Curious Actor Optimizer using Actor model parameters

[someone-ml]

https://github.com/yusukeurakami/plan2explore-pytorch/blob/13c13bd6c206742fd25d68ab693a5b5271b5b34a/main.py#L149

Hello, I was wondering, when using p2e algorithm, why do we pass actor_model parameters into curious_actor_optimizer? Same happens with value_model_optimizer and value_model.

In the way I see it, this is bug because curious_actor_model doesn't learn anything while exploring environment with p2e and the way to fix it is to pass curious_actor_model parameters and curious_value_model parameters into curious_actor_optimizer and value_model_optimizer respectively.

[seolhokim]

you're right