mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 00:21:07 +01:00
5c46600f84
# Motivation This PR intends to extend `cuda_lazy_init` to `device_lazy_init` which is a device-agnostic API that can support any backend. And change `maybe_initialize_cuda` to `maybe_initialize_device` to support lazy initialization for CUDA while maintaining scalability. # Design We maintain a flag for each backend to manage the lazy initialization state separately. # Additional Context No need more UTs. This is a reland PR, the original PR is [refactor lazy init to device-agnostic](https://github.com/pytorch/pytorch/pull/118846). This is a common PR, and does not trigger xpu ciflow. Differential Revision: [D53478332](https://our.internmc.facebook.com/intern/diff/D53478332) Pull Request resolved: https://github.com/pytorch/pytorch/pull/119248 Approved by: https://github.com/EikanWang, https://github.com/gujinghui, https://github.com/jgong5, https://github.com/atalman
94 lines
2.5 KiB
C++
94 lines
2.5 KiB
C++
#define TORCH_ASSERT_ONLY_METHOD_OPERATORS
|
|
// ${generated_comment}
|
|
|
|
// Python bindings for torch.* functions implemented through ATen.
|
|
//
|
|
// The functions are bound as static methods on a class
|
|
// torch._C._VariableFunctions which is also aliased as Variable._torch
|
|
// and also copied into 'torch' module.
|
|
|
|
#include <Python.h>
|
|
|
|
// Undefine the copysign macro so that at::copysign works as intended with MSVC
|
|
// https://github.com/python/cpython/blob/c60394c7fc9cc09b16e9675a3eeb5844b6d8523f/PC/pyconfig.h#L196
|
|
#ifdef _MSC_VER
|
|
#undef copysign
|
|
#endif // _MSC_VER
|
|
|
|
#include "torch/csrc/autograd/python_torch_functions.h"
|
|
#include "torch/csrc/autograd/python_variable.h"
|
|
#include "torch/csrc/autograd/utils/wrap_outputs.h"
|
|
#include "torch/csrc/Dtype.h"
|
|
#include "torch/csrc/DynamicTypes.h"
|
|
#include "torch/csrc/Exceptions.h"
|
|
#include "torch/csrc/utils/out_types.h"
|
|
#include "torch/csrc/utils/pybind.h"
|
|
#include "torch/csrc/utils/pycfunction_helpers.h"
|
|
#include "torch/csrc/utils/python_arg_parser.h"
|
|
#include "torch/csrc/utils/tensor_layouts.h"
|
|
#include "torch/csrc/utils/tensor_new.h"
|
|
#include "torch/csrc/utils/tensor_numpy.h"
|
|
#include "torch/csrc/jit/frontend/tracer.h"
|
|
#include "torch/csrc/autograd/generated/variable_factories.h"
|
|
#include "torch/csrc/utils/structseq.h"
|
|
#include "torch/csrc/utils/device_lazy_init.h"
|
|
#include "torch/csrc/autograd/generated/python_return_types.h"
|
|
|
|
#include <ATen/core/Tensor.h>
|
|
|
|
#ifndef AT_PER_OPERATOR_HEADERS
|
|
#include <ATen/Functions.h>
|
|
#else
|
|
$ops_headers
|
|
#endif
|
|
|
|
#include <functional>
|
|
#include <initializer_list>
|
|
#include <stdexcept>
|
|
#include <utility>
|
|
|
|
using at::Tensor;
|
|
using at::Device;
|
|
using at::Layout;
|
|
using at::Scalar;
|
|
using at::ScalarType;
|
|
using at::Backend;
|
|
using at::OptionalDeviceGuard;
|
|
using at::DeviceGuard;
|
|
using at::TensorOptions;
|
|
using at::IntArrayRef;
|
|
using at::Generator;
|
|
using at::TensorList;
|
|
using at::Dimname;
|
|
using at::DimnameList;
|
|
using at::ArrayRef;
|
|
|
|
using torch::utils::check_out_type_matches;
|
|
using namespace torch::autograd::utils;
|
|
|
|
// NOTE: See [Sharded File] comment in VariableType
|
|
|
|
namespace torch::autograd {
|
|
|
|
// generated forward declarations start here
|
|
|
|
${py_forwards}
|
|
|
|
static PyMethodDef torch_functions_shard[] = {
|
|
${py_method_defs}
|
|
};
|
|
|
|
void gatherTorchFunctions${shard_id}(std::vector<PyMethodDef> &torch_functions) {
|
|
constexpr size_t num_functions = sizeof(torch_functions_shard) / sizeof(torch_functions_shard[0]);
|
|
torch_functions.insert(
|
|
torch_functions.end(),
|
|
torch_functions_shard,
|
|
torch_functions_shard + num_functions);
|
|
}
|
|
|
|
// generated methods start here
|
|
|
|
${py_methods}
|
|
|
|
} // namespace torch::autograd
|