il_representations.scripts package¶
Module contents¶
The scripts package contains the scripts that we use to perform representation learning, imitation learning, evaluation, dataset generation, and so on.
Submodules¶
il_representations.scripts.collate_configs module¶
il_representations.scripts.grab_some_stuff_and_pickle_it module¶
il_representations.scripts.il_test module¶
il_representations.scripts.il_train module¶
il_representations.scripts.interpret module¶
il_representations.scripts.joint_training module¶
il_representations.scripts.joint_training_cluster module¶
il_representations.scripts.mkdataset_demos module¶
il_representations.scripts.mkdataset_random module¶
il_representations.scripts.pretrain_n_adapt module¶
il_representations.scripts.render_dataset module¶
il_representations.scripts.run_rep_learner module¶
il_representations.scripts.save_traced_net module¶
Trace & save a network for a MAGICAL environment. Capable of automatically figuring out where the encoder is.
- il_representations.scripts.save_traced_net.auto_save_name(module_path: str) str ¶
Use config.json files to automatically come up with a name for the given encoder.
- il_representations.scripts.save_traced_net.fetch_encoder(net: torch.nn.Module) torch.nn.Module ¶
- il_representations.scripts.save_traced_net.get_config_path(module_path: str) Optional[str] ¶
Keep walking up the tree until we find a config.json.
- il_representations.scripts.save_traced_net.load_eval_net(net_path: str) torch.nn.Module ¶
Load a network on disk, making sure it ends up on the CPU & in eval mode.
- il_representations.scripts.save_traced_net.main(args: argparse.Namespace) None ¶
- il_representations.scripts.save_traced_net.pre_order_children(net: torch.nn.Module) Iterator[Tuple[str, torch.nn.Module]] ¶
Traverse module tree in pre-order.
- il_representations.scripts.save_traced_net.trace_encoder(encoder: torch.nn.Module) torch.nn.Module ¶
Generate a random example of the appropriate size & use it to trace the given network.