DispatchKey.CompositeExplicitAutograd
if (
k == DispatchKey.Undefined or is_included_in_alias(k, cand)
) and op.has_kernel_for_dispatch_key(cand):
return cand
has_backend_kernel = op.has_kernel_for_any_dispatch_key(
torch._C._dispatch_get_backend_keyset_from_autograd(k)
) or op.has_kernel_for_dispatch_key(DispatchKey.CompositeExplicitAutograd)
# 2.3. Use CompositeImplicitAutograd kernel if available
cand = DispatchKey.CompositeImplicitAutogradNestedTensor
if (
(k != DispatchKey.Undefined and is_included_in_alias(k, cand))
and op.has_kernel_for_dispatch_key(cand)
and not has_backend_kernel
):
return cand
cand = DispatchKey.CompositeImplicitAutograd
if (
k == DispatchKey.Undefined or is_included_in_alias(k, cand)
) and op.has_kernel_for_dispatch_key(cand):
if k == DispatchKey.AutogradOther and op.has_kernel_for_any_dispatch_key(
torch._C._dispatch_autogradother_backends
):
raise RuntimeError("ambiguous autogradother kernel")
elif not has_backend_kernel:
return cand
# 2.4. For autograd backend keys, use kernel from DispatchKey::Autograd if available
cand = DispatchKey.Autograd
if is_included_in_alias(k, cand) and op.has_kernel_for_dispatch_key(cand):
return cand
# 2.5 Use kernel from DispatchKey::FuncTorchBatchedDecomposition if available
cand = DispatchKey.FuncTorchBatchedDecomposition
if is_included_in_alias(k, cand) and op.has_kernel_for_dispatch_key(cand):
return cand
# Backend fallback
if torch._C._dispatch_has_backend_fallback(k):
# The dispatch key itself will implicitly route to backend fallback.
# This is probably not great for the pure Python implementation.
return k
raise NotImplementedError(f"could not find kernel for {op} at dispatch key {k}")
_higher_order_ops: Dict[str, "HigherOrderOperator"] = {}
_HIGHER_ORDER_OP_DEFAULT_FALLTHROUGH_DISPATCH_KEYS = [
DispatchKey.PythonDispatcher, # type: ignore[attr-defined]
DispatchKey.PythonTLSSnapshot, # type: ignore[attr-defined]
DispatchKey.ADInplaceOrView,
DispatchKey.BackendSelect,
DispatchKey.AutocastCPU, # type: ignore[attr-defined]
DispatchKey.AutocastCUDA, # type: ignore[attr-defined]
]
class HigherOrderOperator(OperatorBase):
# The HigherOrderOperator will appear as torch.ops.higher_order.{name}
#
# If you're creating a new HigherOrderOperator, please do not change the
# default. Adding operators to the global torch.ops namespace is a bad
# practice due to name collisions.
def __init__(self, name):
super().__init__()
self._name = name
# Make _OPNamespace not scream, this whole name based association needs a good hard look
self.__name__ = name
_higher_order_ops[name] = self
self._ns = "higher_order"
# For a normal HigherOrderOperator instance, we will change its __module__ from torch._ops to
# torch._ops.higher_order.
# For an instance of subclass of HigherOrderOperator (e.g. customized higher order op),
# the __module__ attribute will be kept unchanged.
if self.__class__ is HigherOrderOperator:
self_name_space = "." + self.namespace if self.namespace else ""
self.__module__ = self.__module__ + self_name_space
self.non_fallthrough_keys = torch._C._dispatch_keyset_full()
for dispatch_key in _HIGHER_ORDER_OP_DEFAULT_FALLTHROUGH_DISPATCH_KEYS:
self.fallthrough(dispatch_key)
# [NOTE] We have to register pre-dispatch key implementation
# because sometimes HOP use aot-dispatch tracing to detect certaion
# mutations. This is problematic when we are functionalizing HOP
# during pre-dispatch because when the inner tracer starts, it will see
# that PreDispatch key is still active. In that case, we just redispatch
# it to next key. This is only safe to do when PreDispatch key stack has no
# active modes.
def py_impl(self, k):
if isinstance(k, torch._C.DispatchKey) and not self.non_fallthrough_keys.has(k):
self.non_fallthrough_keys = self.non_fallthrough_keys.add(k)
return super().py_impl(k)
@property
def namespace(self):
return self._ns
def fallthrough(self, dispatch_key):
self.non_fallthrough_keys = self.non_fallthrough_keys.remove(dispatch_key)
def dispatch(self, dispatch_key, *args, **kwargs):
from torch.utils._python_dispatch import _get_current_dispatch_mode
if dispatch_key in self._dispatch_cache:
kernel = self._dispatch_cache[dispatch_key]
assert not isinstance(kernel, torch._C.DispatchKey)
return kernel(*args, **kwargs)
if dispatch_key == torch._C.DispatchKey.FuncTorchDynamicLayerFrontMode:
return dispatch_functorch(self, args, kwargs)
if dispatch_key == torch._C.DispatchKey.Python:
# The place to handle ProxyTorchDispatchMode, FakeTensorMode, etc
from torch.utils._python_dispatch import _pop_mode_temporarily
curr_mode = _get_current_dispatch_mode()
assert (
curr_mode is not None
), "Illegal invocation of dispatch on torch._C.DispatchKey.Python without a mode."
