Hi, I have used sample_numpyro_nuts because the sampler is less time consuming. It even works well with pm.Categorial() with observed; however, if I want to make predictions with the new data, it poses a problem. with pm.Model() as model1: sigma_ard = pm.Gamma("sigma_ard", alpha = 2, beta = 0.…

NUTS, which numpyro uses can’t sample discrete variables. To make predictions you can use sample_posterior_predictive instead. Just recreate the model after you’re done sampling with the extra variable and include it in var_names. This may help understand the uses: Out of model predictions with PyM…

pm.Categorical with sample_numpyro_nuts

Questions v5

jessegrabowski November 24, 2023, 8:19am 3

Does this model need an predictive model? It looks like you can just use pm.set_data here

1 Like

Topic		Replies	Views
Use of numpyro/Jax with pymc-dev v5 jax	5	2359	July 28, 2022
TypeError in NUTS Questions	5	666	May 9, 2018
Returning posterior predictive probabilities for multiple categories Questions development	2	662	November 20, 2018
Generating Posterior Predictive Samples Over Forecast Period Questions	13	1267	January 2, 2020
NUTS - FromFunctionOp has no attribute 'grad' Questions theano	2	1005	August 3, 2018

pm.Categorical with sample_numpyro_nuts

Related topics