FM / functions

/usr/local/lib64/python3.6/site-packages/torch/include/torch/csrc/autograd/functions/

/usr/local/lib64/python3.6/site-packages/torch/include/torch/csrc/autograd/functions

Name	Size	Mode	Actions
accumulate_grad.h	11466	0644	edit dl rm
basic_ops.h	2668	0644	edit dl rm
comm.h	1271	0644	edit dl rm
pybind.h	314	0644	edit dl rm
tensor.h	3926	0644	edit dl rm
utils.h	2883	0644	edit dl rm

Edit: /usr/local/lib64/python3.6/site-packages/torch/include/torch/csrc/autograd/functions/utils.h (2883B)

#pragma once

#include <torch/csrc/WindowsTorchApiMacro.h>
#include <torch/csrc/autograd/autograd.h>
#include <torch/csrc/autograd/function.h>
#include <torch/csrc/autograd/variable.h>
#include <torch/csrc/autograd/InferenceMode.h>
#include <torch/csrc/utils/variadic.h>

#include <ATen/ATen.h>

#include <functional>
#include <memory>
#include <vector>

namespace torch { namespace autograd {

using function_constructor = std::function<std::shared_ptr<Node>(edge_list&&)>;

/**
 * Wraps the tensor outputs in variables and creates the grad_fn and sets the
 * grad_fn if necessary.
 */
TORCH_API variable_list wrap_outputs(const variable_list& inputs, tensor_list&& outputs,
                                     const function_constructor& ctr);

///  Checks that inputs contains exactly `args` items and that the first `required_args`
/// items are not nullptr. If not specified, `required_args` defaults to `args`.
TORCH_API void check_input_variables(const char* name, const variable_list& inputs, int args, int required_args=-1, bool allow_undefined=false);

struct ComputeRequiresGrad : IterArgs<ComputeRequiresGrad> {
  bool out = false;
  using IterArgs<ComputeRequiresGrad>::operator();
  void operator()(const at::Tensor& tensor) {
    const auto& var = static_cast<const Variable&>(tensor);
    if (var.defined() && var.requires_grad()) {
      out = true;
    }
  }
  void operator()(const c10::optional<at::Tensor>& tensor) {
    if (tensor.has_value()) {
      (*this)(*tensor);
    }
  }
  bool short_circuit() {
    return out;
  }
};

template <typename... Args>
inline bool compute_requires_grad(Args&&... args) {
  if (!GradMode::is_enabled()) {
    return false;
  }
  return ComputeRequiresGrad().apply(std::forward<Args>(args)...).out;
}

inline void set_history(
    at::Tensor& variable,
    const std::shared_ptr<Node>& grad_fn) {
  AT_ASSERT(grad_fn);
  if (variable.defined()) {
    // If the codegen triggers this, you most likely want to add your newly added function
    // to the DONT_REQUIRE_DERIVATIVE list in tools/autograd/gen_variable_type.py
    TORCH_INTERNAL_ASSERT(isDifferentiableType(variable.scalar_type()));
    auto output_nr =
        grad_fn->add_input_metadata(variable);
    impl::set_gradient_edge(variable, {grad_fn, output_nr});
  } else {
    grad_fn->add_input_metadata(Node::undefined_input());
  }
}

inline void set_history(
    std::vector<Variable>&& variables,
    const std::shared_ptr<Node>& grad_fn) {
  for (auto& variable : variables) {
    set_history(variable, grad_fn);
  }
}

inline void set_history(
    std::vector<Variable>& variables,
    const std::shared_ptr<Node>& grad_fn) {
  for (auto& variable : variables) {
    set_history(variable, grad_fn);
  }
}

inline bool isFwGradDefined(const c10::optional<at::Tensor>& t) {
  return t.has_value() && t->defined() && t->_fw_grad(/*level */ 0).defined();
}

}}

#pragma once #include #include #include #include #include #include #include #include #include #include namespace torch { namespace autograd { using function_constructor = std::function(edge_list&&)>; /** * Wraps the tensor outputs in variables and creates the grad_fn and sets the * grad_fn if necessary. */ TORCH_API variable_list wrap_outputs(const variable_list& inputs, tensor_list&& outputs, const function_constructor& ctr); /// Checks that inputs contains exactly `args` items and that the first `required_args` /// items are not nullptr. If not specified, `required_args` defaults to `args`. TORCH_API void check_input_variables(const char* name, const variable_list& inputs, int args, int required_args=-1, bool allow_undefined=false); struct ComputeRequiresGrad : IterArgs { bool out = false; using IterArgs::operator(); void operator()(const at::Tensor& tensor) { const auto& var = static_cast(tensor); if (var.defined() && var.requires_grad()) { out = true; } } void operator()(const c10::optional& tensor) { if (tensor.has_value()) { (*this)(*tensor); } } bool short_circuit() { return out; } }; template inline bool compute_requires_grad(Args&&... args) { if (!GradMode::is_enabled()) { return false; } return ComputeRequiresGrad().apply(std::forward(args)...).out; } inline void set_history( at::Tensor& variable, const std::shared_ptr& grad_fn) { AT_ASSERT(grad_fn); if (variable.defined()) { // If the codegen triggers this, you most likely want to add your newly added function // to the DONT_REQUIRE_DERIVATIVE list in tools/autograd/gen_variable_type.py TORCH_INTERNAL_ASSERT(isDifferentiableType(variable.scalar_type())); auto output_nr = grad_fn->add_input_metadata(variable); impl::set_gradient_edge(variable, {grad_fn, output_nr}); } else { grad_fn->add_input_metadata(Node::undefined_input()); } } inline void set_history( std::vector&& variables, const std::shared_ptr& grad_fn) { for (auto& variable : variables) { set_history(variable, grad_fn); } } inline void set_history( std::vector& variables, const std::shared_ptr& grad_fn) { for (auto& variable : variables) { set_history(variable, grad_fn); } } inline bool isFwGradDefined(const c10::optional& t) { return t.has_value() && t->defined() && t->_fw_grad(/*level */ 0).defined(); } }}