*/
uint64_t CPUGeneratorImpl::seed() {
  auto random = c10::detail::getNonDeterministicRandom();
  this->set_current_seed(random);
  return random;
}

/**
 * Sets the internal state of CPUGeneratorImpl. The new internal state

            

Reported by FlawFinder.

              uint64_t CPUGeneratorImpl::seed() {
  auto random = c10::detail::getNonDeterministicRandom();
  this->set_current_seed(random);
  return random;
}

/**
 * Sets the internal state of CPUGeneratorImpl. The new internal state
 * must be a strided CPU byte tensor and of the same size as either

            

Reported by FlawFinder.

               *
 * See Note [Acquire lock when using random generators]
 */
uint32_t CPUGeneratorImpl::random() {
  return engine_();
}

/**
 * Gets a random 64 bit unsigned integer from the engine

            

Reported by FlawFinder.

                  accum_state->next_float_normal_sample = *(this->next_float_normal_sample_);
  }

  memcpy(rng_state, accum_state.get(), size);
  return state_tensor.getIntrusivePtr();
}

/**
 * Gets the DeviceType of CPUGeneratorImpl.

            

Reported by FlawFinder.

                  }
    CAFFE_ENFORCE(src);
    CAFFE_ENFORCE(dst);
    memcpy(dst, src, nbytes);
  }

  void CopyBytesFromCPU(size_t nbytes, const void* src, void* dst) override {
    CopyBytesSameDevice(nbytes, src, dst);
  }

            

Reported by FlawFinder.

                }
  CAFFE_ENFORCE(src);
  CAFFE_ENFORCE(dst);
  memcpy(dst, src, nbytes);
}

template <>
inline void IDEEPContext::CopyBytes<CPUContext, IDEEPContext>(
    size_t nbytes,

            

Reported by FlawFinder.

                }
  CAFFE_ENFORCE(src);
  CAFFE_ENFORCE(dst);
  memcpy(dst, src, nbytes);
}

template <>
inline void IDEEPContext::CopyBytes<IDEEPContext, CPUContext>(
    size_t nbytes,

            

Reported by FlawFinder.

                }
  CAFFE_ENFORCE(src);
  CAFFE_ENFORCE(dst);
  memcpy(dst, src, nbytes);
}

} // namespace caffe2

            

Reported by FlawFinder.

                  return model.net.Transpose(blob_in, blob_out, **kwargs)


def sum(model, blob_in, blob_out, **kwargs):
    """Sum"""
    return model.net.Sum(blob_in, blob_out, **kwargs)


def reduce_sum(model, blob_in, blob_out, **kwargs):

            

Reported by Pylint.

              ## @package algebra
# Module caffe2.python.helpers.algebra






def transpose(model, blob_in, blob_out, use_cudnn=False, **kwargs):

            

Reported by Pylint.

                  """ArgMin"""
    return model.net.ArgMin(blob_in, blob_out, **kwargs)

def batch_mat_mul(model, blob_in, blob_out,
                  enable_tensor_core=False, **kwargs):
    if enable_tensor_core:
        kwargs['engine'] = 'TENSORCORE'

    return model.net.BatchMatMul(blob_in, blob_out, **kwargs)

            

Reported by Pylint.

              
    return model.net.BatchMatMul(blob_in, blob_out, **kwargs)

def sparse_lengths_sum_4bit_rowwise_sparse(model, blob_in, blob_out, **kwargs):
    return model.net.SparseLengthsSum4BitRowwiseSparse(blob_in, blob_out, **kwargs)

            

Reported by Pylint.

                  auto* other_per_channel_affine =
        static_cast<PerChannelAffineQuantizer*>(other.get());
    return scalar_type() == other_per_channel_affine->scalar_type() &&
        scales().equal(other_per_channel_affine->scales()) &&
        zero_points().equal(other_per_channel_affine->zero_points()) &&
        axis() == other_per_channel_affine->axis();
  }

 protected:

            

Reported by FlawFinder.

                      static_cast<PerChannelAffineQuantizer*>(other.get());
    return scalar_type() == other_per_channel_affine->scalar_type() &&
        scales().equal(other_per_channel_affine->scales()) &&
        zero_points().equal(other_per_channel_affine->zero_points()) &&
        axis() == other_per_channel_affine->axis();
  }

 protected:
  Tensor scales_;

            

Reported by FlawFinder.

                  auto* other_per_channel_float_qparams =
        static_cast<PerChannelAffineFloatQParamsQuantizer*>(other.get());
    return scalar_type() == other_per_channel_float_qparams->scalar_type() &&
        scales().equal(other_per_channel_float_qparams->scales()) &&
        zero_points().equal(other_per_channel_float_qparams->zero_points()) &&
        axis() == other_per_channel_float_qparams->axis();
  }
};


