Initial commit

README.md

@@ -16,7 +16,7 @@ model-index:
       type: SpaceInvadersNoFrameskip-v4
     metrics:
     - type: mean_reward
-      value: 374.00 +/- 214.89
+      value: 287.50 +/- 64.16
       name: mean_reward
       verified: false
 ---
@@ -62,15 +62,16 @@ python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f lo
 
 ## Hyperparameters
 ```python
-OrderedDict([('batch_size', 32),
-             ('buffer_size', 50000),
+OrderedDict([('batch_size', 16),
+             ('buffer_size', 70000),
              ('env_wrapper',
               ['stable_baselines3.common.atari_wrappers.AtariWrapper']),
              ('exploration_final_eps', 0.01),
              ('exploration_fraction', 0.1),
+             ('exploration_initial_eps', 0.03),
              ('frame_stack', 4),
              ('gradient_steps', 1),
-             ('learning_rate', 0.0001),
+             ('learning_rate', 5e-05),
              ('learning_starts', 100000),
              ('n_timesteps', 1000000.0),
              ('optimize_memory_usage', False),

args.yml

@@ -16,7 +16,10 @@
   - - gym_packages
     - []
   - - hyperparams
-    - null
+    - batch_size: 16
+      buffer_size: 70000
+      exploration_initial_eps: 0.03
+      learning_rate: 5.0e-05
   - - log_folder
     - logs/
   - - log_interval
@@ -54,7 +57,7 @@
   - - save_replay_buffer
     - false
   - - seed
-    - 682855770
+    - 3943982455
   - - storage
     - null
   - - study_name
@@ -64,7 +67,7 @@
   - - track
     - false
   - - trained_agent
-    - ''
+    - logs/dqn/SpaceInvadersNoFrameskip-v4_11/best_model.zip
   - - truncate_last_trajectory
     - true
   - - uuid

config.yml

@@ -1,20 +1,22 @@
 !!python/object/apply:collections.OrderedDict
 - - - batch_size
-    - 32
+    - 16
   - - buffer_size
-    - 50000
+    - 70000
   - - env_wrapper
     - - stable_baselines3.common.atari_wrappers.AtariWrapper
   - - exploration_final_eps
     - 0.01
   - - exploration_fraction
     - 0.1
+  - - exploration_initial_eps
+    - 0.03
   - - frame_stack
     - 4
   - - gradient_steps
     - 1
   - - learning_rate
-    - 0.0001
+    - 5.0e-05
   - - learning_starts
     - 100000
   - - n_timesteps

dqn-SpaceInvadersNoFrameskip-v4.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de3f8e78936e95fe66aabb761622ff2eb0c546ee022b51b59f882cd084b35f1c
-size 13719681
+oid sha256:3ba5deaa07347a31876f8c5b221a849aecd380900bcd15b8f2f250c7d101e2a8
+size 13719558

replay.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:322cf0cd61bf5e0031f1dcdc699552efb08db7bd19365cc56f504f76f4a212eb
-size 234682
+oid sha256:c6f8a7f3081e0696ec7b6575dda514c1e6c01d4dd70027ecd4f3739dbac4cf3f
+size 256560

results.json

@@ -1 +1 @@
-{"mean_reward": 374.0, "std_reward": 214.89299662855464, "is_deterministic": false, "n_eval_episodes": 10, "eval_datetime": "2023-02-20T03:47:57.282577"}
+{"mean_reward": 287.5, "std_reward": 64.15800807381726, "is_deterministic": false, "n_eval_episodes": 10, "eval_datetime": "2023-02-20T20:15:58.443635"}

train_eval_metrics.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eea199f4702bc78439f76abfe3ab53224930438b63c7d7c9223761d773d2084d
-size 6036
+oid sha256:4be628e39e6e1984fd259b3cf123aa7625c8b69a4a63313d379c8fadbc3baebc
+size 7329