Release 0.7.0

Author: yannbouteiller

.gitignore

@@ -19,7 +19,7 @@ __pycache__/
 *$py.class
 
 # C extensions
-*.so
+# *.so
 
 # Distribution / packaging
 .Python

README.md

@@ -242,7 +242,7 @@ This works on any track, using any (sensible) camera configuration.
 }
 ```
 Note that human players can see or hear the features provided by this environment: we provide no "cheat" that would render the approach non-transferable to the real world.
-In case you do wish to cheat, though, you can easily take inspiration from our [rtgym interfaces](https://github.com/trackmania-rl/tmrl/blob/master/tmrl/custom/custom_gym_interfaces.py) to build your own custom environment for TrackMania.
+In case you do wish to cheat, though, you can easily take inspiration from our [rtgym interfaces](https://github.com/trackmania-rl/tmrl/blob/master/tmrl/custom/tm/tm_gym_interfaces.py) to build your own custom environment for TrackMania.
 
 The `Full` environment is used in the official [TMRL competition](https://github.com/trackmania-rl/tmrl/blob/master/readme/competition.md), and custom environments are featured in the "off" competition :wink:
 
@@ -471,7 +471,7 @@ This project uses [Real-Time Gym](https://github.com/yannbouteiller/rtgym) (```r
 
 Time-steps are being elastically constrained to their nominal duration. When this elastic constraint cannot be satisfied, the previous time-step times out and the new time-step starts from the current timestamp.
 
-Custom `rtgym` interfaces for Trackmania used by `tmrl` are implemented in [custom_gym_interfaces.py](https://github.com/yannbouteiller/tmrl/blob/master/tmrl/custom/custom_gym_interfaces.py).
+Custom `rtgym` interfaces for Trackmania used by `tmrl` are implemented in [tm_gym_interfaces.py](https://github.com/yannbouteiller/tmrl/blob/master/tmrl/custom/tm/tm_gym_interfaces.py).
 
 ## Remote training architecture:
 

pyproject.toml

@@ -0,0 +1,5 @@
+[build-system]
+requires = [
+    "setuptools",
+]
+build-backend = "setuptools.build_meta"

readme/tuto_library.md

@@ -169,7 +169,7 @@ We use [Real-Time Gym](https://github.com/yannbouteiller/rtgym) (`rtgym`) to wra
 But you can use an environment of your choice, as long as it is registered as a Gymnasium environment.
 
 To build your own real-time environment (e.g., an environment for your own robot or video game), you can follow the [rtgym tutorial](https://github.com/yannbouteiller/rtgym#tutorial).
-If you need inspiration, you can find our `rtgym` interfaces for TrackMania in [custom_gym_interfaces.py](https://github.com/trackmania-rl/tmrl/blob/master/tmrl/custom/custom_gym_interfaces.py).
+If you need inspiration, you can find our `rtgym` interfaces for TrackMania in [tm_gym_interfaces.py](https://github.com/trackmania-rl/tmrl/blob/master/tmrl/custom/tm/tm_gym_interfaces.py).
 
 For the sake of the `tmrl` tutorial, we will be using the dummy RC drone environment from the `rtgym` tutorial:
 
@@ -558,7 +558,7 @@ A few more parameters are configurable, although they will not be useful in this
 In particular:
 
 `obs_preprocessor` can be used to modify observations returned by the environment (this enables, e.g., converting RGB images into grayscale without modifying the environment).
-Some examples of such preprocessors are available [here](https://github.com/trackmania-rl/tmrl/blob/master/tmrl/custom/custom_preprocessors.py).
+Some examples of such preprocessors are available [here](https://github.com/trackmania-rl/tmrl/blob/master/tmrl/custom/tm/tm_preprocessors.py).
 
 `standalone` can be set to `True` for deployment, in which case the `RolloutWorker` will not attempt to connect to the `Server`.
 

setup.py

@@ -9,6 +9,10 @@
 import urllib.error
 import socket
 
+
+TMRL_VERSION = '0.7.0'
+
+
 if sys.version_info < (3, 7):
     sys.exit('Sorry, Python < 3.7 is not supported.')
 
@@ -137,13 +141,13 @@ def url_retrieve(url: str, outfile: Path, overwrite: bool = False):
 
 setup(
     name='tmrl',
-    version='0.6.6',
