Error modelling Collider -- -- bayesian Regression

Questions v5
,
2

for reference the following model should exclude the confounding collider but I get the same error

with pm.Model() as exclude_collider_model:
    x = pm.Data('x', np.array(data['x'].values,dtype=float))
    y = pm.Data('y', np.array(data['y'].values,dtype=float))
    # y= 
    alpha = pm.Normal('intercept',0,1)
    beta_x = pm.Normal('slope_x',0,1)
    sigma = pm.Exponential('error',lam=1)
    
    mu = alpha + beta_x*x
    y_hat = pm.Normal('y_hat',0,sigma,observed=y-mu)
    trace_exclude_collider = pm.sample(1000)