Efficient spin-charge conversion is indispensable in spintronic computing and memory technologies for achieving low power. Spin Hall and quantum spin Hall effects have been demonstrated in certain atomically thin two-dimensional (2D) materials. In this work, we develop a multiscale simulation method from atomistic quantum transport simulation to the circuit model for the spin Hall effect (SHE) device based on 2D materials. Numerical implementation to speed up atomistic transport simulations with the non-equilibrium Green's function formalism is described for the crossbar SHE device. The multiscale method can treat atomistic scale features and compute the spintronic device performance metrics of the modeled device. As an example, the effect of edge roughness on the SHE devices based on monolayer 2D materials is investigated. The results illustrate that aggressively scaled monolayer SHE devices can efficiently transduce charge to spin in the presence of edge roughness.