tf_agents.agents.data_converter.DataContext

A class that stores useful data for performing data conversions.

tf_agents.agents.data_converter.DataContext(
    time_step_spec: tf_agents.trajectories.TimeStep,
    action_spec: tf_agents.typing.types.NestedTensorSpec,
    info_spec: tf_agents.typing.types.NestedTensorSpec,
    policy_state_spec: tf_agents.typing.types.NestedTensorSpec = (),
    use_half_transition: bool = False
)

Args
`time_step_spec`	A nest of `tf.TimeStep` representing the time_steps.
`action_spec`	A nest of `tf.TypeSpec` representing the actions.
`info_spec`	A nest of `tf.TypeSpec` representing the policy's info. (Typically this is the info emitted by the collect policy).
`policy_state_spec`	A nest of `tf.TypeSpec` representing the policy's state.
`use_half_transition`	A special transition that the next_time_step contains a dummy observation.

Raises
`TypeError`	If any of the specs are not nests containing tf.TypeSpec objects.

Attributes
`action_spec`
`info_spec`
`policy_state_spec`
`time_step_spec`
`trajectory_spec`
`transition_spec`

tf_agents.agents.data_converter.DataContext

Args

Raises

Attributes