OSL for dialogue / cannot do it...

pydial/Simulate.py

@@ -130,11 +130,11 @@ def run_dialogs(self, numDialogs):
         '''
         for i in range(numDialogs):
             logger.info('Dialogue %d' % (i+1))
-            self.run(session_id='simulate_dialog'+str(i), sim_level=self.sim_level)
+            self.run(session_id='simulate_dialog'+str(i), sim_level=self.sim_level, roi=roi)
 
         self.agent_factory.power_down_factory() # Important! -uses FORCE_SAVE on policy- which will finalise learning and save policy.
 
-    def run(self, session_id, agent_id='Smith', sim_level='dial_act'):
+    def run(self, session_id, agent_id='Smith', sim_level='dial_act', roi=False):
         '''
         Runs one episode through the simulator
         
@@ -146,8 +146,12 @@ def run(self, session_id, agent_id='Smith', sim_level='dial_act'):
         '''
 
         # GENERATE A USER PREFERENCE: a * Length + (1-a) * Success
+
+
         preference = torch.randn(2)
         preference = (torch.abs(preference) / torch.norm(preference, p=1)).type(FloatTensor)
+        if roi:
+            pass
         logger.dial('User\'s preference: [{}, {}]'.format(preference[0], preference[1]))
 
         # RESET THE USER SIMULATOR:

pydial/policy/PolicyManager.py

@@ -300,6 +300,9 @@ def _load_domains_policy(self, domainString=None):
             elif policy_type == 'morl':
                 from policy import MORLPolicy
                 self.domainPolicies[domainString] = MORLPolicy.MORLPolicy(in_policy_file, out_policy_file, domainString, learning)
+            elif policy_type == 'roi-morl':
+                from policy import RoiMORLPolicy
+                self.domainPolicies[domainString] = RoiMORLPolicy.RoiMORLPolicy(in_policy_file, out_policy_file, domainString, learning)
             else:
                 try:
                     # try to view the config string as a complete module path to the class to be instantiated