How to model time-dependent variables in pymc3

@junpenglao - and how would you implement an integrated Random Walk in pymc3? My simplistic thinking is:

rw1 = pm.GaussianRandomWalk('rw1',sd=1., shape=lendata)
slope = pm.Deterministic('slope', pm.math.cumsum( rw1))

except there’s nopm.math.cumsum().… My question originated after reading this vignette:

https://cran.r-project.org/web/packages/walker/vignettes/walker.html