assert (
type(curr_mode) in self.python_key_mode_table
), f"Current active mode {curr_mode} not registered"
handler = self.python_key_mode_table[type(curr_mode)]
with _pop_mode_temporarily() as mode:
return handler(mode, *args, **kwargs)
functionality_key = torch._C._to_functionality_key(dispatch_key) # type: ignore[attr-defined]
if functionality_key == torch._C.DispatchKey.PreDispatch:
from torch.utils._python_dispatch import _pop_mode_temporarily
# The check for Python in the exclude set is so we properly respect `with no_dispatch()`
# calls inside of a mode.
if (
_len_torch_dispatch_stack_pre_dispatch() > 0
) and not torch._C._dispatch_tls_is_dispatch_key_excluded(
DispatchKey.Python
):
curr_mode = _get_current_dispatch_mode_pre_dispatch()
assert (
curr_mode is not None
), "Illegal invocation of dispatch on torch._C.DispatchKey.PreDispatch without a mode."
assert (
type(curr_mode) in self.python_key_mode_table
), f"Current active mode {curr_mode} not registered"
handler = self.python_key_mode_table[type(curr_mode)]
with _pop_mode_temporarily(functionality_key) as mode:
return handler(mode, *args, **kwargs)
final_key = resolve_key(self, dispatch_key)
# This can current fail due to backend fallbacks. You just have to
# register them by hand for HigherOrderOperator.
if final_key not in self.py_kernels:
raise NotImplementedError(
f"could not find kernel for HigherOrderOperator {self._name} "
f"at dispatch key {final_key} (resolved from {dispatch_key})"
)
# [NOTE] We shouldn't cache PreDispatch kernel here because depending
# on what modes are active, predispatch behaviour is different.
# Also we do same thing for normal ops:
# See Note [Not Caching Per-Dispatch-Key Mode Handlers]
if dispatch_key != torch._C.DispatchKey.PreDispatch:
self._dispatch_cache[dispatch_key] = self.py_kernels[final_key]
kernel = self.py_kernels[final_key]
# It's illegal to register DispatchKey to py_kernels, since there's no
# C++ kernel to call into
assert not isinstance(kernel, torch._C.DispatchKey)
return kernel(*args, **kwargs)
def __call__(self, *args, **kwargs):
# Dynamo already traces the body of HigherOrderOp beforehand when it
# so no need to trace into it.
import torch._dynamo
from torch._dynamo import disable
@disable
def wrapper():
flat_args = _to_flat_tuple(args, kwargs)
if torch.overrides.has_torch_function(flat_args):
return torch.overrides.handle_torch_function(
self, flat_args, *args, **kwargs
)
dispatch_key_set = _compute_keyset(args, kwargs, self.non_fallthrough_keys)
return self.dispatch(
dispatch_key_set.highestPriorityTypeId(), *args, **kwargs
)
return wrapper()
def __str__(self):
return f"{self.name()}"
def name(self):
return self._name
def _to_flat_tuple(args, kwargs):
return pytree.arg_tree_leaves(*args, **kwargs)
def _compute_keyset(args, kwargs, non_fallthrough_keys):
tensors = _get_tensors(args, kwargs)
return key_extractor(tensors, non_fallthrough_keys)
def _get_tensors(args, kwargs):
flat_all = _to_flat_tuple(args, kwargs)
tensor_args = [t for t in flat_all if isinstance(t, torch.Tensor)]
return tuple(tensor_args)
# Note - this should maintain identical impl to the C++ dispatcher key extraction logic
# at ATen/core/dispatch/DispatchKeyExtractor.h
def key_extractor(tensors, key_mask):
key_set = torch._C._dispatch_tls_local_include_set()
for tensor in tensors:
key_set = key_set | torch._C._dispatch_keys(tensor)
key_set = key_set - torch._C._dispatch_tls_local_exclude_set()
