# default generator. The race from multiple RPC threads could mess up
        # the draw order from the default RNG instance, leading to
        # non-deterministic behavior. Hence, create a dedicated RNG here.
        g_cpu = torch.Generator()
        g_cpu.manual_seed(0)
        self.w = torch.rand((3, 3), requires_grad=requires_grad, generator=g_cpu)

    def forward(self, t1):
        return torch.mm(self.w, t1)

            

Reported by Pylint.

                      # non-deterministic behavior. Hence, create a dedicated RNG here.
        g_cpu = torch.Generator()
        g_cpu.manual_seed(0)
        self.w = torch.rand((3, 3), requires_grad=requires_grad, generator=g_cpu)

    def forward(self, t1):
        return torch.mm(self.w, t1)

    def get_w(self):

            

Reported by Pylint.

                      self.w = torch.rand((3, 3), requires_grad=requires_grad, generator=g_cpu)

    def forward(self, t1):
        return torch.mm(self.w, t1)

    def get_w(self):
        return self.w



            

Reported by Pylint.

                  @dist_init()
    def test_dist_optim_exception(self):
        # distributed version
        owner1 = "worker%d" % ((self.rank + 1) % self.world_size)
        owner2 = "worker%d" % ((self.rank + 2) % self.world_size)

        remote_module1 = rpc.remote(owner1, MyModule)
        remote_module2 = rpc.remote(owner2, MyModule)
        remote_param1 = remote_method(MyModule.get_w, remote_module1)

            

Reported by Pylint.

                  def test_dist_optim_exception(self):
        # distributed version
        owner1 = "worker%d" % ((self.rank + 1) % self.world_size)
        owner2 = "worker%d" % ((self.rank + 2) % self.world_size)

        remote_module1 = rpc.remote(owner1, MyModule)
        remote_module2 = rpc.remote(owner2, MyModule)
        remote_param1 = remote_method(MyModule.get_w, remote_module1)
        remote_param2 = remote_method(MyModule.get_w, remote_module2)

            

Reported by Pylint.

                      )

        with dist_autograd.context() as context_id:
            g_cpu = torch.Generator()
            g_cpu.manual_seed(0)
            t1 = torch.rand((3, 3), requires_grad=True, generator=g_cpu)
            t2 = torch.rand((3, 3), requires_grad=True, generator=g_cpu)
            output1 = rpc_async_method(MyModule.forward, remote_module1, t2)
            output2 = rpc_async_method(MyModule.forward, remote_module2, output1.wait())

            

Reported by Pylint.

                      with dist_autograd.context() as context_id:
            g_cpu = torch.Generator()
            g_cpu.manual_seed(0)
            t1 = torch.rand((3, 3), requires_grad=True, generator=g_cpu)
            t2 = torch.rand((3, 3), requires_grad=True, generator=g_cpu)
            output1 = rpc_async_method(MyModule.forward, remote_module1, t2)
            output2 = rpc_async_method(MyModule.forward, remote_module2, output1.wait())
            loss = torch.add(output2.wait(), t1).sum()


            

Reported by Pylint.

                          g_cpu = torch.Generator()
            g_cpu.manual_seed(0)
            t1 = torch.rand((3, 3), requires_grad=True, generator=g_cpu)
            t2 = torch.rand((3, 3), requires_grad=True, generator=g_cpu)
            output1 = rpc_async_method(MyModule.forward, remote_module1, t2)
            output2 = rpc_async_method(MyModule.forward, remote_module2, output1.wait())
            loss = torch.add(output2.wait(), t1).sum()

            dist_autograd.backward(context_id, [loss])

            

Reported by Pylint.

                          t2 = torch.rand((3, 3), requires_grad=True, generator=g_cpu)
            output1 = rpc_async_method(MyModule.forward, remote_module1, t2)
            output2 = rpc_async_method(MyModule.forward, remote_module2, output1.wait())
            loss = torch.add(output2.wait(), t1).sum()

            dist_autograd.backward(context_id, [loss])
            with self.assertRaisesRegex(Exception, "Error running optimizer"):
                dist_optim.step(context_id)


            

Reported by Pylint.

