updates deps and wrappers for version 0.4.8

.github/workflows/test-ci.yml

@@ -23,7 +23,7 @@ jobs:
       run: |
         python -m pip install --upgrade pip
         # cpu version of pytorch
-        pip install .[test,sb3]
+        pip install .[test]
     - name: Download examples
       run: |
         make download_examples

README.md

@@ -25,7 +25,7 @@ This quickstart guide will get you up and running using the Godot RL Agents libr
 1. Install the Godot RL Agents library: (if you are new to python, pip and conda, read this [guide](https://www.machinelearningplus.com/deployment/conda-create-environment-and-everything-you-need-to-know-to-manage-conda-virtual-environment/))
 
 ```bash
-pip install godot-rl[sb3]
+pip install godot-rl
 ```
 
 1. Download one, or more of [examples](https://github.com/edbeeching/godot_rl_agents_examples), such as BallChase, JumperHard, FlyBy.

godot_rl/main.py

@@ -66,6 +66,7 @@ def get_args():
     parser.add_argument("--num_gpus", default=None, type=int, help="Number of GPUs to use [only for rllib]")
     parser.add_argument("--experiment_name", default=None, type=str, help="The name of the experiment [only for rllib]")
     parser.add_argument("--viz", default=False, action="store_true", help="Whether to visualize one process")
+
     return parser.parse_known_args()
 
 

godot_rl/wrappers/stable_baselines_wrapper.py

@@ -69,7 +69,7 @@ def step_wait(self):
         raise NotImplementedError()
 
 
-def stable_baselines_training(args, extras):
+def stable_baselines_training(args, extras, n_steps=200000):
     # TODO: Add cla etc for sb3
     env = StableBaselinesGodotEnv(env_path=args.env_path, show_window=args.viz, speedup=args.speedup)
 
@@ -81,7 +81,7 @@ def stable_baselines_training(args, extras):
         n_steps=32,
         tensorboard_log="logs/log",
     )
-    model.learn(1000000)
+    model.learn(n_steps)
 
     print("closing env")
     env.close()

pyproject.toml

@@ -3,7 +3,7 @@ requires = ["setuptools>=61.0"]
 build-backend = "setuptools.build_meta"
 [project]
 name = "godot_rl"
-version = "0.4.7"
+version = "0.4.8"
 authors = [
   { name="Edward Beeching", email="edbeeching@gmail.com" },
 ]

setup.cfg

@@ -15,7 +15,9 @@ install_requires =
     tensorboard
     wget
     huggingface_hub>=0.10
-
+    gym==0.26.2
+    stable-baselines3
+    huggingface_sb3
 
 python_requires = >=3.8
 zip_safe = no
@@ -41,7 +43,7 @@ dev =
     pyyaml>=5.3.1
 
 sb3 =
-    gym==0.21
+    gym==0.26.2
     stable-baselines3
     huggingface_sb3
 
@@ -57,3 +59,14 @@ rllib =
 
 clean-rl = 
     wandb
+
+all =     
+    numpy==1.23.5
+    gym==0.26.2
+    stable-baselines3
+    huggingface_sb3
+    sample-factory
+
+    ray==2.2.0
+    ray[rllib]
+    tensorflow_probability

tests/test_sb3_training.py

@@ -0,0 +1,19 @@
+import pytest
+
+from godot_rl.core.godot_env import GodotEnv
+from godot_rl.main import get_args
+
+try:
+    from godot_rl.wrappers.stable_baselines_wrapper import stable_baselines_training
+except ImportError as e:
+
+    def stable_baselines_training(args, extras):
+        print("Import error when trying to use sb3, this is probably not installed try pip install godot-rl[sb3]")
+
+
+def test_sb3_training():
+    args, extras = get_args()
+    args.env = "gdrl"
+    args.env_path = "examples/godot_rl_JumperHard/bin/JumperHard.x86_64"
+
+    stable_baselines_training(args, extras, n_steps=10000)