Vector data for one on-policy trajectory. More...
#include <LibtorchRLTrajectoryBuffer.h>
Public Attributes | |
| std::vector< std::vector< Real > > | observations |
| Observations for each transition. | |
| std::vector< std::vector< Real > > | next_observations |
| Next observations for each transition. | |
| std::vector< std::vector< Real > > | actions |
| Actions applied at each transition. | |
| std::vector< std::vector< Real > > | log_probabilities |
| Action log-probabilities recorded during rollout. | |
| std::vector< Real > | rewards |
| Scalar rewards for each transition. | |
| std::vector< bool > | terminals |
| Terminal flags for each transition. A true value cuts value bootstrapping at that step. | |
| std::vector< Real > | value_targets |
| Critic targets aligned with each transition. | |
| std::vector< Real > | advantages |
| Advantage estimates aligned with each transition. | |
Vector data for one on-policy trajectory.
Definition at line 30 of file LibtorchRLTrajectoryBuffer.h.
| std::vector<std::vector<Real> > LibtorchRLTrajectoryBuffer::Trajectory::actions |
Actions applied at each transition.
Definition at line 37 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryAssembler::assemble(), and LibtorchRLTrajectoryBuffer::validateTrajectory().
| std::vector<Real> LibtorchRLTrajectoryBuffer::Trajectory::advantages |
Advantage estimates aligned with each transition.
Definition at line 47 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryBuffer::validateTrajectory().
| std::vector<std::vector<Real> > LibtorchRLTrajectoryBuffer::Trajectory::log_probabilities |
Action log-probabilities recorded during rollout.
Definition at line 39 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryAssembler::assemble(), and LibtorchRLTrajectoryBuffer::validateTrajectory().
| std::vector<std::vector<Real> > LibtorchRLTrajectoryBuffer::Trajectory::next_observations |
Next observations for each transition.
Definition at line 35 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryAssembler::assemble(), LibtorchRLValueEstimator::estimate(), and LibtorchRLTrajectoryBuffer::validateTrajectory().
| std::vector<std::vector<Real> > LibtorchRLTrajectoryBuffer::Trajectory::observations |
Observations for each transition.
Definition at line 33 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryAssembler::assemble(), LibtorchRLValueEstimator::estimate(), and LibtorchRLTrajectoryBuffer::validateTrajectory().
| std::vector<Real> LibtorchRLTrajectoryBuffer::Trajectory::rewards |
Scalar rewards for each transition.
Definition at line 41 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryAssembler::assemble(), LibtorchRLValueEstimator::estimate(), and LibtorchRLTrajectoryBuffer::validateTrajectory().
| std::vector<bool> LibtorchRLTrajectoryBuffer::Trajectory::terminals |
Terminal flags for each transition. A true value cuts value bootstrapping at that step.
Definition at line 43 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryAssembler::assemble(), LibtorchRLValueEstimator::estimate(), and LibtorchRLTrajectoryBuffer::validateTrajectory().
| std::vector<Real> LibtorchRLTrajectoryBuffer::Trajectory::value_targets |
Critic targets aligned with each transition.
Definition at line 45 of file LibtorchRLTrajectoryBuffer.h.
Referenced by LibtorchRLTrajectoryBuffer::validateTrajectory().