Some more questions

[albertma-evotec]

in gentrl.py, the parameter cond_lb and cond_rb was not used in the body? What is the purpose of them? and why there are two learning-rate variable (lr_lp and lr_dec)? is lr_dec referring to the decay rate?

Thanks

def train_as_rl(self,
                    reward_fn,
                    num_iterations=100000, verbose_step=50,
                    batch_size=200,
                    cond_lb=-2, cond_rb=0,
                    lr_lp=1e-5, lr_dec=1e-6):

[Bibyutatsu]

1. Yeah, you are right cond_lb and cond_rb are not used. Still don't know what is the purpose.

2. lr_lp refers to the learning rate of Learnable Prior and lr_dec refers to the learning rate of Decoder

As in the reinforcement learning step, only the lp and dec are trained so optimizer_lp and optimizer_dec are used to update the parameters