Slow pymc sampling using the introductory example

Questions v5
1

Hi, I am new to PyMC and am trying the new PyMC v4 with the introductory example, but the sampling is taking forever.

Below is how I installed PyMC:

mamba create -n newenv python=3.10 arviz matplotlib numpy pandas seaborn jupyterlab
mamba activate newenv
mamba install -c conda-forge "pymc>=4"

Then I ran the example code:

import pymc as pm
import numpy as np
import arviz as az

RANDOM_SEED = 8927
rng = np.random.default_rng(RANDOM_SEED)

alpha, sigma = 1, 1
beta = [1, 2.5]

size = 100

X1 = np.random.randn(size)
X2 = np.random.randn(size) * 0.2

Y = alpha + beta[0] * X1 + beta[1] * X2 + rng.normal(size=size) * sigma`

basic_model = pm.Model()

with basic_model:
    alpha = pm.Normal('alpha', mu=0, sigma=10)
    beta = pm.Normal('beta', mu=0, sigma=10, shape=2)
    sigma = pm.HalfNormal('sigma', sigma=1)

    mu = alpha + beta[0] * X1 + beta[1] * X2

    Y_obs = pm.Normal('Y_obs', mu=mu, sigma=sigma, observed=Y)

with basic_model:
    idata = pm.sample(draws=1)

Even with only 1 draw the model have been running for almost an hour:

image

What could have gone wrong in either the package installation or the example code?

2

Welcome!

I would begin by following the official installation instructions found here. Are there any warnings or other messages upon importing PyMC or when running your code?

3

Also, which OS is this?

4

There are no warnings when I import PyMC. The official instructions to install is below:

conda create -c conda-forge -n pymc_env "pymc>=4"
conda activate pymc_env

I basically just replaced conda with mamba based on the suggestion in this github issue

My OS is Windows 11.

5

Are you running this in a notebook for as a script invoked from the command line (or some other way?)?

6

I am running this in VS Code Notebook cells.

7

Actually I found the issue, I was running the code in a quarto (qmd) document, moving the code to jupyter notebook fixed the issue.

image
image922×481 36.4 KB

1 Like
8

I might try opening a python at the terminal (e.g., ipython) and importing PyMC there. Notebooks have a nasty habit of hiding warnings and other diagnostics. If nothing comes up there, I might suggest trying to run the following and see what happens:

import pymc as pm
if __name__ == "__main__":
  with pm.Model() as model:
      a = pm.Normal("a")
      b = pm.Normal("b", mu=a, sigma=1, observed=[1,2,3])
      idata = pm.sample()
9

Thank you, I will try that. I think Notebook is working, and Quarto is not working yet with PyMC. I may open up an issue there instead. Thank you for your help!

1 Like