Error in examples/case_studies/probabilistic_matrix_factorization.ipynb

[wd60622]

Error while finding MAP for the model.

────────────────────────────────── Error running examples/case_studies/probabilistic_matrix_factorization.ipynb ──────────────────────────────────

---------------------------------------------------------------------------
Exception encountered at "In [20]":
---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
Cell In[20], line 2
      1 # Find MAP for PMF.
----> 2 pmf.find_map();

Cell In[13], line 6, in _find_map(self)
      4 with self.model:
      5     logging.info("finding PMF MAP using L-BFGS-B optimization...")
----> 6     self._map = pm.find_MAP(method="L-BFGS-B")
      8 elapsed = int(time.time() - tstart)
      9 logging.info("found PMF MAP in %d seconds" % elapsed)

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/tuning/starting.py:134, in find_MAP(start, vars, method, return_raw,
include_transformed, progressbar, progressbar_theme, maxeval, model, seed, *args, **kwargs)
    127 ipfn = make_initial_point_fn(
    128     model=model,
    129     jitter_rvs=set(),
    130     return_transformed=True,
    131     overrides=start,
    132 )
    133 start = ipfn(seed)
--> 134 model.check_start_vals(start)
    136 vars_dict = {var.name: var for var in vars}
    137 x0 = DictToArrayBijection.map(
    138     {var_name: value for var_name, value in start.items() if var_name in vars_dict}
    139 )

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/model/core.py:1742, in Model.check_start_vals(self, start, **kwargs)
   1736     valid_keys = ", ".join(value_names_set)
   1737     raise KeyError(
   1738         "Some start parameters do not appear in the model!\n"
   1739         f"Valid keys are: {valid_keys}, but {extra_keys} was supplied"
   1740     )
-> 1742 initial_eval = self.point_logps(point=elem, **kwargs)
   1744 if not all(np.isfinite(v) for v in initial_eval.values()):
   1745     raise SamplingError(
   1746         "Initial evaluation of model at starting point failed!\n"
   1747         f"Starting values:\n{elem}\n\n"
   1748         f"Logp initial evaluation results:\n{initial_eval}\n"
   1749         "You can call `model.debug()` for more details."
   1750     )

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/model/core.py:1774, in Model.point_logps(self, point, round_vals, **kwargs)
   1771     point = self.initial_point()
   1773 factors = self.basic_RVs + self.potentials
-> 1774 factor_logps_fn =
   1775 return {
   1776     factor.name: np.round(np.asarray(factor_logp), round_vals)
   1777     for factor, factor_logp in zip(
   (...)
   1780     )
   1781 }

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/model/core.py:672, in Model.logp(self, vars, jacobian, sum)
    670 rv_logps: list[TensorVariable] = []
    671 if rvs:
--> 672     rv_logps = transformed_conditional_logp(
    673         rvs=rvs,
    674         rvs_to_values=self.rvs_to_values,
    675         rvs_to_transforms=self.rvs_to_transforms,
    676         jacobian=jacobian,
    677     )
    678     assert isinstance(rv_logps, list)
    680 # Replace random variables by their value variables in potential terms

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/logprob/basic.py:595, in transformed_conditional_logp(rvs, rvs_to_values,
rvs_to_transforms, jacobian, **kwargs)
    592     transform_rewrite = TransformValuesRewrite(values_to_transforms)  # type: ignore
    594 kwargs.setdefault("warn_rvs", False)
--> 595 temp_logp_terms = conditional_logp(
    596     rvs_to_values,
    597     extra_rewrites=transform_rewrite,
    598     use_jacobian=jacobian,
    599     **kwargs,
    600 )
    602 # The function returns the logp for every single value term we provided to it.
    603 # This includes the extra values we plugged in above, so we filter those we
    604 # actually wanted in the same order they were given in.
    605 logp_terms = {}

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/logprob/basic.py:479, in conditional_logp(rv_values, warn_rvs, ir_rewriter,
extra_rewrites, **kwargs)
    417 r"""Create a map between variables and conditional logps such that the sum is their joint logp.
    418
    419 The `rv_values` dictionary specifies a joint probability graph defined by
   (...)
    475
    476 """
    477 warn_rvs, kwargs = _deprecate_warn_missing_rvs(warn_rvs, kwargs)
--> 479 fgraph = construct_ir_fgraph(rv_values, ir_rewriter=ir_rewriter)
    481 if extra_rewrites is not None:
    482     extra_rewrites.rewrite(fgraph)

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/logprob/rewriting.py:249, in construct_ir_fgraph(rv_values, ir_rewriter)
    246 cloned_values = tuple(v.clone() for v in rv_values.values())
    247 ir_rv_values = dict(zip(fgraph.outputs, cloned_values))
--> 249 replacements = tuple((rv, valued_rv(rv, value)) for rv, value in ir_rv_values.items())
    250 toposort_replace(fgraph, replacements, reverse=True)
    252 if ir_rewriter is None:

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/logprob/rewriting.py:249, in <genexpr>(.0)
    246 cloned_values = tuple(v.clone() for v in rv_values.values())
    247 ir_rv_values = dict(zip(fgraph.outputs, cloned_values))
--> 249 replacements = tuple((rv, valued_rv(rv, value)) for rv, value in ir_rv_values.items())
    250 toposort_replace(fgraph, replacements, reverse=True)
    252 if ir_rewriter is None:

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pytensor/graph/op.py:304, in Op.__call__(self, name, return_list, *inputs,
**kwargs)
    301             n.name = f"{name}_{i}"
    303 if config.compute_test_value != "off":
--> 304     compute_test_value(node)
    306 if self.default_output is not None:
    307     rval = node.outputs

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pytensor/graph/op.py:125, in compute_test_value(node)
    122 thunk.inputs = [storage_map for v in node.inputs]
    123 thunk.outputs = [storage_map for v in node.outputs]
--> 125 thunk()
    127 for output in node.outputs:
    128     # Check that the output has been computed
    129     assert compute_map[0], (output, storage_map[0])

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pytensor/graph/op.py:524, in Op.make_py_thunk.<locals>.rval(p, i, o, n)
    522 @is_thunk_type
    523 def rval(p=p, i=node_input_storage, o=node_output_storage, n=node):
--> 524     r = p(n, [x[0] for x in i], o)
    525     for o in node.outputs:
    526         compute_map[0] = True

File ~/micromamba/envs/pymc-examples/lib/python3.11/site-packages/pymc/logprob/abstract.py:245, in ValuedRV.perform(self, node, inputs, out)
    244 def perform(self, node, inputs, out):
--> 245     raise NotImplementedError("ValuedVar should not be present in the final graph!")

NotImplementedError: ValuedVar should not be present in the final graph!