What's the difference between the synthetic/crl/energy and synthetic/crl/envelope

[JQuan1999]

Sorry, I have a problem why don't directly output HQ in the synthetic/crl/envelope/meta.py learn()，

__, Q = self.model_(Variable(torch.cat(state_batch, dim=0)),
                                Variable(w_batch), w_num=self.weight_num)
# detach since we don't want gradients to propagate
# HQ, _    = self.model_(Variable(torch.cat(next_state_batch, dim=0), volatile=True),
# 					  Variable(w_batch, volatile=True), w_num=self.weight_num)
 _, DQ = self.model(Variable(torch.cat(next_state_batch, dim=0), requires_grad=False),
                               Variable(w_batch, requires_grad=False))

but in the synthetic/crl/energy/meta.py learn()

__, Q = self.model(Variable(torch.cat(state_batch, dim=0)),
                               Variable(preference_batch), w_num=self.weight_num)
 # detach since we don't want gradients to propagate
HQ, _ = self.model(Variable(torch.cat(next_state_batch, dim=0)),
                               Variable(preference_batch), w_num=self.weight_num)

Why getting HQ takes two different approaches and what is the difference between them