Merge pull request #5320 from ricardoV94/joint_logpt

Refactor `Factor` properties

Author: michaelosthege

.github/workflows/pytest.yml

@@ -49,7 +49,6 @@ jobs:
             --ignore=pymc/tests/test_updates.py
             --ignore=pymc/tests/test_gp.py
             --ignore=pymc/tests/test_model.py
-            --ignore=pymc/tests/test_model_func.py
             --ignore=pymc/tests/test_ode.py
             --ignore=pymc/tests/test_posdef_sym.py
             --ignore=pymc/tests/test_quadpotential.py
@@ -82,7 +81,6 @@ jobs:
             pymc/tests/test_distributions_timeseries.py
             pymc/tests/test_gp.py
             pymc/tests/test_model.py
-            pymc/tests/test_model_func.py
             pymc/tests/test_model_graph.py
             pymc/tests/test_ode.py
             pymc/tests/test_posdef_sym.py
@@ -166,7 +164,6 @@ jobs:
             pymc/tests/test_ode.py
           - |
             pymc/tests/test_model.py
-            pymc/tests/test_model_func.py
             pymc/tests/test_modelcontext.py
             pymc/tests/test_model_graph.py
             pymc/tests/test_pickling.py

RELEASE-NOTES.md

@@ -39,9 +39,13 @@ All of the above apply to:
 - `pm.sample_prior_predictive`, `pm.sample_posterior_predictive` and `pm.sample_posterior_predictive_w` now return an `InferenceData` object by default, instead of a dictionary (see [#5073](https://github.com/pymc-devs/pymc/pull/5073)).
 - `pm.sample_prior_predictive` no longer returns transformed variable values by default. Pass them by name in `var_names` if you want to obtain these draws (see [4769](https://github.com/pymc-devs/pymc/pull/4769)).
 - `pm.sample(trace=...)` no longer accepts `MultiTrace` or `len(.) > 0` traces ([see 5019#](https://github.com/pymc-devs/pymc/pull/5019)).
+- `logpt`, `logpt_sum`, `logp_elemwiset` and `nojac` variations were removed. Use `Model.logpt(jacobian=True/False, sum=True/False)` instead.
+- `dlogp_nojact` and `d2logp_nojact` were removed. Use `Model.dlogpt` and `d2logpt` with `jacobian=False` instead.
+- `logp`, `dlogp`, and `d2logp` and `nojac` variations were removed. Use `Model.compile_logp`, `compile_dlgop` and `compile_d2logp` with `jacobian` keyword instead.
+- `model.makefn` is now called `Model.compile_fn`, and `model.fn` was removed.
+- Methods starting with `fast_*`, such as `Model.fast_logp`, were removed. Same applies to `PointFunc` classes
 - The GLM submodule was removed, please use [Bambi](https://bambinos.github.io/bambi/) instead.
 - `pm.Bound` interface no longer accepts a callable class as argument, instead it requires an instantiated distribution (created via the `.dist()` API) to be passed as an argument. In addition, Bound no longer returns a class instance but works as a normal PyMC distribution. Finally, it is no longer possible to do predictive random sampling from Bounded variables. Please, consult the new documentation for details on how to use Bounded variables (see [4815](https://github.com/pymc-devs/pymc/pull/4815)).
-- `pm.logpt(transformed=...)` kwarg was removed (816b5f).
 - `Model(model=...)` kwarg was removed
 - `Model(theano_config=...)` kwarg was removed
 - `Model.size` property was removed (use `Model.ndim` instead).

pymc/backends/arviz.py

@@ -246,12 +246,26 @@ def _extract_log_likelihood(self, trace):
         # TODO: We no longer need one function per observed variable
         if self.log_likelihood is True:
             cached = [
-                (var, self.model.fn(self.model.logp_elemwiset(var)[0]))
+                (
+                    var,
+                    self.model.compile_fn(
+                        self.model.logpt(var, sum=False)[0],
+                        inputs=self.model.value_vars,
+                        on_unused_input="ignore",
+                    ),
+                )
                 for var in self.model.observed_RVs
             ]
         else:
             cached = [
-                (var, self.model.fn(self.model.logp_elemwiset(var)[0]))
+                (
+                    var,
+                    self.model.compile_fn(
+                        self.model.logpt(var, sum=False)[0],
+                        inputs=self.model.value_vars,
+                        on_unused_input="ignore",
+                    ),
+                )
                 for var in self.model.observed_RVs
                 if var.name in self.log_likelihood
             ]

pymc/backends/base.py

@@ -65,7 +65,7 @@ def __init__(self, name, model=None, vars=None, test_point=None):
 
         self.vars = vars
         self.varnames = [var.name for var in vars]
-        self.fn = model.fastfn(vars)
+        self.fn = model.compile_fn(vars, inputs=model.value_vars, on_unused_input="ignore")
 
         # Get variable shapes. Most backends will need this
         # information.

pymc/distributions/__init__.py

@@ -16,8 +16,7 @@
     logcdf,
     logp,
     logp_transform,
-    logpt,
-    logpt_sum,
+    joint_logpt,
 )
 
 from pymc.distributions.bound import Bound
@@ -191,9 +190,8 @@
     "Censored",
     "CAR",
     "PolyaGamma",
-    "logpt",
+    "joint_logpt",
     "logp",
     "logp_transform",
     "logcdf",
-    "logpt_sum",
 ]

pymc/distributions/logprob.py

@@ -11,7 +11,6 @@
 #   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 #   See the License for the specific language governing permissions and
 #   limitations under the License.
-import warnings
 
 from collections.abc import Mapping
 from functools import singledispatch
@@ -118,7 +117,7 @@ def _get_scaling(total_size, shape, ndim):
 )
 
 
-def logpt(
+def joint_logpt(
     var: Union[TensorVariable, List[TensorVariable]],
     rv_values: Optional[Union[TensorVariable, Dict[TensorVariable, TensorVariable]]] = None,
     *,
@@ -264,17 +263,3 @@ def logcdf(rv, value):
 
     value = at.as_tensor_variable(value, dtype=rv.dtype)
     return logcdf_aeppl(rv, value)
-
-
-def logpt_sum(*args, **kwargs):
-    """Return the sum of the logp values for the given observations.
-
-    Subclasses can use this to improve the speed of logp evaluations
-    if only the sum of the logp values is needed.
-    """
-    warnings.warn(
-        "logpt_sum has been deprecated, you can use logpt instead, which now defaults"
-        "to the same behavior of logpt_sum",
-        DeprecationWarning,
-    )
-    return logpt(*args, sum=True, **kwargs)