Formatting

README.md

@@ -25,7 +25,9 @@ You may need to install/update bazel: [Ubuntu guide](https://docs.bazel.build/ve
 
 ```
 pip install ray[rllib]
-pip install tensorflow # or tensorflow-gpu
+pip install tensorflow
+# with GPU support
+# conda install -c anaconda tensorflow-gpu 
 pip install -r requirements.txt
 ```
 

envs/flatland_sparse.py

@@ -39,10 +39,16 @@ class FlatlandSparse(MultiAgentEnv):
         schedule_generator = sparse_schedule_generator({float(k): float(v)
                                                         for k, v in self._config['speed_ratio_map'].items()})
 
-        env = RailEnv(width=self._config['width'], height=self._config['height'], rail_generator=rail_generator,
-                      schedule_generator=schedule_generator, number_of_agents=self._config['number_of_agents'],
-                      malfunction_generator_and_process_data=malfunction_from_params(stochastic_data),
-                      obs_builder_object=self._observation.builder(), remove_agents_at_target=False)
+        env = RailEnv(
+            width=self._config['width'],
+            height=self._config['height'],
+            rail_generator=rail_generator,
+            schedule_generator=schedule_generator,
+            number_of_agents=self._config['number_of_agents'],
+            malfunction_generator_and_process_data=malfunction_from_params(stochastic_data),
+            obs_builder_object=self._observation.builder(),
+            remove_agents_at_target=False
+        )
         return env
 
     def step(self, action_dict):

experiments/flatland_random_sparse_small/global_obs_conv_net/ppo.yaml

 flatland-random-sparse-small-global-conv-ppo:
-    run: PPO
-    env: flatland_random_sparse_small
-    stop:
-        timesteps_total: 10000000  # 1e7
-    checkpoint_freq: 10
-    checkpoint_at_end: True
-    keep_checkpoints_num: 5
-    checkpoint_score_attr: episode_reward_mean
-    config:
-        clip_rewards: True
-        clip_param: 0.1
-        vf_clip_param: 500.0
-        entropy_coeff: 0.01
-        # effective batch_size: train_batch_size * num_agents_in_each_environment [5, 10]
-        # see https://github.com/ray-project/ray/issues/4628
-        train_batch_size: 1000  # 5000
-        rollout_fragment_length: 50  # 100
-        sgd_minibatch_size: 100  # 500
-        num_sgd_iter: 10
-        num_workers: 7
-        num_envs_per_worker: 5
-        batch_mode: truncate_episodes
-        observation_filter: NoFilter
-        vf_share_layers: True
-        vf_loss_coeff: 0.5
-        num_gpus: 1
+  run: PPO
+  env: flatland_random_sparse_small
+  stop:
+    timesteps_total: 10000000  # 1e7
+  checkpoint_freq: 10
+  checkpoint_at_end: True
+  keep_checkpoints_num: 5
+  checkpoint_score_attr: episode_reward_mean
+  config:
+    clip_rewards: True
+    clip_param: 0.1
+    vf_clip_param: 500.0
+    entropy_coeff: 0.01
+    # effective batch_size: train_batch_size * num_agents_in_each_environment [5, 10]
+    # see https://github.com/ray-project/ray/issues/4628
+    train_batch_size: 1000  # 5000
+    rollout_fragment_length: 50  # 100
+    sgd_minibatch_size: 100  # 500
+    num_sgd_iter: 10
+    num_workers: 7
+    num_envs_per_worker: 5
+    batch_mode: truncate_episodes
+    observation_filter: NoFilter
+    vf_share_layers: True
+    vf_loss_coeff: 0.5
+    num_gpus: 1
 