key_set = key_set & key_mask
return key_set
# Mode stack for PreDispatchKey
# it should always have three keys with
# priority given to FunctionalTensorMode and
# then ProxyTorchDispatchMode. It means that
# slot 0 belongs to ProxyTorchDispatchMode and
# slot 1 belongs to FunctionalTensorMode.
#
# SchemaCheckMode is separate from the other 2,
# and is only valid when the stack is empty.
# SchemaCheckMode is for testing purposes, and
# is meant to run in eager mode on concrete inputs,
# checking for incorrect schemas in regards to
# aliasing or mutating ops.
class _ModeStackStateForPreDispatch:
def __init__(self):
self.__infra_modes = [None, None]
self._schema_check_mode = None
def set(self, index, mode):
assert index < len(self.__infra_modes)
self.__infra_modes[index] = mode
def get(self, index):
assert index < len(self.__infra_modes)
return self.__infra_modes[index]
def count(self):
return len([i for i in self.__infra_modes if i is not None]) + int(
self._schema_check_mode is not None
)
_mode_stack_state_for_pre_dispatch = _ModeStackStateForPreDispatch()
def unset_mode_pre_dispatch(mode_key, schema_check=False):
current_mode_stack_pre_dispatch = mode_stack_state_for_pre_dispatch()
assert mode_key is None or mode_key in (
torch._C._TorchDispatchModeKey.PROXY,
torch._C._TorchDispatchModeKey.FUNCTIONAL,
)
if schema_check:
assert mode_key is None
def _unset_mode():
if mode_key == torch._C._TorchDispatchModeKey.PROXY:
current_mode = current_mode_stack_pre_dispatch.get(0)
mode_stack_state_for_pre_dispatch().set(0, None)
return current_mode
elif mode_key == torch._C._TorchDispatchModeKey.FUNCTIONAL:
current_mode = current_mode_stack_pre_dispatch.get(1)
mode_stack_state_for_pre_dispatch().set(1, None)
return current_mode
else:
current_mode = mode_stack_state_for_pre_dispatch()._schema_check_mode
mode_stack_state_for_pre_dispatch()._schema_check_mode = None
return current_mode
current_mode = _unset_mode()
new_pre_dispatch_len = _len_torch_dispatch_stack_pre_dispatch()
# When we are unsetting a mode, we need to check if there is
# active mode left on the PreDispatch key. If there is nothing
# active, we need to remove PreDispatch key from local dispatch include
# set.
if new_pre_dispatch_len == 0:
torch._C._dispatch_tls_set_dispatch_key_included(
torch._C.DispatchKey.PreDispatch, False
)
return current_mode
def _set_mode_pre_dispatch(mode):
from torch._subclasses.functional_tensor import FunctionalTensorMode
from torch._subclasses.schema_check_mode import SchemaCheckMode
from torch.fx.experimental.proxy_tensor import ProxyTorchDispatchMode
assert isinstance(
mode,
(
FunctionalTensorMode,
ProxyTorchDispatchMode,
SchemaCheckMode,
),
)
previous_mode_stack_len = _len_torch_dispatch_stack_pre_dispatch()
if isinstance(mode, SchemaCheckMode):
current_mode = mode_stack_state_for_pre_dispatch()._schema_check_mode
if previous_mode_stack_len > 0:
raise AssertionError(
"SchemaCheckMode for pre-dispatch must be used exclusively, found other modes on the stack"
)
mode_stack_state_for_pre_dispatch()._schema_check_mode = mode
elif isinstance(mode, FunctionalTensorMode):
current_mode = mode_stack_state_for_pre_dispatch().get(1)
assert current_mode is None
mode_stack_state_for_pre_dispatch().set(1, mode)
else:
current_mode = mode_stack_state_for_pre_dispatch().get(0)
assert current_mode is None
mode_stack_state_for_pre_dispatch().set(0, mode)
# When we are setting a mode, we need to check if there is
# active mode left on the PreDispatch key. If there was nothing
# active before setting this mode, it means that PreDispatch key
# was turned off. So we need to turn it on again.
if previous_mode_stack_len == 0:
torch._C._dispatch_tls_set_dispatch_key_included(
torch._C.DispatchKey.PreDispatch, True
)
def _pop_mode_from_pre_dispatch():
mode_stack = mode_stack_state_for_pre_dispatch()
pre_dispatch_len = _len_torch_dispatch_stack_pre_dispatch()
if pre_dispatch_len == 0:
raise AssertionError("Trying to pop empty mode stack")
if mode_stack._schema_check_mode is not None:
return unset_mode_pre_dispatch(None, schema_check=True)
if mode_stack.get(1) is not None:
return unset_mode_pre_dispatch(torch._C._TorchDispatchModeKey.FUNCTIONAL)
if mode_stack.get(0) is not None:
return unset_mode_pre_dispatch(torch._C._TorchDispatchModeKey.PROXY)
def _len_torch_dispatch_stack_pre_dispatch():
return mode_stack_state_for_pre_dispatch().count()
def _get_dispatch_mode_pre_dispatch(mode_key):
assert mode_key in (
torch._C._TorchDispatchModeKey.PROXY,
torch._C._TorchDispatchModeKey.FUNCTIONAL,
)
if mode_key == torch._C._TorchDispatchModeKey.PROXY:
return mode_stack_state_for_pre_dispatch().get(0)
else:
return mode_stack_state_for_pre_dispatch().get(1)
def _get_current_dispatch_mode_pre_dispatch():
if mode_stack_state_for_pre_dispatch()._schema_check_mode is not None:
return mode_stack_state_for_pre_dispatch()._schema_check_mode
else:
stack_len = mode_stack_state_for_pre_dispatch().count()
if stack_len == 2:
return mode_stack_state_for_pre_dispatch().get(1)
if stack_len == 1:
return (
mode_stack_state_for_pre_dispatch().get(1)
if mode_stack_state_for_pre_dispatch().get(1) is not None
else mode_stack_state_for_pre_dispatch().get(0)
)
return None
def mode_stack_state_for_pre_dispatch():
global _mode_stack_state_for_pre_dispatch
return _mode_stack_state_for_pre_dispatch
cached_ops: Set["OpOverload"] = set()
def add_cached_op(op_overload):
global cached_ops
cached_ops.add(op_overload)
def reset_cached_ops():
global cached_ops
cached_ops.clear()
def get_cached_ops():
global cached_ops
return cached_ops
# Each OpOverload object contains pointer to a a specific operator overload, a pointer to the parent `OpOverloadPacket` object.
