Merge pull request #116 from Unity-Technologies/develop-4

[WIP] Update to 4.0 release

README.md

@@ -44,6 +44,10 @@ To learn more, please read our AAAI Workshop paper:
    * Added evaluation script for benchmarking, along with official benchmarking guidelines.
 * v3.1 Minor release.
    * Updates Unity project to 2019.2 and ML-Agents version to 0.10.
+* 4.0 Major project updates
+   * Unity project updated to 2019.4 LTS.
+   * ML-Agents package updated to 1.1.0 (Release 3).
+   * Gym wrapper now uses registry by default to auto-download binary.
 
 
 ## Installation
@@ -54,19 +58,21 @@ The Obstacle Tower environment runs on Mac OS X, Windows, or Linux.
 
 Python dependencies (also in [setup.py](https://github.com/Unity-Technologies/obstacle-tower-env/blob/master/setup.py)):
 
-* Python 3.6
-* Unity ML-Agents v0.10
+* Python 3.6+
+* Unity ML-Agents 1.x
 * OpenAI Gym
 
-### Download the environment
+### Download the environment (optional)
+
+By default, the binary will be automatically downloaded when the Obstacle Tower gym is first instantiated. The binaries for each platform can be separately downloaded at the following links.
 
 | *Platform*     | *Download Link*                                                                     |
 | --- | --- |
-| Linux (x86_64) | https://storage.googleapis.com/obstacle-tower-build/v3.1/obstacletower_v3.1_linux.zip   |
-| Mac OS X       | https://storage.googleapis.com/obstacle-tower-build/v3.1/obstacletower_v3.1_osx.zip     |
-| Windows        | https://storage.googleapis.com/obstacle-tower-build/v3.1/obstacletower_v3.1_windows.zip |
+| Linux (x86_64) | https://storage.googleapis.com/obstacle-tower-build/v4.0/obstacletower_v4.0_linux.zip   |
+| Mac OS X       | https://storage.googleapis.com/obstacle-tower-build/v4.0/obstacletower_v4.0_osx.zip     |
+| Windows        | https://storage.googleapis.com/obstacle-tower-build/v4.0/obstacletower_v4.0_windows.zip |
 
-For checksums on these files, see [here](https://storage.googleapis.com/obstacle-tower-build/v3.1/ote-v3.1-checksums.txt).
+For checksums on these files, see [here](https://storage.googleapis.com/obstacle-tower-build/v4.0/ote-v4.0-checksums.txt).
 
 ### Install the Gym interface
 

examples/basic_usage.ipynb

@@ -23,7 +23,7 @@
    "metadata": {},
    "source": [
     "## Launching the environment\n",
-    "Ensure that the Obstacle Tower binary has been downloaded (https://github.com/Unity-Technologies/obstacle-tower-env#download-the-environment), and placed in the correct sub-folder. Here we use the `examples/ObstacleTower` sub-folder."
+    "The binary for Obstacle Tower will be automatically downloaded when the environment is first initialized."
    ]
   },
   {
@@ -36,7 +36,7 @@
     "# as well as whether the environment will run at realtime speed. Set this to `True`\n",
     "# to visual the agent behavior as you would in player mode.\n",
     "\n",
-    "env = ObstacleTowerEnv('./ObstacleTower/obstacletower', retro=False, realtime_mode=False)"
+    "env = ObstacleTowerEnv(retro=False, realtime_mode=False)"
    ]
   },
   {
@@ -174,7 +174,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "env = ObstacleTowerEnv('./ObstacleTower/obstacletower', retro=True)"
+    "env = ObstacleTowerEnv(retro=True)"
    ]
   },
   {
@@ -247,7 +247,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.6"
+   "version": "3.8.3"
   }
  },
  "nbformat": 4,

examples/evaluation.py

@@ -4,33 +4,33 @@
 def run_episode(env):
     done = False
     episode_return = 0.0
-    
+
     while not done:
         action = env.action_space.sample()
         obs, reward, done, info = env.step(action)
         episode_return += reward
     return episode_return
 
 
-if __name__ == '__main__':
-    # In this example we use the seeds used for evaluating submissions 
+if __name__ == "__main__":
+    # In this example we use the seeds used for evaluating submissions
     # to the Obstacle Tower Challenge.
     eval_seeds = [1001, 1002, 1003, 1004, 1005]
 
     # Create the ObstacleTowerEnv gym and launch ObstacleTower
-    env = ObstacleTowerEnv('./ObstacleTower/obstacletower')
+    env = ObstacleTowerEnv("./ObstacleTower/obstacletower")
 
     # Wrap the environment with the ObstacleTowerEvaluation wrapper
     # and provide evaluation seeds.
     env = ObstacleTowerEvaluation(env, eval_seeds)
 
-    # We can run episodes (in this case with a random policy) until 
+    # We can run episodes (in this case with a random policy) until
     # the "evaluation_complete" flag is True.  Attempting to step or reset after
     # all of the evaluation seeds have completed will result in an exception.
     while not env.evaluation_complete:
         episode_rew = run_episode(env)
 
-    # Finally the evaluation results can be fetched as a dictionary from the 
+    # Finally the evaluation results can be fetched as a dictionary from the
     # environment wrapper.
     print(env.results)