import contextlib import ctypes import inspect import sys import types from abc import ABC from typing import Any, Dict import torch._C from torch import _utils_internal from torch._functorch.pyfunctorch import dispatch_functorch # 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) def has_key(op, k): return ( torch._C._dispatch_has_kernel_for_dispatch_key(op.name(), k) or k in op.py_kernels ) # TODO(voz) We are missing an entire axis of registration - Modes for the python key class PyOperatorABC(ABC): def __call__(self, *args, **kwargs): pass def py_impl(self, dispatch_key, fn): pass def name(self): pass is_included_in_alias = torch._C._dispatch_is_included_in_alias DispatchKey = torch._C.DispatchKey # Equivalent to computeDispatchTableEntryWithDebug def resolve_key(op: PyOperatorABC, k: DispatchKey): # type: ignore[valid-type] # 1. (Direct) operator registration if has_key(op, 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 has_key( op, 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 has_key( op, cand ): return cand has_backend_kernel = torch._C._dispatch_has_kernel_for_any_dispatch_key( op.name(), torch._C._dispatch_get_backend_keyset_from_autograd(k) ) or has_key(op, DispatchKey.CompositeExplicitAutograd) # 2.3. Use CompositeImplicitAutograd kernel if available cand = DispatchKey.CompositeImplicitAutogradNestedTensor if ( (k != DispatchKey.Undefined and is_included_in_alias(k, cand)) and has_key(op, cand) and not has_backend_kernel ): return cand cand = DispatchKey.CompositeImplicitAutograd if (k == DispatchKey.Undefined or is_included_in_alias(k, cand)) and has_key( op, cand ): if ( k == DispatchKey.AutogradOther and torch._C._dispatch_has_kernel_for_any_dispatch_key( op.name(), 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 has_key(op, 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}") pyop_namespace = {} class PyOperator(PyOperatorABC): def __init__(self, name): self._name = name self.table = {} self.python_key_mode_table = {} self.functorch_table = {} # Make _OPNamespace not scream, this whole name based association needs a good hard look self.__name__ = name pyop_namespace[name] = self def fallthrough(self, dispatch_key): self.table[dispatch_key] = self._fallthrough_fn(self, dispatch_key) def py_impl(self, dispatch_key_or_mode_or_transform): def inner(fn): if inspect.isclass(dispatch_key_or_mode_or_transform) and issubclass( dispatch_key_or_mode_or_transform, torch.utils._python_dispatch.TorchDispatchMode, ): mode = dispatch_key_or_mode_or_transform assert mode 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[mode] = fn return fn if isinstance( dispatch_key_or_mode_or_transform, torch._C._functorch.TransformType ): transform = dispatch_key_or_mode_or_transform self.functorch_table[transform] = fn return fn dispatch_key = dispatch_key_or_mode_or_transform assert ( dispatch_key != torch._C.DispatchKey.Python ), "Please register a mode for the torch._C.DispatchKey.Python key instead." assert isinstance(dispatch_key, torch._C.DispatchKey) assert dispatch_key not in self.table self.table[dispatch_key] = fn return fn return inner def dispatch(self, dispatch_key, *args, **kwargs): from torch.utils._python_dispatch import _get_current_dispatch_mode if dispatch_key == torch._C.DispatchKey.FuncTorchDynamicLayerFrontMode: return dispatch_functorch(self, args, kwargs) if dispatch_key == torch._C.DispatchKey.Python: # TODO(voz): We should walk all the nodes here / turn it into a list, topmode is ok for now. 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" # TODO(voz): The idea behind this is that we do not yet support dispatch by key + mode, only key. return self.python_key_mode_table[type(curr_mode)](*args, **kwargs) assert dispatch_key in self.table, dispatch_key return self.table[dispatch_key](*args, **kwargs) def __call__(self, *args, **kwargs): 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) return self.dispatch(dispatch_key_set.highestPriorityTypeId(), *args, **kwargs) def name(self): return self.name # TODO(voz): Should rewrite fallthrough register as the impl for keys we do not specify # as opposed to being this sort of explicit thing where ops are a little too key aware... def _fallthrough_fn(self, operator, dispatch_key): def inner(*args, **kwargs): all_keys_after_current = torch._C._dispatch_keyset_full_after(dispatch_key) all_keys_after_current_masked = all_keys_after_current & _compute_keyset( args, kwargs ) return self.dispatch( all_keys_after_current_masked.highestPriorityTypeId(), *args, **kwargs ) return inner def _to_flat_tuple(args, kwargs): flat_args, _ = torch.utils._pytree.tree_flatten(args) flat_kwargs, _ = torch.utils._pytree.tree_flatten(kwargs) flat_all = flat_args + flat_kwargs return flat_all def _compute_keyset(args, kwargs): tensors = _get_tensors(args, kwargs) return key_extractor(tensors) 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_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() return key_set # 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(PyOperatorABC): def __init__(self, overloadpacket, op, op_dk, schema, tags): 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.py_kernels: Dict[torch._C.DispatchKey, Any] = {} # type: ignore[name-defined] self.