From c2c24b4268a26d69da682a9fe6b63d75cbc140a3 Mon Sep 17 00:00:00 2001
From: liangrz <liangrz15@mails.tsinghua.edu.cn>
Date: Tue, 9 Jun 2020 11:28:11 +0800
Subject: [PATCH] fix the policy training bug
---
convlab2/dialog_agent/env.py | 1 +
1 file changed, 1 insertion(+)
diff --git a/convlab2/dialog_agent/env.py b/convlab2/dialog_agent/env.py
index f4ed9f8..62c02aa 100755
--- a/convlab2/dialog_agent/env.py
+++ b/convlab2/dialog_agent/env.py
@@ -28,6 +28,7 @@ class Environment():
self.evaluator.add_sys_da(self.usr.get_in_da())
self.evaluator.add_usr_da(self.usr.get_out_da())
dialog_act = self.sys_nlu.predict(observation) if self.sys_nlu else observation
+ self.sys_dst.state['user_action'] = dialog_act
state = self.sys_dst.update(dialog_act)
if self.evaluator:
--
GitLab