ReturnEstimates#

class hybrid_aif.train.ReturnEstimates[source]#

Bases: NamedTuple

Per-state λ-returns and critic values over a batch of imagined rollouts.

Entry [i, k] is about the k-th state of trajectory i.

ret: Array#

The return estimate at each state, (N, H).

critic: Array#

The critic’s value of each state, (N, H).

value: Array#

The per-step G as a value that the return was folded from.

static __new__(_cls, ret, critic, value)#

Create new instance of ReturnEstimates(ret, critic, value)