Add graphviz support for SymbolicRVs

[larryshamalama]

Closes #5766.

Prior to this PR, symbolic distributions were labelled as "Deterministic" in the graph produced by Graphviz. The recent addition of symbolic RVs in #6072 allows pymc/model_graph.py to be adapted more easily. Here is an example of what is obtained using symbolic distributions such as pm.Mixture, pm.Censored, pm.ZeroInflatedPoisson.

with pm.Model() as model:
    psi = pm.Beta("psi", 1., 1.)
    mu_zip = pm.Gamma("mu_zip", 1., 1.)
    zip_var = pm.ZeroInflatedPoisson("zip", psi=psi, mu=mu_zip)
    
    mu_cens = pm.Normal("mu_cens", 0., 5.)
    sigma_cens = pm.HalfCauchy("sigma_cens", 2.5)
    normal_dist = pm.Normal.dist(mu=mu_cens, sigma=sigma_cens)
    censored_normal = pm.Censored("censored_normal", normal_dist, lower=-2., upper=2., observed=[0, 0.5, 1])
    
    mu_grw = pm.Normal("mu_grw", 0., 5.)
    sigma_grw = pm.HalfCauchy("sigma_grw", 2.5)
    pm.GaussianRandomWalk("grw", init_dist=pm.Normal.dist(mu_grw, sigma_grw), steps=10)
    
pm.model_to_graphviz(zip_model)

Tests in pymc/tests/test_model_graph.py checks that the arrows are pointing to and from the right variables and that the overall shape of the Graphviz Digraph is correct. Perhaps an additional check that the distribution names are correct can be added, because this is what this PR is about.

[codecov]

Codecov Report

Merging #6149 (20cfc98) into main (ce89404) will increase coverage by 0.76%.
The diff coverage is 31.42%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6149      +/-   ##
==========================================
+ Coverage   88.84%   89.60%   +0.76%     
==========================================
  Files         101      101              
  Lines       22235    22264      +29     
==========================================
+ Hits        19755    19950     +195     
+ Misses       2480     2314     -166     

Impacted Files	Coverage Δ
pymc/model_graph.py	59.50% <31.42%> (-19.45%)	⬇️
pymc/tests/distributions/test_timeseries.py	0.00% <0.00%> (-99.51%)	⬇️
pymc/tests/step_methods/hmc/test_quadpotential.py	0.00% <0.00%> (-95.82%)	⬇️
pymc/distributions/timeseries.py	28.46% <0.00%> (-66.00%)	⬇️
pymc/step_methods/hmc/quadpotential.py	73.76% <0.00%> (-6.94%)	⬇️
pymc/distributions/logprob.py	91.09% <0.00%> (-6.17%)	⬇️
pymc/distributions/shape_utils.py	97.42% <0.00%> (-0.43%)	⬇️
pymc/smc/smc.py	97.37% <0.00%> (+0.74%)	⬆️
pymc/distributions/censored.py	100.00% <0.00%> (+7.50%)	⬆️
... and 9 more

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[larryshamalama]

Closed in favor of a new PR that will use a dispatching approach