# You can obtain an OpOverload object through attribute query on OpOverloadPacket.
class OpOverload(OperatorBase):
def __init__(self, overloadpacket, op, op_dk, schema, tags):
super().__init__()
self._op = op
self._op_dk = op_dk
self._schema = schema
self._overloadpacket = overloadpacket
self._tags = tags
self._overloadname = (
"default" if schema.overload_name == "" else schema.overload_name
)
self._name = self._schema.name
if schema.overload_name:
self._name += "." + schema.overload_name
self.__name__ = f"{self._schema.name.split('::')[1]}.{self._overloadname}"
self.__module__ = overloadpacket.__module__
op.__module__ = overloadpacket.__module__
self.__qualname__ = self._name
self.__annotations__ = {}
# Only compute the OperatorHandle when we need it. Not all OpOverloads have
# OperatorHandles (the TorchScript ones don't...)
self._lazy_handle = None
# If the OpOverload was constructed from a Library.def in Python.
self._defined_in_python = self.__qualname__ in torch.library._defs
# Logic replicated from aten/src/ATen/native/MathBitsFallback.h
is_write = None
for a in self._schema.arguments:
if a.alias_info is None:
continue
if is_write is None:
is_write = a.alias_info.is_write
else:
# We will conservatively call mixed mutable/non-mutable
# aliased inputs as NOT a view
is_write = a.alias_info.is_write or is_write
self.is_view = is_write is not None and not is_write
@property
def _namespace(self):
return self._schema.name.split("::")[0]
@property
def _opname(self):
return self._schema.name.split("::")[1]
@property
def _handle(self):
if self._lazy_handle is None:
self._lazy_handle = torch._C._dispatch_find_schema_or_throw(
self._schema.name, self._schema.overload_name
)
return self._lazy_handle
# it's a no-op since OpOverload object is immutable and must be unique for a given op overload.
def __deepcopy__(self, memo=None):
return self
def __repr__(self):
return "<OpOverload(op='{}.{}', overload='{}')>".format(
*self._schema.name.split("::"), self._overloadname
)
def __call__(self_, *args, **kwargs): # noqa: B902
# use `self_` to avoid naming collide with aten ops arguments that
# are named "self". This way, all the aten ops can be called by kwargs.
return self_._op(*args, **kwargs)
def redispatch(self_, keyset, *args, **kwargs): # noqa: B902
# use `self_` to avoid naming collide with aten ops arguments that
# are named "self". This way, all the aten ops can be called by kwargs.
return self_._handle.redispatch_boxed(keyset, *args, **kwargs)
def __hash__(self):
return hash(self._op)
# `my_namespace.my_op_name.overload_name`
def __str__(self):
return "{}.{}.{}".format(*self._schema.name.split("::"), self._overloadname)
def has_kernel_for_dispatch_key(self, k):
return super().has_kernel_for_dispatch_key(
k
) or torch._C._dispatch_has_kernel_for_dispatch_key(self.name(), k)
def has_kernel_for_any_dispatch_key(self, ks):
return torch._C._dispatch_has_kernel_for_any_dispatch_key(
self.name(), ks
) or super().has_kernel_for_any_dispatch_key(ks)
@property
def namespace(self):
return self._schema.name.split("::")[0]
def decompose(self, *args, **kwargs):
dk = torch._C.DispatchKey.CompositeImplicitAutograd
if dk in self.py_kernels:
# NB: This branch is not too necessary anymore, because we can
# apply Python CompositeImplicitAutograd *before* tracing
# using Python dispatcher (also taking advantage of the autograd
# formula). But it's included for completeness
return self.py_kernels[dk](*args, **kwargs)
elif torch._C._dispatch_has_kernel_for_dispatch_key(self.name(), dk):
return self._op_dk(dk, *args, **kwargs)
else:
