Not loading the statistics_pred model weights

[khchul]

Thank you for your great work!

I'm wondering, why are there only self.model.load_state_dict() and no self.statistics_pred.load_state_dict() in def test()? Isn't it just using some randomly initialized weights?

[icantnamemyself]

Sry for the late response. The common practice is to first train the SAN framework and then evaluate its performance immediately, with test=0 for the test() function (also see in run_LongExp.py). As the statistics_pred module is already loaded before the training loop, the parameters are correctly used in the test procedure. Also, if you want to conduct only testing, it is required to reload statistics_pred explicitly in the test() function.