__name__ = "{}.{}".format( self._schema.name.split("::")[1], self._overloadname ) # TODO(voz): Lots of shared logic around python_key_mode_table, maybe pull into base... self.python_key_mode_table = {} self.__module__ = overloadpacket.__module__ op.__module__ = overloadpacket.__module__ self.__qualname__ = self._name self.__annotations__ = {} # NB: This name is hard-coded in torch/csrc/autograd/python_variable.cpp self._dispatch_cache = {} # 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 # 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 "".format( *self._schema.name.split("::"), self._overloadname ) def __call__(self, *args, **kwargs): return self._op(*args, **kwargs or {}) 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) @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 def py_impl(self, dispatch_key_or_mode): def inner(fn): if inspect.isclass(dispatch_key_or_mode) and issubclass( dispatch_key_or_mode, torch.utils._python_dispatch.TorchDispatchMode ): mode = dispatch_key_or_mode assert mode 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[mode] = fn self._dispatch_cache.clear() return fn assert isinstance(dispatch_key_or_mode, torch._C.DispatchKey) assert ( dispatch_key_or_mode != torch._C.DispatchKey.Python ), "Please register a mode for the torch._C.DispatchKey.Python key instead." if dispatch_key_or_mode in self.py_kernels: raise RuntimeError( f"Trying to override a python impl for {dispatch_key_or_mode} on operator {self._name}" ) self.py_kernels[dispatch_key_or_mode] = fn self._dispatch_cache.clear() return fn return inner # 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 self.python_key_mode_table: self._dispatch_cache[key] = key 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: # TODO: This path is slow, should generally encourage this # case to not happen return self._op_dk(key, *args, **kwargs) # TODO(voz): The idea behind this is that we do not yet support dispatch by key + mode, only key. return self.python_key_mode_table[curr_mode](*args, **kwargs) self._dispatch_cache[key] = handler return handler final_key = resolve_key(self, key) # 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) self._dispatch_cache[key] = handler return handler # print(self, key, final_key) r = self.py_kernels.get(final_key, final_key) self._dispatch_cache[key] = r 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 # 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 = [] # 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 "".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 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( "'{}' can't have an overload name beginning with '__' and the " "underlying op {} has no attribute {} either.".format( str(self), str(self._op), key ) ) 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) # cache the overload object setattr(self, key, overload) self._dir.append(key) return overload except RuntimeError: raise AttributeError( "The underlying op of '{}' has no overload name '{}'".format( str(self), key ) ) from None def __iter__(self): return iter(self._dir) def __call__(self, *args, **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. 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] # 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 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 == "__origin__": raise AttributeError() # 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 = "{}::{}".format(namespace_name, op_name) try: op, overload_names = torch._C._jit_get_operation(qualified_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 # let the script frontend know that op is identical to the builtin op # with qualified_op_name torch.jit._builtins._register_builtin(op, qualified_op_name) op.__module__ = self.__module__ + "." + namespace_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 class _PyOpNamespace(_OpNamespace): def __init__(self): super().__init__("torch.ops") self.pyop_namespace = pyop_namespace class _Ops(types.ModuleType): __file__ = "_ops.py" def __init__(self): super().__init__("torch.ops") self.loaded_libraries = set() self.pyops = _PyOpNamespace() self._dir = [] def __getattr__(self, name): # Check if the name is a pyop if name in self.pyops.pyop_namespace: return self.pyops.pyop_namespace[name] # 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 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 sys.executable == "torch_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()