NUTS and zero derivative error

Questions
8 
with pm.Model() as GVmodel:
    mu = pm.Normal('mu', mu=0, sigma=.01)
    nu = pm.Normal('nu', (K/3 - 1)*dt, sigma= 0.1) # K:Kurtois of nu  
    gammadiff = pm.Gamma('gammadiff', alpha=dt/nu, beta=nu) 
    sigma = pm.HalfNormal('sigma', np.sqrt(V/dt)) # V: variance of the data
    theta = pm.Normal('theta', mu=S*sigma*np.sqrt(dt)/3*nu,sigma=1 )
    likelihood = pm.Normal('likelihood', mu = mu*dt + theta*gammadiff, sigma=(sigma**2)*gammadiff, observed= data.Return) #log_return data
    step = pm.HamiltonianMC()
    trace = pm.sample(step = step)

The data is from kaggle under maxi24/eurusd
I have the same can someone help me please.