                          loss = torch.add(output2.wait(), t1).sum()

            dist_autograd.backward(context_id, [loss])
            with self.assertRaisesRegex(Exception, "Error running optimizer"):
                dist_optim.step(context_id)

    @dist_init()
    def test_dist_optim_exception_on_constructor(self):
        # distributed version

            

Reported by Pylint.

              
import torch
import torch.distributed.rpc as rpc
from torch import Tensor, device, dtype, nn
from torch.distributed.nn.jit import instantiator
from torch.distributed import _remote_device
from torch.distributed.rpc.internal import _internal_rpc_pickler
from torch.nn import Module
from torch.nn.parameter import Parameter

            

Reported by Pylint.

              
import torch
import torch.distributed.rpc as rpc
from torch import Tensor, device, dtype, nn
from torch.distributed.nn.jit import instantiator
from torch.distributed import _remote_device
from torch.distributed.rpc.internal import _internal_rpc_pickler
from torch.nn import Module
from torch.nn.parameter import Parameter

            

Reported by Pylint.

                      # If ``enable_moving_cpu_tensors_to_cuda`` is true, but the device map is not set,
        # then any CPU tensors can still be moved to a cuda device to run forward,
        # but the output must be moved back to CPU before being sent over the wire.
        enable_moving_cpu_tensors_to_cuda = torch.device(self.device).type == "cuda"
        return enable_moving_cpu_tensors_to_cuda

    def _init_template(self, module_interface_cls, enable_moving_cpu_tensors_to_cuda):
        """
        Instantiates template on local side.

            

Reported by Pylint.

                  )


def _create_module(module_cls, args, kwargs, device):
    module = module_cls(*args, **kwargs)
    if not isinstance(module, nn.Module):
        raise ValueError(
            "Expect `module_cls(*args, **kwargs)` returns an instance of <class nn.Module>, "
            f"but it returns an instance of {type(module)}."

            

Reported by Pylint.

              

def _create_module_with_interface(
    module_cls, args, kwargs, device, module_interface_cls
):
    module = _create_module(module_cls, args, kwargs, device)
    if module_interface_cls is not None:
        module = torch.jit.script(module)
    return rpc.RRef(module, module_interface_cls)

            

Reported by Pylint.

                  raise ValueError("Method ``{}`` not supported for RemoteModule".format(name))


class _RemoteModule(nn.Module):
    def __init__(
        self,
        remote_device: str,
        module_cls: Type[nn.Module],
        args: Tuple = None,

            

Reported by Pylint.

                          # Create the module on the remote side.
            fut.wait()  # Ensure remote_module_cls is available on remote side.

            # TODO: We need to change this to rpc.remote, and make it async (see the else branch below).
            # For that we need to be able to apply _module_interface_cls to the RRef returned by rpc.remote
            # See https://github.com/pytorch/pytorch/issues/58098 for more context.
            self.module_rref = rpc.rpc_sync(
                self.on,
                _create_module_with_interface,

            

Reported by Pylint.

                  def apply(self: T, fn: Callable[[Module], None]) -> T:  # type: ignore[return]
        _raise_not_supported(self.apply.__name__)

    def cuda(self: T, device: Optional[Union[int, device]] = None) -> T:  # type: ignore[return]
        _raise_not_supported(self.cuda.__name__)

    def xpu(self: T, device: Optional[Union[int, device]] = None) -> T:  # type: ignore[return]
        _raise_not_supported(self.xpu.__name__)


            

Reported by Pylint.

                  def cuda(self: T, device: Optional[Union[int, device]] = None) -> T:  # type: ignore[return]
        _raise_not_supported(self.cuda.__name__)

    def xpu(self: T, device: Optional[Union[int, device]] = None) -> T:  # type: ignore[return]
        _raise_not_supported(self.xpu.__name__)

    def cpu(self: T) -> T:  # type: ignore[return]
        _raise_not_supported(self.cpu.__name__)


            

Reported by Pylint.

                  def bfloat16(self: T) -> T:  # type: ignore[return]
        _raise_not_supported(self.bfloat16.__name__)

    def to(self, *args, **kwargs) -> T:  # type: ignore[return]
        _raise_not_supported(self.to.__name__)

    def register_backward_hook(  # type: ignore[return]
        self, hook: Callable[[Module, _grad_t, _grad_t], Union[None, Tensor]]
    ) -> RemovableHandle:

            

Reported by Pylint.

              from torch.fx.graph import Graph, Node
from torch.quantization.fx.utils import get_new_attr_name_with_prefix

from .utils import (
    get_node_first_input_and_output_type,
    getattr_from_fqn,
    NodeInputOrOutputType,
    return_first_non_observer_node,
    get_number_of_non_param_args,

            

Reported by Pylint.

                  get_node_input_qparams,
)

from .ns_types import (
    NSSingleResultValuesType,
    NSSubgraph,
    NSNodeTargetType,
)
from torch.quantization.ns.mappings import (

            

Reported by Pylint.

