UpdatingPriors

updating_priors:
[Notebook url] (https://docs.pymc.io/en/v3/pymc-examples/examples/pymc3_howto/updating_priors.html):
For the following code section, to get to the posterior trace, I had to change to trace.posterior.alpha.mean(axis=0), etc.
I am not sure if this is correct. Check the code:

for _ in range(10):
    # generate more data
    X1 = np.random.randn(size)
    X2 = np.random.randn(size) * 0.2
    Y = alpha_true + beta0_true * X1 + beta1_true * X2 + np.random.randn(size)
    model = Model()
    with model:
        # Priors are posteriors from previous iteration
        alpha = from_posterior("alpha", trace.posterior.alpha.mean(axis=0)) #CA
        beta0 = from_posterior("beta0", trace.posterior.beta0.mean(axis=0)) #CA
        beta1 = from_posterior("beta1", trace.posterior.beta1.mean(axis=0)) #CA
        # Expected value of outcome
        mu = alpha + beta0 * X1 + beta1 * X2
        # Likelihood (sampling distribution) of observations
        Y_obs = Normal("Y_obs", mu=mu, sigma=1, observed=Y)
        # draw 10000 posterior samples
        trace = sample(1000)
        traces.append(trace)

But, I got the following error on trace = sample(1000):

SamplingError: Initial evaluation of model at starting point failed!
Starting values:
{'alpha_interval__': array(nan), 'beta0_interval__': array(nan), 'beta1_interval__': array(nan)}

Initial evaluation results:
{'alpha': nan, 'beta0': nan, 'beta1': nan, 'Y_obs': nan}

Start with the updating_priors notebook linked in the issue and run the code around the repeated sampling loop. Inspect the posterior trace values and dimensions used by from_posterior, then verify that the example reaches sampling without NaN initial values and that the notebook produces valid updated traces.