mirror of
https://github.com/pytorch/pytorch.git
synced 2025-10-21 21:49:24 +08:00
73604eed0c
Follows #133067 Pull Request resolved: https://github.com/pytorch/pytorch/pull/133399 Approved by: https://github.com/Skylion007
25 lines
745 B
C++
25 lines
745 B
C++
#pragma once
|
|
|
|
#include <torch/csrc/Export.h>
|
|
#include <torch/csrc/jit/codegen/fuser/arg_spec.h>
|
|
#include <torch/csrc/jit/codegen/fuser/partition_desc.h>
|
|
#include <torch/csrc/jit/codegen/fuser/tensor_desc.h>
|
|
#include <torch/csrc/jit/ir/ir.h>
|
|
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
namespace torch::jit::fuser {
|
|
|
|
// Creates a CPU or CUDA kernel for the given graph.
|
|
// Returns the C++ or CUDA string implementing the kernel.
|
|
TORCH_API std::string generateKernel(
|
|
const std::string& name,
|
|
const Graph& graph,
|
|
const std::vector<std::pair<const Value*, const std::optional<TensorDesc>>>&
|
|
inputs,
|
|
const std::vector<std::pair<const Value*, const TensorDesc>>& outputs,
|
|
const bool use_cuda);
|
|
|
|
} // namespace torch::jit::fuser
|