+    version=TMRL_VERSION,
     description='Network-based framework for real-time robot learning',
     long_description=README,
     long_description_content_type='text/markdown',
     keywords='reinforcement learning, robot learning, trackmania, self driving, roborace',
     url='https://github.com/trackmania-rl/tmrl',
-    download_url='https://github.com/trackmania-rl/tmrl/archive/refs/tags/v0.6.6.tar.gz',
+    download_url=f'https://github.com/trackmania-rl/tmrl/archive/refs/tags/v{TMRL_VERSION}.tar.gz',
     author='Yann Bouteiller, Edouard Geze',
     author_email='[email protected], [email protected]',
     license='MIT',

tmrl/actor.py

@@ -134,7 +134,7 @@ def save(self, path):
 
     def load(self, path, device):
         self.device = device
-        self.load_state_dict(torch.load(path, map_location=self.device))
+        self.load_state_dict(torch.load(path, map_location=self.device, weights_only=True))
         return self
 
     def act_(self, obs, test=False):

tmrl/config/config_objects.py

@@ -2,15 +2,14 @@
 # from tmrl.custom.custom_checkpoints import load_run_instance_images_dataset, dump_run_instance_images_dataset
 # third-party imports
 
-import numpy as np
 import rtgym
 
 # local imports
 import tmrl.config.config_constants as cfg
 from tmrl.training_offline import TorchTrainingOffline
-from tmrl.custom.custom_gym_interfaces import TM2020Interface, TM2020InterfaceLidar, TM2020InterfaceLidarProgress
+from tmrl.custom.tm.tm_gym_interfaces import TM2020Interface, TM2020InterfaceLidar, TM2020InterfaceLidarProgress
 from tmrl.custom.custom_memories import MemoryTMFull, MemoryTMLidar, MemoryTMLidarProgress, get_local_buffer_sample_lidar, get_local_buffer_sample_lidar_progress, get_local_buffer_sample_tm20_imgs
-from tmrl.custom.custom_preprocessors import obs_preprocessor_tm_act_in_obs, obs_preprocessor_tm_lidar_act_in_obs,obs_preprocessor_tm_lidar_progress_act_in_obs
+from tmrl.custom.tm.tm_preprocessors import obs_preprocessor_tm_act_in_obs, obs_preprocessor_tm_lidar_act_in_obs, obs_preprocessor_tm_lidar_progress_act_in_obs
 from tmrl.envs import GenericGymEnv
 from tmrl.custom.custom_models import SquashedGaussianMLPActor, MLPActorCritic, REDQMLPActorCritic, RNNActorCritic, SquashedGaussianRNNActor, SquashedGaussianVanillaCNNActor, VanillaCNNActorCritic, SquashedGaussianVanillaColorCNNActor, VanillaColorCNNActorCritic
 from tmrl.custom.custom_algorithms import SpinupSacAgent as SAC_Agent

tmrl/custom/custom_algorithms.py

@@ -11,7 +11,7 @@
 # local imports
 import tmrl.custom.custom_models as core
 from tmrl.custom.utils.nn import copy_shared, no_grad
-from tmrl.util import cached_property, partial
+from tmrl.util import cached_property
 from tmrl.training import TrainingAgent
 import tmrl.config.config_constants as cfg
 

tmrl/custom/tm/__init__.py

@@ -15,12 +15,12 @@
 
 # local imports
 import tmrl.config.config_constants as cfg
-from tmrl.custom.utils.compute_reward import RewardFunction
-from tmrl.custom.utils.control_gamepad import control_gamepad, gamepad_reset, gamepad_close_finish_pop_up_tm20
-from tmrl.custom.utils.control_mouse import mouse_close_finish_pop_up_tm20
-from tmrl.custom.utils.control_keyboard import apply_control, keyres
-from tmrl.custom.utils.window import WindowInterface
-from tmrl.custom.utils.tools import Lidar, TM2020OpenPlanetClient, save_ghost
+from tmrl.custom.tm.utils.compute_reward import RewardFunction
+from tmrl.custom.tm.utils.control_gamepad import control_gamepad, gamepad_reset, gamepad_close_finish_pop_up_tm20
+from tmrl.custom.tm.utils.control_mouse import mouse_close_finish_pop_up_tm20
+from tmrl.custom.tm.utils.control_keyboard import apply_control, keyres
+from tmrl.custom.tm.utils.window import WindowInterface
+from tmrl.custom.tm.utils.tools import Lidar, TM2020OpenPlanetClient, save_ghost
 
 # Globals ==============================================================================================================