Rddl wip functional

myGym/__init__.py

@@ -1,3 +1,3 @@
 # This file was generated by the setup.py and will be rewritten at the next build
-__version__ = "2.2.2.dev279+g20c5920"
+__version__ = "2.2.2.dev316+gb2cc0b1.d20241118"
 __name__ = "myGym"

myGym/configs/rllib_debug.json

@@ -19,7 +19,7 @@
     "task_type"             :"reach",
     "natural_language"      :0,
     "task_objects"          :[{"init":{"obj_name":"null"},
-                               "goal":{"obj_name":"cube_holes","fixed":1,"rand_rot":0, "sampling_area":[-0.5, 0.2, 0.3, 0.6, 0.1, 0.4]}}],
+                               "goal":{"obj_name":"apple_target","fixed":1,"rand_rot":0, "sampling_area":[-0.5, 0.2, 0.3, 0.6, 0.1, 0.4]}}],
     "used_objects"          :{"num_range":[0,0], "obj_list":[]},
     // Observation
     // actual_state options (pick one): "endeff_xyz", "endeff_6D" (robot end effector), "obj_xyz", "obj_6D", "vae", "yolact", "voxel" or "dope"

myGym/configs/train_rddl.json

@@ -0,0 +1,44 @@
+{
+    #Environment
+    "env_name"  :"Gym-v0",
+    "workspace" :"table",
+    "engine"    :"pybullet",
+    "render"    :"opengl",
+    "camera"    :1,
+    "gui"       :1,
+    "visualize" :1,
+    "visgym"    :0,
+    #Robot
+    "robot"         :"kuka",
+    "robot_action"  :"step",
+    "robot_init"    :[0.0, 0.4, 0.4],
+    "max_velocity"  :1,
+    "max_force"     :70,
+    "action_repeat" :1,
+    #Task
+    "natural_language"      :0,
+    "rddl": {"protoactions":["Approach", "Grasp", "Move", "Drop"], "num_task_range":[1,3], "allowed_objects":["Apple", "Tuna", "Banana"], "allowed_predicates":["Near"]},
+    "observation"           : {"actual_state":"obj_xyz", "goal_state":"obj_xyz"}, # this is ignored
+    #Distractor
+    "distractors"           : {"list":null, "moveable":1, "constant_speed":0, "movement_dims":3,
+                                           "movement_endpoints":[-0.3, 0.3, 0.4, 0.7, 0.1, 0.3]},
+    #Reward
+    "vae_path"      :null,
+    "yolact_path"   :null,
+    "yolact_config" :null,
+    #Train
+    "train_framework"   :"pytorch",
+    "algo"              :"ppo",
+    "num_networks" : 1,
+    "max_episode_steps" :1024,
+    "algo_steps"        :1024,
+    "steps"             :200000,
+    "pretrained_model"  :null,
+    "multiprocessing"   :false,
+    #Evaluation
+    "eval_freq"     :10000,
+    "eval_episodes" :10,
+    #Saving and logging
+    "logdir"    :"trained_models/test",
+    "record"    :0
+    }

myGym/env.txt

@@ -57,7 +57,6 @@ matplotlib                    3.1.3
 mdurl                         0.1.2
 mpi4py                        3.1.4
 msgpack                       1.0.5
-mujoco-py                     2.1.2.14
 myGym                         2.2.2.dev32+gd0bbd2d /home/michal/code/myGym
 networkx                      2.6.3
 numpy                         1.18.5
@@ -221,7 +220,6 @@ mistune                       0.8.4
 mpi4py                        3.1.4
 mpmath                        1.3.0
 msgpack                       1.0.5
-mujoco-py                     2.1.2.14
 myGym                         2.2.2.dev32+gd0bbd2d /home/michal/code/myGym
 networkx                      2.6.3
 nicomotion                    1.0

myGym/envs/__init__.py

@@ -1,4 +1,4 @@
-from gym.envs.registration import register
+from gymnasium.envs.registration import register
 
 register(
     id="CrowWorkspaceEnv-v0",

myGym/envs/base_env.py

@@ -4,14 +4,13 @@
 import pybullet_utils.bullet_client as bc
 import time
 import numpy as np
-from gym.utils import seeding
-import gym
+from gymnasium.utils import seeding
+import gymnasium as gym
 import inspect
 from myGym.envs.camera import Camera
-import pkg_resources
-currentdir = pkg_resources.resource_filename("myGym", "envs")
-repodir = pkg_resources.resource_filename("myGym", "./")
-
+import importlib.resources as resources
+currentdir = resources.files("myGym").joinpath("envs")
+repodir = resources.files("myGym").parent
 
 class BaseEnv(gym.Env):
     """
@@ -30,7 +29,7 @@ class BaseEnv(gym.Env):
 
     def __init__(self,
                  gui_on=True,
-                 objects_dir_path=pkg_resources.resource_filename("myGym", "envs/"),
+                 objects_dir_path=resources.files("myGym").joinpath("envs"),
                  max_steps=1024,
                  show_bounding_boxes_gui=False,
                  changing_light_gui=False,
@@ -73,6 +72,7 @@ def __init__(self,
         self._set_physics()
         self._setup_scene()
         self._set_action_space()
+        self.obsdim = self.check_obs_template()
         self._set_observation_space()
 
 
@@ -197,7 +197,7 @@ def _restart_episode(self):
         self.episode_reward = 0.0
         self.episode_steps = 0
 
-    def reset(self, hard=False):
+    def reset(self, hard=False,options=None):
         """
         Reset the state of the environment
         """