Merge pull request #59 from WiemKhlifi/fix/fix_cql_loss

fix: fix CQL loss

Author: jcformanek

og_marl/tf2_systems/offline/iql_cql.py

@@ -226,7 +226,7 @@ def _tf_train_step(self, train_step: int, experience: Dict[str, Any]) -> Dict[st
             #############
 
             # Mask out zero-padded timesteps
-            loss = td_loss + self.cql_weight + cql_loss
+            loss = td_loss + self.cql_weight * cql_loss
 
         # Get trainable variables
         variables = (*self.q_network.trainable_variables,)