            

Reported by FlawFinder.

                      static_cast<PerChannelAffineFloatQParamsQuantizer*>(other.get());
    return scalar_type() == other_per_channel_float_qparams->scalar_type() &&
        scales().equal(other_per_channel_float_qparams->scales()) &&
        zero_points().equal(other_per_channel_float_qparams->zero_points()) &&
        axis() == other_per_channel_float_qparams->axis();
  }
};

// This is an internal utility function for getting at the QTensorImpl,

            

Reported by FlawFinder.

                    struct Bundle final {
        struct Buffer final {
          VkPipelineStageFlags stage;
          VkAccessFlags access;

          operator bool() const;
        } staging, buffer;

        struct Image final {

            

Reported by FlawFinder.

              
        struct Image final {
          VkPipelineStageFlags stage;
          VkAccessFlags access;
          VkImageLayout layout;

          operator bool() const;
        } image;
      };

            

Reported by FlawFinder.

              
inline vTensor::View::State::Bundle::Buffer::operator bool() const {
  return (0u != stage) &&
         (0u != access);
}

inline vTensor::View::State::Bundle::Image::operator bool() const {
  return (0u != stage) &&
         (0u != access) &&

            

Reported by FlawFinder.

              
inline vTensor::View::State::Bundle::Image::operator bool() const {
  return (0u != stage) &&
         (0u != access) &&
         (VK_IMAGE_LAYOUT_UNDEFINED != layout);
}

inline bool vTensor::View::State::is_available(
    const Component::Flags components) const {

            

Reported by FlawFinder.

              from importlib import import_module
import pkgutil
import sys
from . import layers


def import_recursive(package):
    """
    Takes a package and imports all modules underneath it

            

Reported by Pylint.

              def find_subclasses_recursively(base_cls, sub_cls):
    cur_sub_cls = base_cls.__subclasses__()
    sub_cls.update(cur_sub_cls)
    for cls in cur_sub_cls:
        find_subclasses_recursively(cls, sub_cls)


import_recursive(sys.modules[__name__])


            

Reported by Pylint.

              




from importlib import import_module
import pkgutil
import sys
from . import layers

            

Reported by Pylint.

                          import_recursive(module)


def find_subclasses_recursively(base_cls, sub_cls):
    cur_sub_cls = base_cls.__subclasses__()
    sub_cls.update(cur_sub_cls)
    for cls in cur_sub_cls:
        find_subclasses_recursively(cls, sub_cls)


            

Reported by Pylint.

              ## @package last_n_window_collector
# Module caffe2.python.layers.last_n_window_collector

from caffe2.python import core, schema
from caffe2.python.layers.layers import ModelLayer


class LastNWindowCollector(ModelLayer):
    """

            

Reported by Pylint.

              
    def __init__(self, model, input_record, num_to_collect,
                 name='last_n_window_collector', **kwargs):
        super(LastNWindowCollector, self).__init__(
            model, name, input_record, **kwargs)
        assert num_to_collect > 0
        self.num_to_collect = num_to_collect
        assert isinstance(input_record, schema.Scalar), \
            "Got {!r}".format(input_record)

            

Reported by Pylint.

                               name='last_n_window_collector', **kwargs):
        super(LastNWindowCollector, self).__init__(
            model, name, input_record, **kwargs)
        assert num_to_collect > 0
        self.num_to_collect = num_to_collect
        assert isinstance(input_record, schema.Scalar), \
            "Got {!r}".format(input_record)

        self.last_n = self.create_param(param_name='last_n',

            

Reported by Bandit.

                          model, name, input_record, **kwargs)
        assert num_to_collect > 0
        self.num_to_collect = num_to_collect
        assert isinstance(input_record, schema.Scalar), \
            "Got {!r}".format(input_record)

        self.last_n = self.create_param(param_name='last_n',
                                        shape=[0],
                                        initializer=('ConstantFill', {}),

            

Reported by Bandit.

                      const auto iy = y * stride_h + kh;
        const auto ix = kw;
        if (stride_w == 1) {
          memcpy(
              dst + (y * output_w),
              src + (iy * width + ix),
              sizeof(T) * output_w);
        } else {
          for (auto x = 0; x < output_w; x++) {

            

Reported by FlawFinder.