-        env_config:
-            min_seed: 1002
-            max_seed: 213783
-            min_test_seed: 0
-            max_test_seed: 100
-            reset_env_freq: 1
-            observation: global
-            observation_config:
-                max_width: 45
-                max_height: 45
-            regenerate_rail_on_reset: True
-            regenerate_schedule_on_reset: True
-            render: False
+    env_config:
+      min_seed: 1002
+      max_seed: 213783
+      min_test_seed: 0
+      max_test_seed: 100
+      reset_env_freq: 1
+      observation: global
+      observation_config:
+        max_width: 45
+        max_height: 45
+      regenerate_rail_on_reset: True
+      regenerate_schedule_on_reset: True
+      render: False
 
-        model:
-            custom_model: global_obs_model
-            custom_options:
-                architecture: impala
+    model:
+      custom_model: global_obs_model
+      custom_options:
+        architecture: impala

experiments/flatland_random_sparse_small/tree_obs_fc_net/ppo.yaml

 flatland-random-sparse-small-tree-fc-ppo:
-    run: PPO
-    env: flatland_random_sparse_small
-    stop:
-        timesteps_total: 10000000  # 1e7
-    checkpoint_freq: 10
-    checkpoint_at_end: True
-    keep_checkpoints_num: 5
-    checkpoint_score_attr: episode_reward_mean
-    config:
-        clip_rewards: True
-        clip_param: 0.1
-        vf_clip_param: 500.0
-        entropy_coeff: 0.01
-        # effective batch_size: train_batch_size * num_agents_in_each_environment [5, 10]
-        # see https://github.com/ray-project/ray/issues/4628
-        train_batch_size: 1000  # 5000
-        rollout_fragment_length: 50  # 100
-        sgd_minibatch_size: 100  # 500
-        num_sgd_iter: 10
-        num_workers: 7
-        num_envs_per_worker: 5
-        batch_mode: truncate_episodes
-        observation_filter: NoFilter
-        vf_share_layers: True
-        vf_loss_coeff: 0.5
-        num_gpus: 1
+  run: PPO
+  env: flatland_random_sparse_small
+  stop:
+    timesteps_total: 10000000  # 1e7
+  checkpoint_freq: 10
+  checkpoint_at_end: True
+  keep_checkpoints_num: 5
+  checkpoint_score_attr: episode_reward_mean
+  config:
+    clip_rewards: True
+    clip_param: 0.1
+    vf_clip_param: 500.0
+    entropy_coeff: 0.01
+    # effective batch_size: train_batch_size * num_agents_in_each_environment [5, 10]
+    # see https://github.com/ray-project/ray/issues/4628
+    train_batch_size: 1000  # 5000
+    rollout_fragment_length: 50  # 100
+    sgd_minibatch_size: 100  # 500
+    num_sgd_iter: 10
+    num_workers: 7
+    num_envs_per_worker: 5
+    batch_mode: truncate_episodes
+    observation_filter: NoFilter
+    vf_share_layers: True
+    vf_loss_coeff: 0.5
+    num_gpus: 1
 
-        env_config:
-            min_seed: 1002
-            max_seed: 213783
-            min_test_seed: 0
-            max_test_seed: 100
-            # After how many episodes the level should be regenerated:
-            reset_env_freq: 1
-            observation: tree
-            observation_config:
-                max_depth: 2
-                shortest_path_max_depth: 30
-            regenerate_rail_on_reset: True
-            regenerate_schedule_on_reset: True
-            render: False
+    env_config:
+      min_seed: 1002
+      max_seed: 213783
+      min_test_seed: 0
+      max_test_seed: 100
+      # After how many episodes the level should be regenerated:
+      reset_env_freq: 1
+      observation: tree
+      observation_config:
+        max_depth: 2
+        shortest_path_max_depth: 30
+      regenerate_rail_on_reset: True
+      regenerate_schedule_on_reset: True
+      render: False
 
-        model:
-            fcnet_activation: relu
-            fcnet_hiddens: [256, 256]
-            vf_share_layers: True  # False
+    model:
+      fcnet_activation: relu
+      fcnet_hiddens: [256, 256]
+      vf_share_layers: True  # False