|
pulsatrix 1.1.0
|
Optimizer parameter groups: a learning rate and weight decay per set of parameters, chosen by name (roadmap TRN-1). More...
#include <cmath>#include <functional>#include <stdexcept>#include <string>#include <unordered_map>#include <utility>#include <vector>#include "pulsatrix/module.hpp"

Go to the source code of this file.
Classes | |
| struct | pulsatrix::ParamGroup |
Optimizer settings for the parameters selects picks. More... | |
| class | pulsatrix::ParamGroupSet |
| The group machinery SGDOptimizer and AdamOptimizer share. More... | |
| struct | pulsatrix::ParamGroupSet::Assignment |
| One parameter with the settings that apply to it this step. More... | |
Namespaces | |
| namespace | pulsatrix |
| namespace | pulsatrix::param_select |
| Common selectors. | |
Typedefs | |
| using | pulsatrix::ParamSelector = std::function< bool(const std::string &name, const Tensor &value)> |
| Picks parameters by their named_parameters() name and value. | |
Functions | |
| ParamSelector | pulsatrix::param_select::name_prefix (std::string prefix) |
The parameter named exactly prefix, and every parameter under it (dot boundaries). | |
| ParamSelector | pulsatrix::param_select::one_dimensional () |
| Parameters of rank 0 or 1: biases and normalization scales. The usual recipe gives these no weight decay. | |
| void | pulsatrix::check_optimizer_setting (float value, const std::string &what) |
Throws std::invalid_argument unless value is finite and non-negative. | |
Optimizer parameter groups: a learning rate and weight decay per set of parameters, chosen by name (roadmap TRN-1).