                      'get_attr', zero_point_node_name, (), {}, zero_point_node_name)
    # create the quantize_per_tensor call
    return graph_c.create_node(
        'call_function', torch.quantize_per_tensor,
        (prev_node_c, scale_node, zero_point_node, torch.quint8), {},
        dtype_cast_name)

def _insert_dtype_cast_after_node(
    node_a: Node,

            

Reported by Pylint.

                  # create the quantize_per_tensor call
    return graph_c.create_node(
        'call_function', torch.quantize_per_tensor,
        (prev_node_c, scale_node, zero_point_node, torch.quint8), {},
        dtype_cast_name)

def _insert_dtype_cast_after_node(
    node_a: Node,
    node_c: Node,

            

Reported by Pylint.

                      (node_input_type_a == NodeInputOrOutputType.FP32 and
         node_input_type_c == NodeInputOrOutputType.FP32_OR_INT8)
    ):
        dtype_cast_op = torch.dequantize
    elif (
        node_input_type_a == node_input_type_c and
        node_input_type_a != NodeInputOrOutputType.UNKNOWN
    ):
        dtype_cast_mod_cls = torch.nn.Identity

            

Reported by Pylint.

                      node_a_input_qparams = get_node_input_qparams(
            node_a, gm_a, node_type_to_io_type_map)
        if node_a_input_qparams is not None:
            dtype_cast_op = torch.quantize_per_tensor  # type: ignore[assignment]
            dtype_cast_scale, dtype_cast_zero_point = node_a_input_qparams
    else:
        raise AssertionError(
            f"dtype cast from {node_input_type_c} {node_c.format_node()} to " +
            f"{node_input_type_a} {node_a.format_node()} needs to be implemented")

            

Reported by Pylint.

                          if is_activation_post_process(module):
                assert isinstance(node.args[0], Node)
                node_to_use_for_fqn = node.args[0]
        fqn = gm._node_name_to_scope[node_to_use_for_fqn.name][0]  # type: ignore[index]
    return fqn  # type: ignore[return-value]

def _insert_logger_after_node(
    node: Node,
    gm: GraphModule,

            

Reported by Pylint.

              
    new_graph = Graph()
    env: Dict[str, Any] = {}
    modules = dict(gm.named_modules())

    def load_arg(a):
        return map_arg(a, lambda node: env[node.name])

    for node in gm.graph.nodes:

            

Reported by Pylint.

                       node_input_type_c == NodeInputOrOutputType.INT8) or
        (node_input_type_a == NodeInputOrOutputType.FP32 and
         node_input_type_c == NodeInputOrOutputType.FP16) or
        # TODO(future PR): determine the actual dtype of node_c,
        # the current code only works because dequantize works with
        # multiple input dtypes.
        (node_input_type_a == NodeInputOrOutputType.FP32 and
         node_input_type_c == NodeInputOrOutputType.FP32_OR_INT8)
    ):

            

Reported by Pylint.

                          new_dtype_cast_name = \
                get_new_attr_name_with_prefix(node_name_prefix)(gm_b)
            if dtype_cast_op:
                # TODO(future PR): add handling for quantize_per_tensor
                new_dtype_cast_node = graph_c.create_node(
                    'call_function', dtype_cast_op, (prev_node_c_inner,), {},
                    new_dtype_cast_name)
                results.append(new_dtype_cast_node)
            else:

            

Reported by Pylint.

              import torch
from pathlib import Path
from scipy import sparse
import math


def to_coo_scipy(x):
    indices_1 = x._indices().numpy()
    values_1 = x._values().numpy()

            

Reported by Pylint.

              import torch
from pathlib import Path
from scipy import sparse
import math


def to_coo_scipy(x):
    indices_1 = x._indices().numpy()
    values_1 = x._values().numpy()

            

Reported by Pylint.

              

def to_coo_scipy(x):
    indices_1 = x._indices().numpy()
    values_1 = x._values().numpy()
    return sparse.coo_matrix((values_1, (indices_1[0], indices_1[1])),
                             shape=x.shape)



            

Reported by Pylint.

              
def to_coo_scipy(x):
    indices_1 = x._indices().numpy()
    values_1 = x._values().numpy()
    return sparse.coo_matrix((values_1, (indices_1[0], indices_1[1])),
                             shape=x.shape)


def sparse_grad_output(a, b):

            

Reported by Pylint.

              def read_matrix_params(path):
    with open(path, 'r') as file:
        line = file.readline()
        nrows, ncols, nnz = map(lambda el: int(el), line.split(', '))
        return (nrows, ncols), nnz


def csr_to_coo(indices, indptr, shape):
    n_rows, n_cols = shape

            

Reported by Pylint.

