PyMC3 on AWS EC2

Questions
9

This is what gave me the error.

with pm.Model() as multinomial_model:
   global_alpha_mean = pm.Normal("global_alpha_mean", mu=0, sd=100)
   global_alpha_sigma = pm.HalfCauchy("global_alpha_sigma", beta=10)
   
   global_beta_mean = pm.Normal("global_beta_mean", mu=0, sd=100)
   global_beta_sigma = pm.HalfCauchy("global_beta_sigma", beta=10)

   #centered
   local_alphas = pm.Normal("local_alphas", mu=global_alpha_mean, sd=global_alpha_sigma, shape= n_labels)
   local_betas = pm.Normal("local_betas", mu=global_beta_mean, sd=global_beta_sigma, shape=(n_labels,n_cols))
   
   mu = local_alphas[labels] + (local_betas[labels] * input_var).sum(axis=1)
   
   probs = T.nnet.softmax(mu)

   y = pm.Categorical('y', p=probs, observed=target_var)
   #y = pm.Multinomial('y', n=n_counts, p=probs, observed=target_var)

with multinomial_model:
   start = pm.find_MAP()
   step = pm.HamiltonianMC()
   trace = pm.sample(3000, step=step, start=start, tune=1000, njobs=4)