not reward-weighted, same as diffusion.py
( self, x_start, cond, t, )
source not stored for this graph (policy: none)
nothing calls this directly
no test coverage detected