              

def csr_to_coo(indices, indptr, shape):
    n_rows, n_cols = shape
    cols = indices
    rows = [0] * len(cols)
    for i in range(n_rows):
        for j in range(indptr[i], indptr[i + 1]):
            rows[j] = i

            

Reported by Pylint.

              
def load_sparse_matrix(path, device):
    with open(path, 'r') as file:
        nrows, ncols, nnz = map(lambda el: int(el), file.readline().split(', '))
        index_pointers = map(lambda el: int(el), file.readline().split())
        indices = map(lambda el: int(el), file.readline().split())

    index_pointers = list(index_pointers)
    indices = list(indices)

            

Reported by Pylint.

              def load_sparse_matrix(path, device):
    with open(path, 'r') as file:
        nrows, ncols, nnz = map(lambda el: int(el), file.readline().split(', '))
        index_pointers = map(lambda el: int(el), file.readline().split())
        indices = map(lambda el: int(el), file.readline().split())

    index_pointers = list(index_pointers)
    indices = list(indices)
    data = torch.randn(nnz, dtype=torch.double)

            

Reported by Pylint.

                  with open(path, 'r') as file:
        nrows, ncols, nnz = map(lambda el: int(el), file.readline().split(', '))
        index_pointers = map(lambda el: int(el), file.readline().split())
        indices = map(lambda el: int(el), file.readline().split())

    index_pointers = list(index_pointers)
    indices = list(indices)
    data = torch.randn(nnz, dtype=torch.double)
    shape = (nrows, ncols)

            

Reported by Pylint.

              
def gen_vector(path, device):
    with open(path, 'r') as file:
        nrows, ncols, nnz = map(lambda el: int(el), file.readline().split(', '))
        index_pointers = map(lambda el: int(el), file.readline().split())
        indices = map(lambda el: int(el), file.readline().split())
        return torch.randn(nrows, dtype=torch.double, device=device)



            

Reported by Pylint.

                              serialized_device_map = {}
                for k in device_map:
                    serialized_device_map[k] = device_map[k].SerializeToString()
                self._NNModule = C.NNModuleFromProtobufDistributed(
                    serialized_proto, serialized_device_map
                )
            # Default
            elif serialized_proto:
                self._NNModule, self._OpList = C.NNModuleFromProtobuf(serialized_proto)

            

Reported by Pylint.

                              )
            # Default
            elif serialized_proto:
                self._NNModule, self._OpList = C.NNModuleFromProtobuf(serialized_proto)
            else:
                raise Exception(
                    "NNModule can be constructed with core.Net or caffe2_pb2.NetDef types"
                )
        else:

            

Reported by Pylint.

                                  "NNModule can be constructed with core.Net or caffe2_pb2.NetDef types"
                )
        else:
            self._NNModule = C.NNModule()

    @property
    def dataFlow(self):
        return self._NNModule.dataFlow()


            

Reported by Pylint.

                      return self._NNModule.dataFlow().replaceNode(old_node, new_node)

    def replaceProducer(self, tensor, new_producer):
        C.replaceProducer(tensor, new_producer)

    def replaceAllUsesWith(self, old_tensor, new_tensor):
        C.replaceAllUsesWith(old_tensor, new_tensor)

    def replaceAsConsumer(self, old_consumer, new_consumer):

            

Reported by Pylint.

                      C.replaceProducer(tensor, new_producer)

    def replaceAllUsesWith(self, old_tensor, new_tensor):
        C.replaceAllUsesWith(old_tensor, new_tensor)

    def replaceAsConsumer(self, old_consumer, new_consumer):
        C.replaceAsConsumer(old_consumer, new_consumer)

    def replaceSubgraph(self, subgraph, new_node, inputs, outputs):

            

Reported by Pylint.