return NotImplemented
# Remove a dispatch key from the dispatch cache. This will force it to get
# recomputed the next time. Does nothing
# WARNING: if you register a dispatch key to py_kernels of an OpOverload,
# calling _del_dispatch on that key is NOT sufficient to apply your change,
# because a single registration may affect MULTIPLE dispatch keys (e.g.,
# registering Autograd affects AutogradCPU). del_dispatch is to be used
# only if you are specifically modifying how get_dispatch handles a
# particular input 'key'.
def _uncache_dispatch(self, key):
self._dispatch_cache.pop(key, None)
# This implements the pre-computation logic for the Python dispatcher.
def _get_dispatch(self, key):
# This is only called upon a cache miss
assert key not in self._dispatch_cache, f"{self} {key}"
if key == torch._C.DispatchKey.Python:
if (
not isinstance(self, TorchBindOpOverload)
and not self.python_key_mode_table
):
self._dispatch_cache[key] = key
add_cached_op(self)
return key
def handler(*args, **kwargs):
from torch.utils._python_dispatch import _get_current_dispatch_mode
# TODO: We also need to handle tensor subclasses here
# TODO(voz): We should walk all the nodes here / turn it into a list, topmode is ok for now.
curr_mode = type(_get_current_dispatch_mode())
assert (
curr_mode is not None
), "Illegal invocation of dispatch on torch._C.DispatchKey.Python without a mode."
if curr_mode not in self.python_key_mode_table:
if isinstance(self, TorchBindOpOverload):
with torch.utils._python_dispatch._pop_mode_temporarily() as mode:
return torch._library.utils.handle_dispatch_mode(
mode, self, *args, **kwargs
)
else:
return self._op_dk(key, *args, **kwargs)
with torch.utils._python_dispatch._pop_mode_temporarily() as mode:
return self.python_key_mode_table[curr_mode](mode, *args, **kwargs)
self._dispatch_cache[key] = handler
add_cached_op(self)
return handler
functionality_key = torch._C._to_functionality_key(key) # type: ignore[attr-defined]
if functionality_key == torch._C.DispatchKey.PreDispatch:
curr_stack_len = _len_torch_dispatch_stack_pre_dispatch()
# The check for Python in the exclude set is so we properly respect `with no_dispatch()`
# calls inside of a mode.
if (
curr_stack_len > 0
and not torch._C._dispatch_tls_is_dispatch_key_excluded(
DispatchKey.Python
)
):
def handler(*args, **kwargs):
@contextlib.contextmanager
def _temporarily_pop_modes_from_pre_dispatch():
top_mode = _pop_mode_from_pre_dispatch()
try:
yield top_mode
finally:
_set_mode_pre_dispatch(top_mode)
with _temporarily_pop_modes_from_pre_dispatch() as curr_mode:
return torch._library.utils.handle_dispatch_mode(
curr_mode, self, *args, **kwargs
)
# Note [Not Caching Per-Dispatch-Key Mode Handlers]
# Note that we're not caching this handler. There isn't really a point, since the slow bit
# is the handler itself (in python).
# Also, not caching means that we don't have to reset the cache when any existing
# modes go out of scope (which in of itself takes time to loop through all operators).
return handler
final_key = resolve_key(self, key)
# See Note [Not Caching Per-Dispatch-Key Mode Handlers]
cache_result = key != torch._C.DispatchKey.PreDispatch
# TODO: We could potentially have lots of debugging wrappers against
# dispatch keys; design some general registration mechanism instead of
# having if statement for each of them
if key == torch._C.DispatchKey.Functionalize:
import torch._dispatch.python as pydispatch
if pydispatch.CROSSREF_FUNCTIONALIZE:
handler = pydispatch.make_crossref_functionalize(self, final_key)
if cache_result:
self._dispatch_cache[key] = handler
add_cached_op(self)
return handler
r = self.py_kernels.get(final_key, final_key)
if cache_result:
self._dispatch_cache[key] = r
add_cached_op(self)
return r
def name(self):
return self._name
@property
def overloadpacket(self):
return self._overloadpacket
@property
def op(self):
return self._op
@property
def tags(self):
return self._tags
# TODO: add more methods to expose information about input and output arguments
# TorchBindOpOverload are those custom ops which have at least one overload's
# schema consists of torch.ScriptObject (i.e. custom class) input.
# TorchBindOpOverload will skip C++ dispatcher and purely dispatched in python
# when its inputs contain FakeScriptObject in a similar way as higher order ops.
class TorchBindOpOverload(OpOverload):
def _fallthrough_keys(self) -> List[DispatchKey]:
