Traceback (most recent call last):
File "example_agent.py", line 22, in <module>
sys_agent.response("I want to find a moderate hotel")
File "/exp/tatk/tatk/dialog_agent/agent.py", line 109, in response
self.output_action = self.policy.predict(state)
File "/exp/tatk/tatk/policy/rule/multiwoz/rule.py", line 31, in predict
return self.policy.predict(state)
File "/exp/tatk/tatk/policy/rule/multiwoz/rule_based_multiwoz_bot.py", line 97, in predict
domain, intent_type = user_act.split('-')
AttributeError: 'list' object has no attribute 'split'
To Reproduce
I try to run a agent according tutorials/Getting_Started/Getting_Started.ipynb as follows:
import sys
import os
from tatk.dialog_agent import PipelineAgent, BiSession
from tatk.nlu.svm.multiwoz import SVMNLU
from tatk.nlu.jointBERT.multiwoz import BERTNLU
from tatk.dst.rule.multiwoz import RuleDST
from tatk.policy.rule.multiwoz import Rule
from tatk.nlg.template.multiwoz import TemplateNLG
from tatk.evaluator.multiwoz_eval import MultiWozEvaluator
import random
import numpy as np
from pprint import pprint
sys_nlu = SVMNLU(mode='usr')
sys_dst = RuleDST()
sys_policy = Rule(character='sys')
sys_nlg = TemplateNLG(is_user=False)
sys_agent = PipelineAgent(sys_nlu, sys_dst, sys_policy, sys_nlg, "chatbot")
sys_agent.response("I want to find a moderate hotel")
Describe the bug
To Reproduce I try to run a agent according tutorials/Getting_Started/Getting_Started.ipynb as follows:
before run this, I change some code to fix
self.history not define
bug: change https://github.com/thu-coai/tatk/blob/fafabc45d02ad889f59354acac4e3b1367e7d4bf/tatk/dialog_agent/agent.py#L96 toExpected behavior print 'I have 18 options for you. Will acorn guest house be alright ? The parking is free . How about acorn guest house ?'