                      C.replaceAllUsesWith(old_tensor, new_tensor)

    def replaceAsConsumer(self, old_consumer, new_consumer):
        C.replaceAsConsumer(old_consumer, new_consumer)

    def replaceSubgraph(self, subgraph, new_node, inputs, outputs):
        self._NNModule.replaceSubgraph(subgraph, new_node, inputs, outputs)

    def deleteSubgraph(self, subgraph):

            

Reported by Pylint.

              
    def match(self, pattern):
        for n in self.dataFlow.getMutableNodes():
            m = C.matchSubgraph(n, pattern)
            if m:
                yield m


def render(s):

            

Reported by Pylint.

                      print(s)


NeuralNetOperator = C.NeuralNetOperator
Operator = C.NeuralNetOperator
NeuralNetData = C.NeuralNetData
Data = C.NeuralNetData
NNSubgraph = C.NNSubgraph
NNMatchGraph = C.NNMatchGraph

            

Reported by Pylint.

              

NeuralNetOperator = C.NeuralNetOperator
Operator = C.NeuralNetOperator
NeuralNetData = C.NeuralNetData
Data = C.NeuralNetData
NNSubgraph = C.NNSubgraph
NNMatchGraph = C.NNMatchGraph
Graph = C.Graph

            

Reported by Pylint.

              
NeuralNetOperator = C.NeuralNetOperator
Operator = C.NeuralNetOperator
NeuralNetData = C.NeuralNetData
Data = C.NeuralNetData
NNSubgraph = C.NNSubgraph
NNMatchGraph = C.NNMatchGraph
Graph = C.Graph
Annotation = C.Annotation

            

Reported by Pylint.

                              self.max_width = max(self.max_width, len(value_str))

        else:
            nonzero_finite_vals = torch.masked_select(tensor_view, torch.isfinite(tensor_view) & tensor_view.ne(0))

            if nonzero_finite_vals.numel() == 0:
                # no valid number, do nothing
                return


            

Reported by Pylint.

                              self.max_width = max(self.max_width, len(value_str))

        else:
            nonzero_finite_vals = torch.masked_select(tensor_view, torch.isfinite(tensor_view) & tensor_view.ne(0))

            if nonzero_finite_vals.numel() == 0:
                # no valid number, do nothing
                return


            

Reported by Pylint.

                          nonzero_finite_max = nonzero_finite_abs.max().double()

            for value in nonzero_finite_vals:
                if value != torch.ceil(value):
                    self.int_mode = False
                    break

            if self.int_mode:
                # in int_mode for floats, all numbers are integers, and we append a decimal to nonfinites

            

Reported by Pylint.

                  if summarize and self.size(0) > 2 * PRINT_OPTS.edgeitems:
        data = ([_val_formatter(val) for val in self[:PRINT_OPTS.edgeitems].tolist()] +
                [' ...'] +
                [_val_formatter(val) for val in self[-PRINT_OPTS.edgeitems:].tolist()])
    else:
        data = [_val_formatter(val) for val in self.tolist()]

    data_lines = [data[i:i + elements_per_line] for i in range(0, len(data), elements_per_line)]
    lines = [', '.join(line) for line in data_lines]

            

Reported by Pylint.

                  if self.is_neg():
        self = self.resolve_neg()

    if self.dtype is torch.float16 or self.dtype is torch.bfloat16:
        self = self.float()

    if self.dtype.is_complex:
        # handle the conjugate bit
        self = self.resolve_conj()

            

Reported by Pylint.

                  if self.is_neg():
        self = self.resolve_neg()

    if self.dtype is torch.float16 or self.dtype is torch.bfloat16:
        self = self.float()

    if self.dtype.is_complex:
        # handle the conjugate bit
        self = self.resolve_conj()

            

Reported by Pylint.

                      return self
    if dim == 1:
        if self.size(0) > 2 * PRINT_OPTS.edgeitems:
            return torch.cat((self[:PRINT_OPTS.edgeitems], self[-PRINT_OPTS.edgeitems:]))
        else:
            return self
    if self.size(0) > 2 * PRINT_OPTS.edgeitems:
        start = [self[i] for i in range(0, PRINT_OPTS.edgeitems)]
        end = ([self[i]

            

Reported by Pylint.

                      return self
    if dim == 1:
        if self.size(0) > 2 * PRINT_OPTS.edgeitems:
            return torch.cat((self[:PRINT_OPTS.edgeitems], self[-PRINT_OPTS.edgeitems:]))
        else:
            return self
    if self.size(0) > 2 * PRINT_OPTS.edgeitems:
        start = [self[i] for i in range(0, PRINT_OPTS.edgeitems)]
        end = ([self[i]

            

Reported by Pylint.

                      start = [self[i] for i in range(0, PRINT_OPTS.edgeitems)]
        end = ([self[i]
               for i in range(len(self) - PRINT_OPTS.edgeitems, len(self))])
        return torch.stack([get_summarized_data(x) for x in (start + end)])
    else:
        return torch.stack([get_summarized_data(x) for x in self])

def _str_intern(inp):
    prefix = 'tensor('

            

Reported by Pylint.

