10 #ifdef LIBTORCH_ENABLED 33 virtual void execute()
override;
51 const torch::Tensor & output_tensor);
virtual void execute() override
We compute the actions in this function together with the corresponding logarithmic probabilities...
const std::vector< Real > _action_std
Standard deviation for the actions, supplied by the user.
Real getSignalLogProbability(const unsigned int signal_index) const
Get the logarithmic probability of (signal_index)-th signal of the control neural net...
static InputParameters validParams()
LibtorchDRLControl(const InputParameters ¶meters)
Construct using input parameters.
DIE A HORRIBLE DEATH HERE typedef LIBMESH_DEFAULT_SCALAR_TYPE Real
const InputParameters & parameters() const
torch::Tensor _std
Standard deviations converted to a 2D diagonal tensor that can be used by Libtorch routines...
A time-dependent, neural-network-based controller which is associated with a Proximal Policy Optimiza...
std::vector< Real > _current_control_signal_log_probabilities
The log probability of control signals from the last evaluation of the controller.
torch::Tensor computeLogProbability(const torch::Tensor &action, const torch::Tensor &output_tensor)
Function which computes the logarithmic probability of given actions.