Merge pull request #22 from rte-france/bd-dev

Update to version  0.5.0

CHANGELOG.rst

@@ -4,8 +4,10 @@ Change Log
 --------
 - stack multiple states in `utils/DeepQAgent`
 
-[0.5.0] - 2020-08-??
+[0.5.0] - 2020-08-18
 --------------------
+- [BREAKING] remove the SAC baseline that was not correct. For backward compatibility, its code
+  can still be accessed with SACOld
 - [FIXED] the counting of the action types frequency in tensorboard (for some baselines)
 - [FIXED] a broken Replay buffer `utils.ReplayBuffer` (used in some baselines)
 - [FIXED] a bug in using multiple environments for some baselines

docs/conf.py

@@ -22,8 +22,8 @@
 author = 'Benjamin DONNOT'
 
 # The full version, including alpha/beta/rc tags
-release = '0.4.4'
-version = '0.4'
+release = '0.5.0'
+version = '0.5'
 
 # -- General configuration ---------------------------------------------------
 

docs/index.rst

@@ -28,7 +28,6 @@ Baseline already Available
    DeepQSimple
    DoubleDuelingDQN
    DuelQSimple
-   SAC
 
 
 More advanced baselines

l2rpn_baselines/LeapNetEncoded/LeapNetEncoded_NN.py

@@ -243,7 +243,7 @@ def predict_movement(self, data, epsilon, batch_size=None, training=False):
         if batch_size is None:
             batch_size = data.shape[0]
         data_nn, true_output_grid = self._make_x_tau(data)
-        res = super().predict_movement(data_nn, epsilon=epsilon, batch_size=batch_size, training=False)
+        res = super().predict_movement(data_nn, epsilon=epsilon, batch_size=batch_size, training=training)
         return res
 
     def train(self, s_batch, a_batch, r_batch, d_batch, s2_batch, tf_writer=None, batch_size=None):

l2rpn_baselines/LeapNetEncoded/LeapNetEncoded_NNParam.py

@@ -175,3 +175,12 @@ def center_reduce(self, env):
         self._center_reduce_vect(env.get_obs(), "tau")
         self._center_reduce_vect(env.get_obs(), "gm_out")
         self._center_reduce_vect(env.get_obs(), "input_q")
+
+    def _get_adds_mults_from_name(self, obs, attr_nm):
+        add_tmp, mult_tmp = super()._get_adds_mults_from_name(obs, attr_nm)
+        if attr_nm in ["line_status"]:
+            # transform time step overflow into (1. - timestep_overflow) [similar to the leap net papers]
+            # 0 powerline is connected, 1 powerline is NOT connected
+            add_tmp = -1.0
+            mult_tmp = -1.0
+        return add_tmp, mult_tmp

l2rpn_baselines/LeapNetEncoded/study.py

@@ -56,10 +56,10 @@ def study(env,
     # Run
     # Create agent
     agent = LeapNetEncoded(action_space=env.action_space,
-                        name=name,
-                        store_action=nb_process == 1,
-                        nn_archi=nn_archi,
-                        observation_space=env.observation_space)
+                           name=name,
+                           store_action=nb_process == 1,
+                           nn_archi=nn_archi,
+                           observation_space=env.observation_space)
 
     # Load weights from file
     agent.load(load_path)

l2rpn_baselines/LeapNetEncoded/train.py

@@ -389,7 +389,8 @@ def __call__(self, action, env, has_error, is_done, is_illegal, is_ambiguous):
                            "day_of_week",
                            "hour_of_day",
                            "minute_of_hour",
-                           "rho"]
+                           "rho",
+                           ]
     li_attr_obs_Tau = ["line_status", "timestep_overflow"]
     list_attr_gm_out = ["a_or", "a_ex", "p_or", "p_ex", "q_or", "q_ex", "prod_q", "load_v"] + li_attr_obs_X
 
@@ -405,7 +406,7 @@ def __call__(self, action, env, has_error, is_done, is_illegal, is_ambiguous):
 
                     'dim_topo': env_init.dim_topo,
 
-                    "sizes_enc": (50, 50, 50, 50),
+                    "sizes_enc": (50, 50, ),
                     "sizes_main": (300, 300, 300),
                     "sizes_out_gm": (100, ),
                     "sizes_Qnet": (200, 200, 200)