ParameterStore.h

// Copyright 2024, UChicago Argonne, LLC
// All Rights Reserved
// Software Name: NEML2 -- the New Engineering material Model Library, version 2
// By: Argonne National Laboratory
// OPEN SOURCE LICENSE (MIT)
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
 
#pragma once
 
#include "neml2/base/NEML2Object.h"
#include "neml2/base/OptionSet.h"
#include "neml2/base/Storage.h"
#include "neml2/tensors/TensorValue.h"
 
// The following are not directly used by ParameterStore itself.
// We put them here so that derived classes can add expected options of these types.
#include "neml2/base/CrossRef.h"
#include "neml2/base/EnumSelection.h"
 
namespace neml2
{
// Forward decl
class VariableBase;
class Model;
 
class ParameterStore
{
public:
  ParameterStore(const OptionSet & options, NEML2Object * object);
 
  const Storage<std::string, TensorValueBase> & named_parameters() const
  {
    return const_cast<ParameterStore *>(this)->named_parameters();
  }
  Storage<std::string, TensorValueBase> & named_parameters();
 
  void set_parameter(const std::string &, const Tensor &);
 
  void set_parameters(const std::map<std::string, Tensor> &);
 
  TensorValueBase & get_parameter(const std::string & name);
 
  const TensorValueBase & get_parameter(const std::string & name) const;
 
  bool has_nl_param() const { return !_nl_params.empty(); }
 
  const VariableBase * nl_param(const std::string &) const;
 
  virtual std::map<std::string, const VariableBase *>
  named_nonlinear_parameters(bool recursive = false) const;
 
  virtual std::map<std::string, Model *>
  named_nonlinear_parameter_models(bool recursive = false) const;
 
protected:
  virtual void send_parameters_to(const torch::TensorOptions & options);
 
  template <typename T, typename = typename std::enable_if_t<std::is_base_of_v<TensorBase<T>, T>>>
  const T & declare_parameter(const std::string & name, const T & rawval);
 
  template <typename T, typename = typename std::enable_if_t<std::is_base_of_v<TensorBase<T>, T>>>
  const T & declare_parameter(const std::string & name,
                              const std::string & input_option_name,
                              bool allow_nonlinear = false);
 
  std::map<std::string, const VariableBase *> _nl_params;
 
  std::map<std::string, Model *> _nl_param_models;
 
private:
  NEML2Object * _object;
 
  const OptionSet _options;
 
  Storage<std::string, TensorValueBase> _param_values;
};
 
template <typename T, typename>
const T &
ParameterStore::declare_parameter(const std::string & name, const T & rawval)
{
  if (_object->host() != _object)
    return _object->host<ParameterStore>()->declare_parameter(
        _object->name() + parameter_name_separator() + name, rawval);
 
  TensorValueBase * base_ptr;
 
  // If the parameter already exists, get it
  if (_param_values.has_key(name))
    base_ptr = &get_parameter(name);
  // If the parameter doesn't exist, create it
  else
  {
    auto val = std::make_unique<TensorValue<T>>(rawval);
    base_ptr = _param_values.set_pointer(name, std::move(val));
  }
 
  auto ptr = dynamic_cast<TensorValue<T> *>(base_ptr);
  neml_assert(ptr, "Internal error: Failed to cast parameter to a concrete type.");
  return ptr->value();
}
 
} // namespace neml2