# TODO: we should be calling the fallback for these, but a fallthrough is almost close
# enough to the fallback in most cases that we care about.
_DEFAULT_FALLTHROUGH_KEYS = [
DispatchKey.Autograd,
DispatchKey.AutogradCPU,
DispatchKey.AutogradCUDA,
DispatchKey.ADInplaceOrView,
DispatchKey.BackendSelect,
DispatchKey.PythonTLSSnapshot,
DispatchKey.PythonDispatcher,
]
def _may_use_fallthrough_instead_of_fallback(key: DispatchKey):
if torch._C._dispatch_has_kernel_for_dispatch_key(self.name(), key):
return torch._C._dispatch_kernel_for_dispatch_key_is_fallthrough(
self.name(), key
)
return (
key not in self.py_kernels
or self.py_kernels[key] is torch.library.fallthrough_kernel
)
return [
key
for key in _DEFAULT_FALLTHROUGH_KEYS
if _may_use_fallthrough_instead_of_fallback(key)
]
@contextlib.contextmanager
def _register_as_effectful_op_temporarily(self):
from torch._higher_order_ops.effects import (
_EffectType,
_register_effectful_op,
SIDE_EFFECTS,
)
try:
if self not in SIDE_EFFECTS:
_register_effectful_op(self, _EffectType.ORDERED)
yield
finally:
if self in SIDE_EFFECTS:
del SIDE_EFFECTS[self]
# use `self_` to avoid naming collide with arguments that
# are named "self". This way, they can be called by kwargs.
def __call__(self_, *args, **kwargs): # noqa: B902
if _must_dispatch_in_python(args, kwargs):
# When any inputs are FakeScriptObject, we need to
# skip c++ dispatcher and dispatch in python through _get_dispatch of python_dispatcher
# because C++ dispatcher will check the schema and cannot recognize FakeScriptObject.
#
# Note:
# 1. We only register the torchbind op temporarily as effectful op because we only want
# the effect token functionalization logic to be applied during tracing. Otherwise, the behavior
# of the eagerly executing the op might change after tracing.
# 2. We don't want to register the op as effectful for all torchbind ops in ctor because this might
# cause unexpected behavior for some autograd.profiler ops e.g. profiler._record_function_exit._RecordFunction.
with self_._register_as_effectful_op_temporarily():
return self_._dispatch_in_python(
args, kwargs, self_._fallthrough_keys()
)
return self_._op(*args, **kwargs)
def _dispatch_in_python(self, args, kwargs, fallthrough_keys):
non_fallthrough_keys = torch._C._dispatch_keyset_full()
for key in fallthrough_keys:
non_fallthrough_keys = non_fallthrough_keys.remove(key)
dispatch_key_set = _compute_keyset(args, kwargs, non_fallthrough_keys)
dispatch_key = dispatch_key_set.highestPriorityTypeId()
handler = (
self._get_dispatch(dispatch_key)
if dispatch_key not in self._dispatch_cache
else self._dispatch_cache[dispatch_key]
)
if isinstance(handler, DispatchKey):
# fallthrough keys can be registered at runtime via torch.library.impl
# so need to add it to fallthrough_keys and re-dispatch.
if torch._C._dispatch_kernel_for_dispatch_key_is_fallthrough(
self.name(), dispatch_key
):
return self._dispatch_in_python(
args, kwargs, fallthrough_keys + [dispatch_key]
)
raise RuntimeError(
f"Torchbind op {self} received a FakeScriptObject input when dispatching {handler}."
f" but no python implementation is found."
f" Please file an issue on this when you encounter this error."
f" This error can happen when you export or compile the model."
f" It can still happpen even if a C++ implementation for {dispatch_key}. "
f" has been registered. That's because FakeScriptObject purely lives in python and cannot work "
f" with a C++ implementation."
)
assert isinstance(handler, Callable) # type: ignore[arg-type]
return handler(*args, **kwargs)
def _must_dispatch_in_python(args, kwargs):
return pytree.tree_any(
lambda obj: isinstance(
obj, torch._library.fake_class_registry.FakeScriptObject
),
(args, kwargs),
)
def _has_script_object_arg(schema: torch.FunctionSchema) -> bool:
return any(isinstance(arg.type, torch.ClassType) for arg in schema.arguments)
# OpOverloadPacket class contains pointer to a base unresolved operator that doesn't correspond to a specific operator
# You can obtain an OpOverload object through attribute query.
class OpOverloadPacket:
def __init__(self, qualified_op_name, op_name, op, overload_names):
# These attributes are accessible on the object through the properties
# defined below but are immutable
self._qualified_op_name = qualified_op_name
self.__name__ = op_name
self._op = op
self._overload_names = overload_names
self._dir = []
self._has_torchbind_op_overload = any(
_has_script_object_arg(schema) for schema in self._schemas.values()
)
# it's a no-op since OpOverloadPacket object is immutable and must be unique for a given op.
def __deepcopy__(self, memo=None):
return self
def __repr__(self):
return "<OpOverloadPacket(op='{}.{}')>".format(
*self._qualified_op_name.split("::")
)
def __hash__(self):
return hash(self._op)
def __str__(self):
return "{}.{}".format(*self._qualified_op_name.split("::"))
@property
def op(self):
return self._op
@property
def _schemas(self):
return {
overload_name: torch._C._get_schema(self._qualified_op_name, overload_name)
for overload_name in self._overload_names
}
def __getattr__(self, key):
# It is not a valid op_name when __file__ is passed in
if key == "__file__":
return "torch.ops"
# ensure that query for dunder attributes that does not exist on
# opoverloadpacket but instead exists on the self._op object does not unnecessarily call
# `_get_operation_overload` (which is an expensive operation).
# This is done to prevent any potential slowdown. This list can be extended
# if there exists other attributes like `__name__` that only exist on self._op and not on the
# opoverloadpacket.
# This is ok since we are guaranteed that an overload name for an aten op can't start with '__'
try:
if key.startswith("__"):
return getattr(self._op, key)
except AttributeError:
# for consistency because it seems weird to
# throw an attribute error with a message containing
# an object name different from the one the attribute
# query was performed on.
raise AttributeError(
f"'{str(self)}' can't have an overload name beginning with '__' and the "
f"underlying op {str(self._op)} has no attribute {key} either."
