viberl.typing

Custom typing classes for reinforcement learning using Pydantic.

Classes:

Name	Description
Action	An action taken by an agent, optionally with log probabilities.
Transition	A single transition in an episode.
Trajectory	A complete trajectory (episode) consisting of multiple transitions.

Action

Action(**data: Any)

Bases: BaseModel

An action taken by an agent, optionally with log probabilities.

Attributes:

Name	Type	Description
model_config
action	int
logprobs	Tensor | None

Source code in .venv/lib/python3.12/site-packages/pydantic/main.py

def __init__(self, /, **data: Any) -> None:
    """Create a new model by parsing and validating input data from keyword arguments.

    Raises [`ValidationError`][pydantic_core.ValidationError] if the input data cannot be
    validated to form a valid model.

    `self` is explicitly positional-only to allow `self` as a field name.
    """
    # `__tracebackhide__` tells pytest and some other tools to omit this function from tracebacks
    __tracebackhide__ = True
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
    if self is not validated_self:
        warnings.warn(
            'A custom validator is returning a value other than `self`.\n'
            "Returning anything other than `self` from a top level model validator isn't supported when validating via `__init__`.\n"
            'See the `model_validator` docs (https://docs.pydantic.dev/latest/concepts/validators/#model-validators) for more details.',
            stacklevel=2,
        )

model_config class-attribute instance-attribute

model_config = ConfigDict(arbitrary_types_allowed=True)

action instance-attribute

action: int

logprobs class-attribute instance-attribute

logprobs: Tensor | None = None

Transition

Transition(**data: Any)

Bases: BaseModel

A single transition in an episode.

Attributes:

Name	Type	Description
model_config
state	ndarray
action	Action
reward	float
next_state	ndarray
done	bool
info	dict[str, Any]

Source code in .venv/lib/python3.12/site-packages/pydantic/main.py

def __init__(self, /, **data: Any) -> None:
    """Create a new model by parsing and validating input data from keyword arguments.

    Raises [`ValidationError`][pydantic_core.ValidationError] if the input data cannot be
    validated to form a valid model.

    `self` is explicitly positional-only to allow `self` as a field name.
    """
    # `__tracebackhide__` tells pytest and some other tools to omit this function from tracebacks
    __tracebackhide__ = True
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
    if self is not validated_self:
        warnings.warn(
            'A custom validator is returning a value other than `self`.\n'
            "Returning anything other than `self` from a top level model validator isn't supported when validating via `__init__`.\n"
            'See the `model_validator` docs (https://docs.pydantic.dev/latest/concepts/validators/#model-validators) for more details.',
            stacklevel=2,
        )

model_config class-attribute instance-attribute

model_config = ConfigDict(arbitrary_types_allowed=True)

state instance-attribute

state: ndarray

action instance-attribute

action: Action

reward instance-attribute

reward: float

next_state instance-attribute

next_state: ndarray

done instance-attribute

done: bool

info class-attribute instance-attribute

info: dict[str, Any] = {}

Trajectory

Trajectory(**data: Any)

Bases: BaseModel

A complete trajectory (episode) consisting of multiple transitions.

Methods:

Name	Description
from_transitions	Create a trajectory from a list of transitions.
to_dict	Convert trajectory to dictionary format for agent learning.

Attributes:

Name	Type	Description
model_config
transitions	list[Transition]
total_reward	float
length	int

Source code in .venv/lib/python3.12/site-packages/pydantic/main.py

def __init__(self, /, **data: Any) -> None:
    """Create a new model by parsing and validating input data from keyword arguments.

    Raises [`ValidationError`][pydantic_core.ValidationError] if the input data cannot be
    validated to form a valid model.

    `self` is explicitly positional-only to allow `self` as a field name.
    """
    # `__tracebackhide__` tells pytest and some other tools to omit this function from tracebacks
    __tracebackhide__ = True
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
    if self is not validated_self:
        warnings.warn(
            'A custom validator is returning a value other than `self`.\n'
            "Returning anything other than `self` from a top level model validator isn't supported when validating via `__init__`.\n"
            'See the `model_validator` docs (https://docs.pydantic.dev/latest/concepts/validators/#model-validators) for more details.',
            stacklevel=2,
        )

model_config class-attribute instance-attribute

model_config = ConfigDict(arbitrary_types_allowed=True)

transitions instance-attribute

transitions: list[Transition]

total_reward instance-attribute

total_reward: float

length instance-attribute

length: int

from_transitions classmethod

from_transitions(transitions: list[Transition]) -> Trajectory

Create a trajectory from a list of transitions.

Source code in viberl/typing.py

@classmethod
def from_transitions(cls, transitions: list[Transition]) -> Trajectory:
    """Create a trajectory from a list of transitions."""
    total_reward = sum(t.reward for t in transitions)
    return cls(transitions=transitions, total_reward=total_reward, length=len(transitions))

to_dict

to_dict() -> dict

Convert trajectory to dictionary format for agent learning.

Source code in viberl/typing.py

def to_dict(self) -> dict:
    """Convert trajectory to dictionary format for agent learning."""
    return {
        'states': [t.state for t in self.transitions],
        'actions': [t.action.action for t in self.transitions],
        'rewards': [t.reward for t in self.transitions],
        'next_states': [t.next_state for t in self.transitions],
        'dones': [t.done for t in self.transitions],
        'logprobs': [
            t.action.logprobs for t in self.transitions if t.action.logprobs is not None
        ],
        'infos': [t.info for t in self.transitions],
    }