                            sizeof(T) * output_w);
        } else {
          for (auto x = 0; x < output_w; x++) {
            memcpy(
                dst + (y * output_w + x),
                src + (iy * width + ix + x * stride_w),
                sizeof(T));
          }
        }

            

Reported by FlawFinder.

                      for (int iw = w_pad; iw < w_pad + dkernel_w; iw += dilation_w, ++s) {
          if (ih >= 0 && ih < height && iw >= 0 && iw < width) {
            for (int g = 0; g < groups; ++g) {
              memcpy(
                  data_col_temp +
                      ((g * kernel_h + r) * kernel_w + s) * (channels / groups),
                  data_im + (ih * width + iw) * channels +
                      g * (channels / groups),
                  sizeof(T) * (channels / groups));

            

Reported by FlawFinder.

                            if (it >= 0 && it < num_frames && ih >= 0 && ih < height &&
                  iw >= 0 && iw < width) {
                for (int g = 0; g < groups; ++g) {
                  memcpy(
                      data_col_temp +
                          (((g * kernel_t + q) * kernel_h + r) * kernel_w + s) *
                              (channels / groups),
                      data_im + ((it * height + ih) * width + iw) * channels +
                          g * (channels / groups),

            

Reported by FlawFinder.

              using namespace ::c10d;

int main(int argc, char** argv) {
  int rank = atoi(getenv("RANK"));
  int size = atoi(getenv("SIZE"));
  auto store = c10::make_intrusive<FileStore>("/tmp/c10d_example", size);
  ProcessGroupGloo pg(store, rank, size);

  // Create some tensors

            

Reported by FlawFinder.

              
int main(int argc, char** argv) {
  int rank = atoi(getenv("RANK"));
  int size = atoi(getenv("SIZE"));
  auto store = c10::make_intrusive<FileStore>("/tmp/c10d_example", size);
  ProcessGroupGloo pg(store, rank, size);

  // Create some tensors
  const auto ntensors = 10;

            

Reported by FlawFinder.

              using namespace ::c10d;

int main(int argc, char** argv) {
  int rank = atoi(getenv("RANK"));
  int size = atoi(getenv("SIZE"));
  auto store = c10::make_intrusive<FileStore>("/tmp/c10d_example", size);
  ProcessGroupGloo pg(store, rank, size);

  // Create some tensors

            

Reported by FlawFinder.

              
int main(int argc, char** argv) {
  int rank = atoi(getenv("RANK"));
  int size = atoi(getenv("SIZE"));
  auto store = c10::make_intrusive<FileStore>("/tmp/c10d_example", size);
  ProcessGroupGloo pg(store, rank, size);

  // Create some tensors
  const auto ntensors = 10;

            

Reported by FlawFinder.

                        memcmp(deser.first.data(), payload.data(), payload.size()) == 0);
    }
    for (size_t i = 0; i < tensors.size(); ++i) {
      EXPECT_TRUE(torch::equal(tensors[i], deser.second[i]));
    }
  };
  run("", {});
  run("hi", {});
  run("", {torch::randn({5, 5})});

            

Reported by FlawFinder.

                EXPECT_EQ(tiny.storage().nbytes() / tiny.dtype().itemsize(), k1K * k1K);
  auto ser = torch::distributed::rpc::wireSerialize({}, {tiny});
  auto deser = torch::distributed::rpc::wireDeserialize(ser.data(), ser.size());
  EXPECT_TRUE(torch::equal(tiny, deser.second[0]));
  EXPECT_LT(ser.size(), (tiny.element_size() * k1K) + k1K);
}

TEST(WireSerialize, CloneSparseTensors) {
  constexpr size_t k1K = 1024;

            

Reported by FlawFinder.

                at::Tensor tiny = big.select(0, 2); // Select a row in the middle
  auto v2 = torch::distributed::rpc::cloneSparseTensors({tiny});
  EXPECT_NE(&v2.get(0).storage(), &tiny.storage()); // Cloned.
  EXPECT_TRUE(torch::equal(v2.get(0), tiny));

  at::Tensor sparse = at::empty({2, 3}, at::dtype<float>().layout(at::kSparse));
  auto v3 = torch::distributed::rpc::cloneSparseTensors({sparse});
  // There is no storage() to compare, but at least confirm equality.
  EXPECT_TRUE(v3.get(0).is_same(sparse));

            

Reported by FlawFinder.

                at::Tensor main = at::empty({2, 3}, at::dtype<float>().layout(at::kSparse));
  auto ser = torch::distributed::rpc::wireSerialize({}, {main.to(at::kSparse)});
  auto deser = torch::distributed::rpc::wireDeserialize(ser.data(), ser.size());
  EXPECT_TRUE(torch::equal(main, deser.second[0]));
}

            

Reported by FlawFinder.