hello!
When I was reading a2ccommon source code. I find a question. Why we need to initialize self.dones tensor ones? That will mark first translation (s,a,s,r, done=1). I don't know if this is a mistake. Or I don't understand the code logic correctly.
hello! When I was reading a2ccommon source code. I find a question. Why we need to initialize self.dones tensor ones? That will mark first translation (s,a,s,r, done=1). I don't know if this is a mistake. Or I don't understand the code logic correctly.