Update benchmark + doc

araffin · Nov 21, 2018 · 866e273 · 866e273
1 parent fe36594
commit 866e273
Show file tree

Hide file tree

Showing 5 changed files with 18 additions and 34 deletions.
diff --git a/README.md b/README.md
@@ -1,6 +1,6 @@
 [![Build Status](https://travis-ci.com/araffin/rl-baselines-zoo.svg?branch=master)](https://travis-ci.com/araffin/rl-baselines-zoo)
 
-# RL Baselines Zoo: a Collection of Trained Reinforcement Learning Agents
+# RL Baselines Zoo: a Collection of Pre-Trained Reinforcement Learning Agents
 
 <img src="images/BipedalWalkerHardcorePPO2.gif" align="right" width="35%"/>
 
@@ -52,6 +52,15 @@ Continue training (here, load pretrained agent for Breakout and continue trainin
 python train.py --algo a2c --env BreakoutNoFrameskip-v4 -i trained_agents/a2c/BreakoutNoFrameskip-v4.pkl -n 5000
 ```
 
+## Record a Video of a Trained Agent
+
+Record 1000 steps:
+
+```
+python -m utils.record_video --algo ppo2 --env BipedalWalkerHardcore-v2 -n 1000
+```
+
+
 ## Current Collection: 70+ Trained Agents!
 
 Scores can be found in `benchmark.md`. To compute them, simply run `python -m utils.benchmark`.
@@ -64,7 +73,7 @@ Scores can be found in `benchmark.md`. To compute them, simply run `python -m ut
 |----------|--------------------|--------------------|--------------------|-------|-------|--------------------|--------------------|
 | A2C      | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |:heavy_check_mark:| :heavy_check_mark: | :heavy_check_mark:| :heavy_check_mark:|
 | ACER     | :heavy_check_mark: |                    |:heavy_check_mark:|:heavy_check_mark: |:heavy_check_mark:|:heavy_check_mark:| :heavy_check_mark: |
-| ACKTR    |:heavy_check_mark:| :heavy_check_mark:|                    |:heavy_check_mark:| :heavy_check_mark:| :heavy_check_mark:|  :heavy_check_mark: |
+| ACKTR    |:heavy_check_mark:| :heavy_check_mark:|:heavy_check_mark:|:heavy_check_mark:| :heavy_check_mark:| :heavy_check_mark:|  :heavy_check_mark: |
 | PPO2     |:heavy_check_mark:|:heavy_check_mark:| :heavy_check_mark: |:heavy_check_mark: |:heavy_check_mark:|:heavy_check_mark:|  :heavy_check_mark: |
 | DQN     |:heavy_check_mark:| :heavy_check_mark: |:heavy_check_mark:| :heavy_check_mark:|:heavy_check_mark:|:heavy_check_mark:|:heavy_check_mark:|
 
@@ -73,7 +82,7 @@ Additional Atari Games (to be completed):
 |  RL Algo |  MsPacman   |
 |----------|-------------|
 | A2C      |:heavy_check_mark:|
-| ACER     | |
+| ACER     |:heavy_check_mark:|
 | ACKTR    | |
 | PPO2     |:heavy_check_mark: |
 | DQN      | |

diff --git a/benchmark.md b/benchmark.md
@@ -18,6 +18,7 @@
 |acer |EnduroNoFrameskip-v4               |      0.000|     0.000|     149574|        45|
 |acer |LunarLander-v2                     |    185.210|    64.829|     149415|       248|
 |acer |MountainCar-v0                     |   -131.213|    32.541|     149976|      1143|
+|acer |MsPacmanNoFrameskip-v4             |   3908.105|   585.407|     148924|        95|
 |acer |PongNoFrameskip-v4                 |     20.667|     0.507|     148275|        57|
 |acer |QbertNoFrameskip-v4                |  18880.469|  1648.937|     148617|        64|
 |acer |SeaquestNoFrameskip-v4             |    872.121|    25.555|     149650|        66|
@@ -26,6 +27,7 @@
 |acktr|BeamRiderNoFrameskip-v4            |   3760.976|  1826.059|     147414|        41|
 |acktr|BreakoutNoFrameskip-v4             |    448.514|    88.882|     143118|        37|
 |acktr|CartPole-v1                        |    487.573|    63.866|     149685|       307|
+|acktr|EnduroNoFrameskip-v4               |      0.000|     0.000|     149574|        45|
 |acktr|LunarLander-v2                     |     96.822|    64.020|     149905|       176|
 |acktr|MountainCar-v0                     |   -111.917|    21.422|     149969|      1340|
 |acktr|PongNoFrameskip-v4                 |     19.224|     3.697|     147753|        67|

diff --git a/hyperparams/ddpg.yml b/hyperparams/ddpg.yml
@@ -30,44 +30,17 @@ BipedalWalker-v2:
 Walker2DBulletEnv-v0:
   n_timesteps: !!float 2e6
   policy: 'LnMlpPolicy'
-  noise_type: 'adaptive-param'
-  noise_std: 0.2
+  noise_type: 'ornstein-uhlenbeck'
+  noise_std: 0.1
   memory_limit: 1000000
   batch_size: 64
   normalize_observations: True
 
 HalfCheetahBulletEnv-v0:
   n_timesteps: !!float 2e6
   policy: 'LnMlpPolicy'
-  noise_type: 'adaptive-param'
-  noise_std: 0.2
-  memory_limit: 1000000
-  batch_size: 64
-  normalize_observations: True
-
-AntBulletEnv-v0:
-  n_timesteps: !!float 2e6
-  policy: 'LnMlpPolicy'
-  noise_type: 'adaptive-param'
-  noise_std: 0.2
-  memory_limit: 1000000
-  batch_size: 64
-  normalize_observations: True
-
-HopperBulletEnv-v0:
-  n_timesteps: !!float 2e6
-  policy: 'LnMlpPolicy'
-  noise_type: 'adaptive-param'
-  noise_std: 0.2
-  memory_limit: 1000000
-  batch_size: 64
-  normalize_observations: True
-
-ReacherBulletEnv-v0:
-  n_timesteps: !!float 2e6
-  policy: 'LnMlpPolicy'
-  noise_type: 'adaptive-param'
-  noise_std: 0.2
+  noise_type: 'ornstein-uhlenbeck'
+  noise_std: 0.1
   memory_limit: 1000000
   batch_size: 64
   normalize_observations: True
diff --git a/trained_agents/acer/MsPacmanNoFrameskip-v4.pkl b/trained_agents/acer/MsPacmanNoFrameskip-v4.pkl
diff --git a/trained_agents/acktr/EnduroNoFrameskip-v4.pkl b/trained_agents/acktr/EnduroNoFrameskip-v4.pkl