                             for i in range(len(self) - PRINT_OPTS.edgeitems, len(self))])
        return torch.stack([get_summarized_data(x) for x in (start + end)])
    else:
        return torch.stack([get_summarized_data(x) for x in self])

def _str_intern(inp):
    prefix = 'tensor('
    indent = len(prefix)
    suffixes = []

            

Reported by Pylint.

              
import torch
from torch import Tensor
from .. import functional as F
from .module import Module
from .activation import MultiheadAttention
from .container import ModuleList
from ..init import xavier_uniform_
from .dropout import Dropout

            

Reported by Pylint.

              import torch
from torch import Tensor
from .. import functional as F
from .module import Module
from .activation import MultiheadAttention
from .container import ModuleList
from ..init import xavier_uniform_
from .dropout import Dropout
from .linear import Linear

            

Reported by Pylint.

              from torch import Tensor
from .. import functional as F
from .module import Module
from .activation import MultiheadAttention
from .container import ModuleList
from ..init import xavier_uniform_
from .dropout import Dropout
from .linear import Linear
from .normalization import LayerNorm

            

Reported by Pylint.

              from .. import functional as F
from .module import Module
from .activation import MultiheadAttention
from .container import ModuleList
from ..init import xavier_uniform_
from .dropout import Dropout
from .linear import Linear
from .normalization import LayerNorm


            

Reported by Pylint.

              from .module import Module
from .activation import MultiheadAttention
from .container import ModuleList
from ..init import xavier_uniform_
from .dropout import Dropout
from .linear import Linear
from .normalization import LayerNorm



            

Reported by Pylint.

              from .activation import MultiheadAttention
from .container import ModuleList
from ..init import xavier_uniform_
from .dropout import Dropout
from .linear import Linear
from .normalization import LayerNorm


class Transformer(Module):

            

Reported by Pylint.

              from .container import ModuleList
from ..init import xavier_uniform_
from .dropout import Dropout
from .linear import Linear
from .normalization import LayerNorm


class Transformer(Module):
    r"""A transformer model. User is able to modify the attributes as needed. The architecture

            

Reported by Pylint.

              from ..init import xavier_uniform_
from .dropout import Dropout
from .linear import Linear
from .normalization import LayerNorm


class Transformer(Module):
    r"""A transformer model. User is able to modify the attributes as needed. The architecture
    is based on the paper "Attention Is All You Need". Ashish Vaswani, Noam Shazeer,

            

Reported by Pylint.

                      r"""Generate a square mask for the sequence. The masked positions are filled with float('-inf').
            Unmasked positions are filled with float(0.0).
        """
        return torch.triu(torch.full((sz, sz), float('-inf')), diagonal=1)

    def _reset_parameters(self):
        r"""Initiate parameters in the transformer model."""

        for p in self.parameters():

            

Reported by Pylint.

                      r"""Generate a square mask for the sequence. The masked positions are filled with float('-inf').
            Unmasked positions are filled with float(0.0).
        """
        return torch.triu(torch.full((sz, sz), float('-inf')), diagonal=1)

    def _reset_parameters(self):
        r"""Initiate parameters in the transformer model."""

        for p in self.parameters():

            

Reported by Pylint.

              from textwrap import dedent
from unittest import skipIf

from torch.package import EmptyMatchError, Importer, PackageExporter, PackageImporter
from torch.package.package_exporter import PackagingError
from torch.testing._internal.common_utils import IS_WINDOWS, run_tests

try:
    from .common import PackageTestCase

            

Reported by Pylint.

              from unittest import skipIf

from torch.package import EmptyMatchError, Importer, PackageExporter, PackageImporter
from torch.package.package_exporter import PackagingError
from torch.testing._internal.common_utils import IS_WINDOWS, run_tests

try:
    from .common import PackageTestCase
except ImportError:

            

Reported by Pylint.

              
from torch.package import EmptyMatchError, Importer, PackageExporter, PackageImporter
from torch.package.package_exporter import PackagingError
from torch.testing._internal.common_utils import IS_WINDOWS, run_tests

try:
    from .common import PackageTestCase
except ImportError:
    # Support the case where we run this file directly.

            

Reported by Pylint.