) from None
try:
# This is ok since we are guaranteed that an overload name for an aten op can't be 'default'
use_key = "" if key == "default" else key
# TODO: disallow access to overloads registered by JIT
op_, op_dk_, tags = torch._C._get_operation_overload(
self._qualified_op_name, use_key
)
schema = torch._C._get_schema(self._qualified_op_name, use_key)
overload = (
OpOverload(self, op_, op_dk_, schema, tags)
if not _has_script_object_arg(schema)
else TorchBindOpOverload(self, op_, op_dk_, schema, tags)
)
# cache the overload object
setattr(self, key, overload)
self._dir.append(key)
return overload
except RuntimeError:
raise AttributeError(
f"The underlying op of '{str(self)}' has no overload name '{key}'"
) from None
def __iter__(self):
return iter(self._dir)
def __call__(self_, *args, **kwargs): # noqa: B902
# use `self_` to avoid naming collide with aten ops arguments that
# named "self". This way, all the aten ops can be called by kwargs.
# overloading __call__ to ensure torch.ops.foo.bar()
# is still callable from JIT
# We save the function ptr as the `op` attribute on
# OpOverloadPacket to access it here.
# Directly calling OverloadPacket goes into C++, which will check
# the schema and cause an error for torchbind op when inputs consist of FakeScriptObject so we
# intercept it here and call TorchBindOpverload instead.
if self_._has_torchbind_op_overload and _must_dispatch_in_python(args, kwargs):
return _call_overload_packet_from_python(self_, args, kwargs)
return self_._op(*args, **(kwargs or {}))
# TODO: use this to make a __dir__
def overloads(self):
return [n if n else "default" for n in self._overload_names]
# Note - this mirrors the logic of the cpp_function defined in jit/python/init.cpp
# _jit_get_operations, which calls _get_operation_for_overload_or_packet.
def _call_overload_packet_from_python(op: OpOverloadPacket, args, kwargs):
# Re-use the torch function handling logic in cpp
torch_function_called, ret = torch._C._maybe_call_torch_function_for_op_packet(
op, *args, **kwargs
)
if torch_function_called:
return ret
# The following mirrors getOpWithStack.
# In cpp, we do a schema matching for the arguments, and call ToIValue to
# to check whether the arguments are valid. But need to do similar things here
# and check the schema whether the FakeScriptObject is the corresponding fake class
# of the actual class used in schema.
exceptions = {}
found_op = None
for overload_name in op.overloads():
op_overload = getattr(op, overload_name)
try:
_ = torch._C._check_schema_allow_fake_script_object(
op_overload._schema, *args, **kwargs
)
found_op = op_overload
break
except RuntimeError as e:
exceptions[overload_name] = e
if found_op:
return found_op(*args, **kwargs)
err_msg = (
f"Fail to match any TorchBindOverload of {op} with following exceptions:\n"
)
for i, (key, msg) in enumerate(exceptions.items()):
err_msg += f"Overload name {key}:\n {msg}\n"
raise RuntimeError(err_msg)
# Resolution of torch.fn is different from torch.ops.aten.fn
# torch.fn uses the Python argparser, matches with the
# appropriate schema, and calls into the unboxed version of the method
# torch.ops.aten.fn resolution is done via the mechanism defined in JIT.
# JIT creates a stack of all the overloads and then tries to match the
# correct one at runtime and always calls into the boxed version of the method
# Autograd codegen creates VariableType, TracerType,
# inplace or view type and python bindings.
# Aten codegen generates tensor methods for the tensor class.
# _OpNamespace is a subclass of ModuleType because the torch script
# allows attribute lookups on modules only. Since we want torch.ops.foo.bar()
# to work from script, we need to ensure ops and foo are modules
class _OpNamespace(types.ModuleType):
"""
An op namespace to dynamically bind Operators into Python.
Say a user has created a custom Operator called "my_namespace::my_op". To
call this op, the user will write torch.ops.my_namespace.my_op(...).
At startup, this operation will not yet be bound into Python. Instead, the
following sequence of magic tricks will occur:
1. `torch.ops.my_namespace` will invoke the `__getattr__` magic method
on the `torch.ops` object, which will create a new `_OpNamespace`
object called `my_namespace` and set it as an attribute on the `ops`
object.
