usersim_params = {}
usersim_params['max_turn'] = max_turn
usersim_params['slot_err_probability'] = params['slot_err_prob']
usersim_params['slot_err_mode'] = params['slot_err_mode']
usersim_params['intent_err_probability'] = params['intent_err_prob']
usersim_params['simulator_run_mode'] = params['run_mode']
usersim_params['simulator_act_level'] = params['act_level']
usersim_params['learning_phase'] = params['learning_phase']
usersim_params['hidden_size'] = params['dqn_hidden_size']
if usr == 0: # real user
user_sim = RealUser(movie_dictionary, act_set, slot_set, goal_set, usersim_params)
elif usr == 1:
user_sim = RuleSimulator(movie_dictionary, act_set, slot_set, goal_set, usersim_params)
world_model = ModelBasedSimulator(movie_dictionary, act_set, slot_set, goal_set, usersim_params)
agent.set_user_planning(world_model)
# elif usr == 2:
# user_sim = ModelBasedSimulator(movie_dictionary, act_set, slot_set, goal_set, usersim_params)
################################################################################
# Add your user simulator here
################################################################################
else:
pass
################################################################################
# load trained NLG model
################################################################################
nlg_model_path = params['nlg_model_path']
diaact_nl_pairs = params['diaact_nl_pairs']
nlg_model = nlg()
usersim_params['slot_err_mode'] = params['slot_err_mode']
usersim_params['intent_err_probability'] = params['intent_err_prob']
usersim_params['simulator_run_mode'] = params['run_mode']
usersim_params['simulator_act_level'] = params['act_level']
usersim_params['learning_phase'] = params['learning_phase']
usersim_params['hidden_size'] = params['dqn_hidden_size']
usersim_params['world_model_nn_type'] = params['world_model_nn_type']
usersim_params['buffer_size_unit'] = params['buffer_size_unit']
# print usr
if usr == 0:# real user
user_sim = RealUser(movie_dictionary, act_set, slot_set, goal_set, usersim_params)
elif usr == 1:
user_sim = RuleSimulator(movie_dictionary, act_set, slot_set, goal_set, usersim_params)
user_sim_planning = ModelBasedSimulator(movie_dictionary, act_set, slot_set, goal_set, usersim_params, discriminator)
agent.set_user_planning(user_sim_planning)
# elif usr == 2:
# user_sim = ModelBasedSimulator(movie_dictionary, act_set, slot_set, goal_set, usersim_params)
################################################################################
# Add your user simulator here
################################################################################
else:
pass
################################################################################
# load trained NLG model
################################################################################
nlg_model_path = params['nlg_model_path']
diaact_nl_pairs = params['diaact_nl_pairs']
nlg_model = nlg()