FM / torch

/usr/local/lib64/python3.6/site-packages/torch/

/usr/local/lib64/python3.6/site-packages/torch

Name	Size	Mode	Actions
ao/	-	0755	rm
autograd/	-	0755	rm
backends/	-	0755	rm
bin/	-	0755	rm
contrib/	-	0755	rm
cpu/	-	0755	rm
cuda/	-	0755	rm
distributed/	-	0755	rm
distributions/	-	0755	rm
fft/	-	0755	rm
for_onnx/	-	0755	rm
futures/	-	0755	rm
fx/	-	0755	rm
include/	-	0755	rm
jit/	-	0755	rm
lib/	-	0755	rm
linalg/	-	0755	rm
multiprocessing/	-	0755	rm
nn/	-	0755	rm
onnx/	-	0755	rm
optim/	-	0755	rm
package/	-	0755	rm
profiler/	-	0755	rm
quantization/	-	0755	rm
share/	-	0755	rm
sparse/	-	0755	rm
special/	-	0755	rm
testing/	-	0755	rm
utils/	-	0755	rm
_C/	-	0755	rm
__pycache__/	-	0755	rm
autocast_mode.py	9663	0644	edit dl rm
functional.py	71257	0644	edit dl rm
hub.py	23371	0644	edit dl rm
overrides.py	84146	0644	edit dl rm
py.typed	0	0644	edit dl rm
quasirandom.py	7489	0644	edit dl rm
random.py	4828	0644	edit dl rm
serialization.py	36544	0644	edit dl rm
storage.py	5762	0644	edit dl rm
torch_version.py	3468	0644	edit dl rm
types.py	1552	0644	edit dl rm
version.py	125	0644	edit dl rm
_appdirs.py	26245	0644	edit dl rm
_C.cpython-36m-x86_64-linux-gnu.so	29296	0755	edit dl rm
_classes.py	1717	0644	edit dl rm
_deploy.py	3100	0644	edit dl rm
_dl.cpython-36m-x86_64-linux-gnu.so	29832	0755	edit dl rm
_jit_internal.py	46910	0644	edit dl rm
_linalg_utils.py	2373	0644	edit dl rm
_lobpcg.py	44076	0644	edit dl rm
_lowrank.py	11031	0644	edit dl rm
_namedtensor_internals.py	5351	0644	edit dl rm
_ops.py	4457	0644	edit dl rm
_python_dispatcher.py	7011	0644	edit dl rm
_six.py	1858	0644	edit dl rm
_sources.py	3882	0644	edit dl rm
_storage_docs.py	1298	0644	edit dl rm
_tensor.py	50854	0644	edit dl rm
_tensor_docs.py	113474	0644	edit dl rm
_tensor_str.py	18091	0644	edit dl rm
_torch_docs.py	364678	0644	edit dl rm
_utils.py	21617	0644	edit dl rm
_utils_internal.py	1687	0644	edit dl rm
_VF.py	656	0644	edit dl rm
_vmap_internals.py	13219	0644	edit dl rm
__config__.py	551	0644	edit dl rm
__future__.py	813	0644	edit dl rm
__init__.py	31199	0644	edit dl rm

Edit: /usr/local/lib64/python3.6/site-packages/torch/_ops.py (4457B)

import torch._C

import contextlib
import ctypes
import sys
import types

import torch.jit
import torch._utils_internal

# 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 _SET_GLOBAL_FLAGS:
        old_flags = sys.getdlopenflags()
        sys.setdlopenflags(old_flags | ctypes.RTLD_GLOBAL)
    yield
    if _SET_GLOBAL_FLAGS:
        sys.setdlopenflags(old_flags)

# _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(_OpNamespace, self).__init__('torch.ops.' + name)
        self.name = name

def __getattr__(self, op_name):
        # It is not a valid op_name when __file__ is passed in
        if op_name == '__file__':
            return 'torch.ops'
        # 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.
        qualified_op_name = '{}::{}'.format(self.name, op_name)
        op = torch._C._jit_get_operation(qualified_op_name)
        # 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)
        setattr(self, op_name, op)
        op.__module__ = self.__module__ + "." + self.name
        return op

class _Ops(types.ModuleType):
    __file__ = '_ops.py'

def __init__(self):
        super(_Ops, self).__init__('torch.ops')
        self.loaded_libraries = set()

def __getattr__(self, name):
        # Here we are creating `torch.ops.my_namespace`
        namespace = _OpNamespace(name)
        setattr(self, name, namespace)
        return namespace

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 = torch._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()

cmd: