def reset(self):
# _reward = np.nan
env_info_dict = self.u_env.reset(train_mode=(util.get_lab_mode() != 'dev'), config=self.env_spec.get('multiwoz'))
a, b = 0, 0 # default singleton aeb
env_info_a = self._get_env_info(env_info_dict, a)
state = env_info_a.states[b]
self.done = False
logger.debug(f'Env {self.e} reset state: {state}')
return state
def analyze_trial(trial_spec, session_metrics_list):
'''Analyze trial and save data, then return metrics'''
info_prepath = trial_spec['meta']['info_prepath']
# calculate metrics
trial_metrics = calc_trial_metrics(session_metrics_list, info_prepath)
# plot graphs
viz.plot_trial(trial_spec, trial_metrics)
# zip files
if util.get_lab_mode() == 'train':
predir, _, _, _, _, _ = util.prepath_split(info_prepath)
shutil.make_archive(predir, 'zip', predir)
logger.info(f'All trial data zipped to {predir}.zip')
return trial_metrics
def to_check_train_step():
'''Condition for running assert_trained'''
return os.environ.get('PY_ENV') == 'test' or util.get_lab_mode() == 'dev'