dynamic_model dimensionality problem

[akattapuram-insitro]

dimensions seem to be non-conformable when scaling dm.alpha, beta, gamma, t, tau, pars

.pars[:3] is a numpy array of dimension (3,10) (in our dataset) whereas the other variables are floats. Safer to scale individually to catch edge cases such as this one.

if dm is not None and dm.recoverable:
        dm.m = m[idx]

        #dm.alpha, dm.beta, dm.gamma, dm.pars[:3] = (                                                                                                   
        #    np.array([dm.alpha, dm.beta, dm.gamma, dm.pars[:3]]) / dm.m[-1]                                                                            
        #)                                                                                                                                              

        dm.alpha=dm.alpha/dm.m[-1]
        dm.beta=dm.beta/dm.m[-1]
    dm.gamma=dm.gamma/dm.m[-1]
        dm.pars[:3]=dm.pars[:3]/dm.m[-1]

        #dm.t, dm.tau, dm.t_, dm.pars[4] = (                                                                                                            
        #    np.array([dm.t, dm.tau, dm.t_, dm.pars[4]]) * dm.m[-1]                                                                                     
    #)                                                                                                                                              
        dm.t=dm.t*dm.m[-1]
        dm.tau=dm.tau*dm.m[-1]
        dm.t_=dm.t_*dm.m[-1]
        dm.pars[4]=dm.pars[4]*dm.m[-1]

(the commented out lines are what's in scvelo, the non-commented out lines are the proposed fix)

[WeilerP]

Duplicate of #1087. Please check open/resolved issues and PRs before opening new ones and use the provided issue templates.