Files
PrimAITE/docs/source/how_to_guides/custom_rewards.rst
2025-03-11 12:27:45 +00:00

49 lines
1.9 KiB
ReStructuredText

.. only:: comment
© Crown-owned copyright 2025, Defence Science and Technology Laboratory UK
.. _custom_rewards:
Creating Custom Rewards in PrimAITE
***********************************
Rewards within PrimAITE are contained within ``rewards.py``, which details the rewards available for all agents within training sessions, how they are calculated and any other specific information where necessary.
Rewards within PrimAITE have been updated to facilitate extensibility and the creation of plugins with the release of PrimAITE version 4.0. Additional information about this is covered within :ref:`extensible_rewards`.
Custom Rewards within PrimAITE should inherit from the ``AbstractReward`` class, found in ``rewards.py``. It's important to include an identifier for any class created within PrimAITE.
.. code:: Python
class ExampleAward(AbstractReward, identifier="ExampleAward"):
"""Example Reward Class """
def calculate(self, state: Dict, last_action_response: "AgentHistoryItem") -> float:
"""Calculate the reward for the current state."""
return 1.0
@classmethod
def from_config(cls, config: dict) -> "AbstractReward":
"""Create a reward function component from a config dictionary."""
return cls()
Custom rewards that have been created should be added to the ``rew_class_identifiers`` dictionary within the ``RewardFunction`` class in ``rewards.py``.
Including Custom Rewards within PrimAITE configuration
======================================================
Custom rewards can then be included within an agents configuration by it's inclusion within the training session configuration YAML.
.. code:: yaml
agents:
- ref: agent_name
reward_function:
reward_components:
- type: DUMMY
weight: 1.0
More detailed information about rewards within PrimAITE can be found within :ref:`Rewards`