                          he.save_source_string("foo", "import package_a.subpackage; import module_a")
        buffer.seek(0)
        hi = PackageImporter(buffer)
        import module_a
        import package_a.subpackage

        module_a_im = hi.import_module("module_a")
        hi.import_module("package_a.subpackage")
        package_a_im = hi.import_module("package_a")

            

Reported by Pylint.

                      buffer.seek(0)
        hi = PackageImporter(buffer)
        import module_a
        import package_a.subpackage

        module_a_im = hi.import_module("module_a")
        hi.import_module("package_a.subpackage")
        package_a_im = hi.import_module("package_a")


            

Reported by Pylint.

                          )
        buffer.seek(0)
        hi = PackageImporter(buffer)
        import module_a
        import package_a.subpackage

        module_a_im = hi.import_module("module_a")
        hi.import_module("package_a.subpackage")
        package_a_im = hi.import_module("package_a")

            

Reported by Pylint.

                      buffer.seek(0)
        hi = PackageImporter(buffer)
        import module_a
        import package_a.subpackage

        module_a_im = hi.import_module("module_a")
        hi.import_module("package_a.subpackage")
        package_a_im = hi.import_module("package_a")


            

Reported by Pylint.

                      Test that an error is thrown when a extern glob is specified with allow_empty=True
        and no matching module is required during packaging.
        """
        import package_a.subpackage  # noqa: F401

        buffer = BytesIO()
        with self.assertRaisesRegex(EmptyMatchError, r"did not match any modules"):
            with PackageExporter(buffer) as exporter:
                exporter.extern(include=["package_b.*"], allow_empty=False)

            

Reported by Pylint.

                          he.save_source_string("foo", "import package_a.subpackage")
        buffer.seek(0)
        hi = PackageImporter(buffer)
        import package_a.subpackage

        _ = package_a.subpackage
        import module_a

        _ = module_a

            

Reported by Pylint.

                      import package_a.subpackage

        _ = package_a.subpackage
        import module_a

        _ = module_a

        m = hi.import_module("package_a.subpackage")
        r = m.result

            

Reported by Pylint.

                  registry for this guarantees us the correct answer for all CSIDL_*
    names.
    """
    import winreg as _winreg

    shell_folder_name = {
        "CSIDL_APPDATA": "AppData",
        "CSIDL_COMMON_APPDATA": "Common AppData",
        "CSIDL_LOCAL_APPDATA": "Local AppData",

            

Reported by Pylint.

              

def _get_win_folder_with_pywin32(csidl_name):
    from win32com.shell import shellcon, shell
    dir = shell.SHGetFolderPath(0, getattr(shellcon, csidl_name), 0, 0)
    # Try to make this a unicode path because SHGetFolderPath does
    # not return unicode strings when there is unicode data in the
    # path.
    try:

            

Reported by Pylint.

              
def _get_win_folder_with_jna(csidl_name):
    import array
    from com.sun import jna
    from com.sun.jna.platform import win32

    buf_size = win32.WinDef.MAX_PATH * 2
    buf = array.zeros('c', buf_size)
    shell = win32.Shell32.INSTANCE

            

Reported by Pylint.

              def _get_win_folder_with_jna(csidl_name):
    import array
    from com.sun import jna
    from com.sun.jna.platform import win32

    buf_size = win32.WinDef.MAX_PATH * 2
    buf = array.zeros('c', buf_size)
    shell = win32.Shell32.INSTANCE
    shell.SHGetFolderPath(None, getattr(win32.ShlObj, csidl_name), None, win32.ShlObj.SHGFP_TYPE_CURRENT, buf)

            

Reported by Pylint.

                  from com.sun.jna.platform import win32

    buf_size = win32.WinDef.MAX_PATH * 2
    buf = array.zeros('c', buf_size)
    shell = win32.Shell32.INSTANCE
    shell.SHGetFolderPath(None, getattr(win32.ShlObj, csidl_name), None, win32.ShlObj.SHGFP_TYPE_CURRENT, buf)
    dir = jna.Native.toString(buf.tostring()).rstrip("\0")

    # Downgrade to short path name if have highbit chars. See

            

Reported by Pylint.

                          has_high_char = True
            break
    if has_high_char:
        buf = array.zeros('c', buf_size)
        kernel = win32.Kernel32.INSTANCE
        if kernel.GetShortPathName(dir, buf, buf_size):
            dir = jna.Native.toString(buf.tostring()).rstrip("\0")

    return dir

            

Reported by Pylint.

              SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
"""

"""Utilities for determining application-specific dirs.

