diff --git a/predicators/settings.py b/predicators/settings.py index 4852530f2d..f329c57ad5 100644 --- a/predicators/settings.py +++ b/predicators/settings.py @@ -653,7 +653,7 @@ def get_arg_specific_settings(cls, args: Dict[str, Any]) -> Dict[str, Any]: # the horizon to be shorter. "touch_point": 15, # Ditto for the simple grid row environment. - "grid_row": cls.grid_row_num_cells + 5, + "grid_row": cls.grid_row_num_cells + 2, })[args.get("env", "")], # Maximum number of steps to roll out an option policy. diff --git a/scripts/configs/active_sampler_learning.yaml b/scripts/configs/active_sampler_learning.yaml index 1fea71fd54..b4afc6ee3d 100644 --- a/scripts/configs/active_sampler_learning.yaml +++ b/scripts/configs/active_sampler_learning.yaml @@ -73,7 +73,7 @@ FLAGS: active_sampler_learning_model: "myopic_classifier_mlp" active_sampler_learning_use_teacher: False online_nsrt_learning_requests_per_cycle: 1 - max_num_steps_interaction_request: 500 + max_num_steps_interaction_request: 150 num_online_learning_cycles: 10 active_sampler_learning_explore_length_base: 100000 # effectively disable sesame_task_planner: "fdopt-costs"