Merge pull request #27 from airboxlab/deps-upgrade

bump deps

.github/workflows/tests.yml

@@ -18,15 +18,15 @@ jobs:
     timeout-minutes: 5
     steps:
       - name: Checkout repo
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
 
       - name: Setup Poetry
         uses: abatilo/actions-poetry@v2.0.0
         with:
           poetry-version: "1.5.1"
 
       - name: Setup python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
         with:
           python-version: "3.10"
           cache: "poetry"

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "rl-energyplus"
-version = "0.7.0"
+version = "0.8.0"
 description = "EnergyPlus Gym Environments for Reinforcement Learning"
 authors = ["Antoine Galataud <antoine@foobot.io>"]
 packages = [
@@ -13,7 +13,7 @@ priority = "primary"
 
 [tool.poetry.dependencies]
 python = ">=3.8,<3.12"
-ray = { version = "2.8.0", extras = ["default", "rllib"] }
+ray = { version = "2.12.0", extras = ["default", "rllib"] }
 gymnasium = "0.28.1"
 numpy = "1.23.5"
 protobuf = "3.20.3"

rleplus/train/pearl.py

@@ -12,7 +12,9 @@
 from pearl.history_summarization_modules.lstm_history_summarization_module import (
     LSTMHistorySummarizationModule,
 )
-from pearl.neural_networks.common.value_networks import EnsembleQValueNetwork
+from pearl.neural_networks.sequential_decision_making.q_value_networks import (
+    EnsembleQValueNetwork,
+)
 from pearl.pearl_agent import PearlAgent
 from pearl.policy_learners.sequential_decision_making.bootstrapped_dqn import (
     BootstrappedDQN,

rleplus/train/rllib.py

@@ -30,7 +30,7 @@ def parse_args() -> argparse.Namespace:
     )
     parser.add_argument(
         "--output",
-        help="EnergyPlus output directory. Default is a generated one in /tmp/",
+        help="EnergyPlus output directory. Default is a generated one in ~/ray_results/",
         required=False,
         default=TemporaryDirectory().name,
     )
@@ -47,16 +47,10 @@ def parse_args() -> argparse.Namespace:
         default=0,
         help="The number of GPUs to use",
     )
-    parser.add_argument(
-        "--alg",
-        default="PPO",
-        choices=["APEX", "DQN", "IMPALA", "PPO", "R2D2"],
-        help="The algorithm to use",
-    )
     parser.add_argument(
         "--use-lstm",
         action="store_true",
-        help="Whether to auto-wrap the model with an LSTM. Only valid option for " "--run=[IMPALA|PPO|R2D2]",
+        help="Whether to auto-wrap the model with an LSTM",
     )
     built_args = parser.parse_args()
     print(f"Running with following CLI args: {built_args}")
@@ -90,8 +84,8 @@ def main():
                 "use_lstm": args.use_lstm,
                 "vf_share_layers": False,
             },
-            _enable_learner_api=True,
         )
+        .experimental(_enable_new_api_stack=True)
         .rl_module(_enable_rl_module_api=True)
         .framework(
             # to use tensorflow, you'll need install it first,

tests/test_train.py

@@ -10,5 +10,4 @@ def test_ray_runner_config(self):
         self.assertEqual(args.timesteps, 1e6)
         self.assertEqual(args.num_workers, 2)
         self.assertEqual(args.num_gpus, 0)
-        self.assertEqual(args.alg, "PPO")
         self.assertFalse(args.use_lstm)