See <https://github.com/ActiveState/appdirs> for details and usage.
"""
# Dev Notes:
# - MSDN on where to store app data files:

            

Reported by Pylint.

              


def user_data_dir(appname=None, appauthor=None, version=None, roaming=False):
    r"""Return full path to the user-specific data dir for this application.

        "appname" is the name of application.
            If None, just the system directory is returned.
        "appauthor" (only used on Windows) is the name of the

            

Reported by Pylint.

              


def user_data_dir(appname=None, appauthor=None, version=None, roaming=False):
    r"""Return full path to the user-specific data dir for this application.

        "appname" is the name of application.
            If None, just the system directory is returned.
        "appauthor" (only used on Windows) is the name of the

            

Reported by Pylint.

                  return path


def site_data_dir(appname=None, appauthor=None, version=None, multipath=False):
    r"""Return full path to the user-shared data dir for this application.

        "appname" is the name of application.
            If None, just the system directory is returned.
        "appauthor" (only used on Windows) is the name of the

            

Reported by Pylint.

              

import numpy as np
from hypothesis import given, assume, settings
import hypothesis.strategies as st

from caffe2.python import core, model_helper, brew, utils
import caffe2.python.hypothesis_test_util as hu
import caffe2.python.serialized_test.serialized_test_util as serial

            

Reported by Pylint.

              
import numpy as np
from hypothesis import given, assume, settings
import hypothesis.strategies as st

from caffe2.python import core, model_helper, brew, utils
import caffe2.python.hypothesis_test_util as hu
import caffe2.python.serialized_test.serialized_test_util as serial


            

Reported by Pylint.

                         epsilon=st.floats(1e-6, 1e-4),
           store_mean=st.booleans(),
           store_inv_stdev=st.booleans())
    def test_instance_norm_layout(self, gc, dc, N, C, H, W, store_mean,
                                  store_inv_stdev, epsilon, seed):
        # force store_inv_stdev if store_mean to match existing forward pass
        # implementation
        store_inv_stdev |= store_mean


            

Reported by Pylint.

                         store_inv_stdev=st.booleans(),
           inplace=st.booleans())
    def test_instance_norm_reference_check(
            self, gc, dc, N, C, H, W, order, store_mean, store_inv_stdev,
            epsilon, seed, inplace):
        np.random.seed(seed)

        # force store_inv_stdev if store_mean to match existing forward pass
        # implementation

            

Reported by Pylint.

              



import numpy as np
from hypothesis import given, assume, settings
import hypothesis.strategies as st

from caffe2.python import core, model_helper, brew, utils

            

Reported by Pylint.

              import caffe2.python.hypothesis_test_util as hu
import caffe2.python.serialized_test.serialized_test_util as serial

import unittest


class TestInstanceNorm(serial.SerializedTestCase):

    def _get_inputs(self, N, C, H, W, order):

            

Reported by Pylint.

              import unittest


class TestInstanceNorm(serial.SerializedTestCase):

    def _get_inputs(self, N, C, H, W, order):
        input_data = np.random.rand(N, C, H, W).astype(np.float32)
        if order == 'NHWC':
            # Allocate in the same order as NCHW and transpose to make sure

            

Reported by Pylint.

              
class TestInstanceNorm(serial.SerializedTestCase):

    def _get_inputs(self, N, C, H, W, order):
        input_data = np.random.rand(N, C, H, W).astype(np.float32)
        if order == 'NHWC':
            # Allocate in the same order as NCHW and transpose to make sure
            # the inputs are identical on freshly-seeded calls.
            input_data = utils.NCHW2NHWC(input_data)

            

Reported by Pylint.

              
class TestInstanceNorm(serial.SerializedTestCase):

    def _get_inputs(self, N, C, H, W, order):
        input_data = np.random.rand(N, C, H, W).astype(np.float32)
        if order == 'NHWC':
            # Allocate in the same order as NCHW and transpose to make sure
            # the inputs are identical on freshly-seeded calls.
            input_data = utils.NCHW2NHWC(input_data)

            

Reported by Pylint.

              
class TestInstanceNorm(serial.SerializedTestCase):

    def _get_inputs(self, N, C, H, W, order):
        input_data = np.random.rand(N, C, H, W).astype(np.float32)
        if order == 'NHWC':
            # Allocate in the same order as NCHW and transpose to make sure
            # the inputs are identical on freshly-seeded calls.
            input_data = utils.NCHW2NHWC(input_data)

            

Reported by Pylint.