imitation.scripts#
Command-line scripts.
Modules
Commands to analyze experimental results. |
|
Configuration settings for scripts. |
|
Converts old-style pickle or npz trajectories to new-style HuggingFace datasets. |
|
Evaluate policies: render policy interactively, save videos, log episode return. |
|
Ingredients for Sacred experiments. |
|
Runs a Sacred experiment in parallel. |
|
Train GAIL or AIRL. |
|
Trains DAgger on synthetic demonstrations generated from an expert policy. |
|
Train a reward model using preference comparisons. |
|
Uses RL to train a policy from scratch, saving rollouts and policy. |
|
Tunes the hyperparameters of the algorithms. |