Closed till2 closed 1 year ago
Hi Danijar,
I wanted ask whether there is a negative sign missing in front of the $sg(R^\lambda_t)/\max(1,S)$ in eq. 11 in the D-V3 paper. I think the negative return (or advantage) should be minimized, right?
Best, Till
You're right, it'll be fixed in the next update.
Hi Danijar,
I wanted ask whether there is a negative sign missing in front of the $sg(R^\lambda_t)/\max(1,S)$ in eq. 11 in the D-V3 paper. I think the negative return (or advantage) should be minimized, right?
Best, Till