tf_agents.agents.data_converter.AsTrajectory

Class that validates and converts other data types to Trajectory.

Note that validation and conversion allows values to contain dictionaries with extra keys as compared to the the specs in the data context. These additional entries / observations are ignored and dropped during conversion.

This non-strict checking allows users to provide additional info and observation keys at input without having to manually prune them before converting.

data_context An instance of DataContext, typically accessed from the TFAgent.data_context property.
sequence_length The required time dimension value (if any), typically determined by the subclass of TFAgent.

Methods

__call__

View source

Convers value to a Trajectory. Performs data validation and pruning.

  • If value is already a Trajectory, only validation is performed.
  • If value is a Transition with tensors containing two ([B, T]) outer dims, then it is simply repackaged to a Trajectory and then validated.
  • If value is a Transition with tensors containing one ([B]) outer dim, a ValueError is raised.

Args
value A Trajectory or Transition object to convert.

Returns
A validated and pruned Trajectory.

Raises
TypeError If value is not one of Trajectory or Transition.
ValueError If value has structure that doesn't match the converter's spec.
TypeError If value has a structure that doesn't match the converter's spec.
ValueError If value is a Transition without a time dimension, as training Trajectories typically have batch and time dimensions.