imitation.scripts.train_imitation#
Trains DAgger on synthetic demonstrations generated from an expert policy.
Functions
- imitation.scripts.train_imitation.bc(bc, _run, _rnd)[source]#
Runs BC training.
- Parameters
bc (
Dict[str,Any]) – Configuration for BC training._run – Sacred run object.
_rnd (
Generator) – Random number generator provided by Sacred.
- Return type
Mapping[str,Mapping[str,float]]- Returns
Statistics for rollouts from the trained policy and demonstration data.
- imitation.scripts.train_imitation.dagger(bc, dagger, _run, _rnd)[source]#
Runs DAgger training.
- Parameters
bc (
Dict[str,Any]) – Configuration for BC training.dagger (
Mapping[str,Any]) – Arguments for DAgger training._run – Sacred run object.
_rnd (
Generator) – Random number generator provided by Sacred.
- Return type
Mapping[str,Mapping[str,float]]- Returns
Statistics for rollouts from the trained policy and demonstration data.