2. `torch.ops.my_namespace.my_op` will then invoke `__getattr__` on
the `my_namespace` object, which will retrieve the operation via
`torch.get_operation`, a function bound from C++, and then in a similar
fashion bind this new object onto the `my_namespace` object.
3. `torch.ops.my_namespace.my_op(...)` then calls this new operation
and subsequent accesses will incur no further lookup (the namespace and
operation will already exist).
"""
def __init__(self, name):
super().__init__("torch.ops." + name)
self.name = name
self._dir = []
def __iter__(self):
return iter(self._dir)
def __getattr__(self, op_name):
# It is not a valid op_name when __file__ is passed in
if op_name == "__file__":
return "torch.ops"
elif op_name in ["__origin__", "__self__"]:
raise AttributeError(
f"Invalid attribute '{op_name}' for '_OpNamespace' '{self.name}'"
)
# Get the op `my_namespace::my_op` if available. This will also check
# for overloads and raise an exception if there are more than one.
namespace_name = self.name
qualified_op_name = f"{namespace_name}::{op_name}"
module_name = self.__module__ + "." + namespace_name
try:
op, overload_names = _get_packet(qualified_op_name, module_name)
if op is None:
raise AttributeError(
f"'_OpNamespace' '{self.name}' object has no attribute '{op_name}'"
)
except RuntimeError as e:
# Turn this into AttributeError so getattr(obj, key, default)
# works (this is called by TorchScript with __origin__)
raise AttributeError(
f"'_OpNamespace' '{self.name}' object has no attribute '{op_name}'"
) from e
op.__module__ = module_name
opoverloadpacket = OpOverloadPacket(
qualified_op_name, op_name, op, overload_names
)
opoverloadpacket.__module__ = self.__module__ + "." + namespace_name
# cache the opoverloadpacket to ensure that each op corresponds to
# a unique OpOverloadPacket object
setattr(self, op_name, opoverloadpacket)
self._dir.append(op_name)
return opoverloadpacket
def _get_packet(qualname, op_module):
op, overload_names = torch._C._jit_get_operation(qualname)
if op is not None:
# let the script frontend know that op is identical to the builtin op
# with qualified_op_name
torch.jit._builtins._register_builtin(op, qualname)
op.__module__ = op_module
return op, overload_names
def _refresh_packet(packet):
op, overload_names = _get_packet(packet._qualified_op_name, packet._op.__module__)
assert op is not None
packet._op = op
packet._overload_names = overload_names
class _PyOpNamespace(_OpNamespace):
def __init__(self, name, ops):
super().__init__(name)
self._ops = ops
def __getattr__(self, name):
# Following _OpNamespace.__getattr__, we cache the op on the _PyOpNamespace object.
op = self._ops.get(name, None)
if op is None:
raise AttributeError(
f"'_PyOpNamespace' '{self.name}' object has no attribute '{name}'"
)
setattr(self, name, op)
return op
class _Ops(types.ModuleType):
__file__ = "_ops.py"
def __init__(self):
super().__init__("torch.ops")
self.loaded_libraries = set()
self._higher_order_op_namespace = _PyOpNamespace(
"torch.ops.higher_order", _higher_order_ops
)
self._dir = []
def __getattr__(self, name):
# Check if the name is a HigherOrderOperator
if name == "higher_order":
return self._higher_order_op_namespace
# Here we are creating `torch.ops.my_namespace`
namespace = _OpNamespace(name)
setattr(self, name, namespace)
self._dir.append(name)
return namespace
def __iter__(self):
return iter(self._dir)
def import_module(self, module):
"""
Imports a Python module that has torch.library registrations.
Generally, to extend PyTorch with custom operators, a user will
create a Python module whose import triggers registration of
the custom operators via a torch.ops.load_library call or a call
to one or more torch.library.* APIs.
It is unexpected for Python modules to have side effects, so some
linters and formatters will complain. Use this API to import Python
modules that contain these torch.library side effects.
Args:
module (str): The name of the Python module to import
"""
importlib.import_module(module)
def load_library(self, path):
"""
Loads a shared library from the given path into the current process.
The library being loaded may run global initialization code to register
custom operators with the PyTorch JIT runtime. This allows dynamically
loading custom operators. For this, you should compile your operator
and the static registration code into a shared library object, and then
call ``torch.ops.load_library('path/to/libcustom.so')`` to load the
shared object.
After the library is loaded, it is added to the
``torch.ops.loaded_libraries`` attribute, a set that may be inspected
for the paths of all libraries loaded using this function.
Args:
path (str): A path to a shared library to load.
"""
if torch._running_with_deploy():
return
path = _utils_internal.resolve_library_path(path)
with dl_open_guard():
# Import the shared library into the process, thus running its
# static (global) initialization code in order to register custom
# operators with the JIT.
ctypes.CDLL(path)
self.loaded_libraries.add(path)
# The ops "namespace"
ops: _Ops = _Ops()