Merge pull request #55 from rte-france/bd-dev

Version 0.7.0

CHANGELOG.rst

@@ -15,6 +15,11 @@ Change Log
 - train sowhere a working baseline (that does better than do nothing)
 - show an example of a baseline that uses a GNN
 
+[0.7.0] - 2023-07-13
+------------------------
+- [ADDED] the "topo oracle agent" (contrib)
+- [ADDED] the "curriculumagent" (contrib)
+
 [0.6.0.post1] - 2022-07-05
 ---------------------------
 - [FIXED] issue with the `PPO_SB3` agent when using a runner, particularly when no "heuristic" are

README.md

@@ -25,16 +25,16 @@ We welcome contributions: see the [contribute guide](/CONTRIBUTE.md) for details
 
 # Get started with a baseline
 
-Say you want to know how you compared with the "DoubleDuelingDQN" baseline implementation in this repository (for the
+Say you want to know how you compared with the "PPO_SB3" baseline implementation in this repository (for the
 sake of this example).
 
 ## Train it (optional)
 As no weights are provided for this baselines by default (yet), you will first need to train such a baseline:
 
 ```python
 import grid2op
-from l2rpn_baselines.DoubleDuelingDQN import train
-env = grid2op.make()
+from l2rpn_baselines.PPO_SB3 import train
+env = grid2op.make("l2rpn_case14_sandbox")
 res = train(env, save_path="THE/PATH/TO/SAVE/IT", iterations=100)
 ```
 
@@ -46,9 +46,9 @@ Once trained, you can reload it and evaluate its performance with the provided "
 
 ```python
 import grid2op
-from l2rpn_baselines.DoubleDuelingDQN import evaluate
-env = grid2op.make()
-res = evaluate(env, load_path="THE/PATH/TO/LOAD/IT.h5", nb_episode=10)
+from l2rpn_baselines.PPO_SB3 import evaluate
+env = grid2op.make("l2rpn_case14_sandbox")
+res = evaluate(env, load_path="THE/PATH/TO/LOAD/IT", nb_episode=10)
 ```
 
 You can have more information about extra argument of the "evaluate" function in the 

docs/conf.py

@@ -22,8 +22,8 @@
 author = 'Benjamin DONNOT'
 
 # The full version, including alpha/beta/rc tags
-release = '0.6.0.post1'
-version = '0.6'
+release = '0.7.0'
+version = '0.7'
 
 # -- General configuration ---------------------------------------------------
 

docs/ppo_stable_baselines.rst

@@ -169,6 +169,19 @@ For example, to create an agent **from scratch**, with some parameters:
     You should probably train it before hand (see the `train` function)
 
 
+Caveats
++++++++++
+
+Be carefull, at time of writing, there is a migration from all major RL packages 
+from the legacy "open ai gym" api to the more recent "gymanisium" api.
+
+This transition will happen soon in grid2op too. But for now, and not to
+break legacy code that might have worked on previous L2RPN competitions, the
+switch from gym to gymanisium is not yet made.
+
+This means that installing stable-baselines3 might cause some issues with grid2op
+and l2rpn baselines due to the gym / gymanisium migration. 
+
 Detailed documentation
 ++++++++++++++++++++++++
 

l2rpn_baselines/__init__.py

@@ -6,7 +6,7 @@
 # SPDX-License-Identifier: MPL-2.0
 # This file is part of L2RPN Baselines, L2RPN Baselines a repository to host baselines for l2rpn competitions.
 
-__version__ = "0.6.0.post1"
+__version__ = "0.7.0"
 
 all_baselines_li = [
     "Template",

l2rpn_baselines/test/test_train_eval.py

@@ -91,6 +91,14 @@
 except ImportError as exc_:
     has_ppo_sb3 = exc_
 
+try:
+    from l2rpn_baselines.CurriculumAgent import train as train_curiculum_agent
+    from l2rpn_baselines.CurriculumAgent import evaluate as eval_curiculum_agent
+    has_curiculum = None
+except ImportError as exc_:
+    has_curiculum = exc_
+
+
 
 class TestDeepQSimple(unittest.TestCase):
     def setUp(self) -> None:
@@ -724,5 +732,35 @@ def test_train_eval(self):
             assert eval_res is not None
 
 
+class TestCuriculumAgent(unittest.TestCase):
+    def test_train_eval(self):
+        tmp_dir = tempfile.mkdtemp()
+        if has_curiculum is not None:
+            raise ImportError(f"CuriculumAgent is not available with error:\n{has_curiculum}")
+
+        with warnings.catch_warnings():
+            warnings.filterwarnings("ignore")
+            env = grid2op.make("l2rpn_case14_sandbox", test=True)
+            nm_ = "TestCuriculumAgent"
+
+            train_curiculum_agent(env,
+                       name=nm_,
+                       iterations=1,
+                       save_path=tmp_dir,
+                       load_path=None,
+                       learning_rate=1e-4,
+                       verbose=False)
+
+            agent, eval_res = eval_curiculum_agent(env,
+                                 load_path=tmp_dir,
+                                 name=nm_,
+                                 logs_path=tmp_dir,
+                                 nb_episode=1,
+                                 nb_process=1,
+                                 max_steps=10,
+                                 verbose=False,
+                                 save_gif=False)
+            assert eval_res is not None
+
 if __name__ == "__main__":
     unittest.main()

setup.py

@@ -8,7 +8,7 @@
 
 import setuptools
 from setuptools import setup
-__version__ = "0.6.0.post1"
+__version__ = "0.7.0"
 
 
 pkgs = {