BUG: model_to_graphviz fails with Flat variables #8024
Description
Describe the issue:
The call to pm.model_to_graphviz or model.to_graphviz fails if the model contains a pm.Flat variable. According to the error message, it tries to sample from the flat variable.
Reproduceable code example:
import numpy as np
import pandas as pd
import pymc as pm
RANDOM_SEED = 8927
rng = np.random.default_rng(RANDOM_SEED)
size = 200
true_intercept = 1
true_slope = 2
x = np.linspace(0, 1, size)
true_regression_line = true_intercept + true_slope * x
y = true_regression_line + rng.normal(scale=0.5, size=size)
data = pd.DataFrame({"x": x, "y": y})
with pm.Model() as model:
intercept = pm.Flat("intercept") # <--- Flat prior here causes model.to_graphviz() to fail
sigma = pm.HalfCauchy("sigma", beta=10)
slope = pm.Normal("slope", 0, sigma=20)
likelihood = pm.Normal("y", mu=intercept + slope * x, sigma=sigma, observed=y)
idata = pm.sample(3000)
pm.model_to_graphviz(model)Error message:
Traceback (most recent call last):
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/link/utils.py", line 197, in streamline_default_f
thunk()
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/graph/op.py", line 545, in rval
r = p(n, [x[0] for x in i], o)
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/tensor/random/op.py", line 430, in perform
self.rng_fn(rng, *args, None if size is None else tuple(size)),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/distributions/continuous.py", line 357, in rng_fn
raise NotImplementedError("Cannot sample from flat variable")
NotImplementedError: Cannot sample from flat variable
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/Users/rebane/dev/salk/turnout-model/model_test/tmp.py", line 25, in <module>
pm.model_to_graphviz(model)
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/model_graph.py", line 763, in model_to_graphviz
plates=graph.get_plates(var_names=var_names),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/model_graph.py", line 349, in get_plates
var_name: tuple(map(int, fast_eval(self.model[var_name].shape)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/model_graph.py", line 77, in fast_eval
return function([], var, mode=_cheap_eval_mode)()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/compile/function/types.py", line 1038, in __call__
outputs = vm() if output_subset is None else vm(output_subset=output_subset)
^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/link/utils.py", line 201, in streamline_default_f
raise_with_op(fgraph, node, thunk)
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/link/utils.py", line 526, in raise_with_op
raise exc_value.with_traceback(exc_trace)
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/link/utils.py", line 197, in streamline_default_f
thunk()
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/graph/op.py", line 545, in rval
r = p(n, [x[0] for x in i], o)
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pytensor/tensor/random/op.py", line 430, in perform
self.rng_fn(rng, *args, None if size is None else tuple(size)),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/distributions/continuous.py", line 357, in rng_fn
raise NotImplementedError("Cannot sample from flat variable")
NotImplementedError: Cannot sample from flat variable
Apply node that caused the error: flat_rv{"->()"}(RNG(<Generator(PCG64) at 0x13B388900>), NoneConst{None})
Toposort index: 0
Inputs types: [RandomGeneratorType, <pytensor.tensor.type_other.NoneTypeT object at 0x1105da480>]
Inputs shapes: ['No shapes', 'No shapes']
Inputs strides: ['No strides', 'No strides']
Inputs values: [Generator(PCG64) at 0x13B388900, None]
Outputs clients: [[], [Shape(intercept)]]
Backtrace when the node is created (use PyTensor flag traceback__limit=N to make it longer):
File "/Users/rebane/dev/salk/turnout-model/model_test/tmp.py", line 19, in <module>
intercept = pm.Flat("intercept") # <--- Flat prior here causes model.to_graphviz() to fail
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/distributions/distribution.py", line 529, in __new__
rv_out = cls.dist(*args, **kwargs)
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/distributions/continuous.py", line 370, in dist
res = super().dist([], **kwargs)
File "/Users/rebane/miniconda3/envs/test/lib/python3.12/site-packages/pymc/distributions/distribution.py", line 598, in dist
return cls.rv_op(*dist_params, size=create_size, **kwargs)
HINT: Use the PyTensor flag `exception_verbosity=high` for a debug print-out and storage map footprint of this Apply node.PyMC version information:
- PyMC: 5.27.0
- PyTensor: 2.36.1
- Python: 3.12.0
- Operating system: macOS Sequoia 15.6.1
- How did you install PyMC: pip
Context for the issue:
Visualizing the model fails. Does not block my work.