No free random variables to sample.

I'm new to pymc3 and trying to learn how to use it. I was following allowing with your video here: https://www.youtube.com/watch?v=rZvro4-nFIk

downloaded your github, and started stepping through the code. Imports and setup section went fine (I have all the requirements) Here is the Linear regression output:

print('Running on PyMC3 v{}'.format(pm.__version__))# Set up basic parameters
num_features = 10
num_observed = 1000

--> Running on PyMC3 v3.7

# Choose random values for the actual alpha and betas
alpha_a = random.normal(size=1)

betas_a = random.normal(size = num_features)

# Create fake predictor data
X_train = random.normal(size=(num_observed, num_features))

# Calculate the actual data, but put a bit of noise in
y_a = alpha_a + nsum(betas_a[None,:] * X_train, 1) + random.normal(size=(num_observed))

# Set up the PyMC model
lin_reg_model = pm.Model()
with lin_reg_model:

    #Note: You can parametrize your functions by either tau or sigma 
    #where tau = 1/sigma^2
    # This is a change from PyMC2
    alpha = pm.Normal('alpha', mu=0, tau=10.**-2, shape=(1))
    betas = pm.Normal('betas', mu=0, tau=10. ** -2, shape=(1, num_features))

    # Simulate the noise
    s = pm.HalfNormal('s', tau=1)

    temp = alpha + T.dot(betas, X_train.T)

    y = pm.Normal('y', mu=temp , tau=s ** -2, observed=y_a)

I had no errors in any of the jupyter cells so far. But the next cell was:

lin_reg_model = pm.Model()# Sample from the model
with lin_reg_model:
    step = pm.NUTS()
    nuts_trace = pm.sample(2e3, step)

and it said this:


ValueError                                Traceback (most recent call last)
<ipython-input-13-da803d520ecc> in <module>
      1 lin_reg_model = pm.Model()# Sample from the model
      2 with lin_reg_model:
----> 3     step = pm.NUTS()
      4     nuts_trace = pm.sample(2e3, step)

~/miniconda3/envs/p37cu10pyToTf/lib/python3.7/site-packages/pymc3/step_methods/arraystep.py in __new__(cls, *args, **kwargs)
     53 
     54         if len(vars) == 0:
---> 55             raise ValueError('No free random variables to sample.')
     56 
     57         if not blocked and len(vars) > 1:

ValueError: No free random variables to sample.

help?

parsing-science / pymc3_quickstart_guide

No free random variables to sample. #2