Make PyMC depend on logprob

Author: pymc-devs

.github/workflows/tests.yml

@@ -85,6 +85,19 @@ jobs:
             pymc/tests/ode/test_utils.py
             pymc/tests/step_methods/hmc/test_quadpotential.py
 
+          - |
+            pymc/tests/logprob/test_abstract.py
+            pymc/tests/logprob/test_censoring.py
+            pymc/tests/logprob/test_composite_logprob.py
+            pymc/tests/logprob/test_cumsum.py
+            pymc/tests/logprob/test_joint_logprob.py
+            pymc/tests/logprob/test_mixture.py
+            pymc/tests/logprob/test_rewriting.py
+            pymc/tests/logprob/test_scan.py
+            pymc/tests/logprob/test_tensor.py
+            pymc/tests/logprob/test_transforms.py
+            pymc/tests/logprob/test_utils.py
+
       fail-fast: false
     runs-on: ${{ matrix.os }}
     env:

.pre-commit-config.yaml

@@ -85,7 +85,6 @@ repos:
           (?x)(arviz-devs.github.io|
                python.arviz.org|
                aesara.readthedocs.io|
-               aeppl.readthedocs.io|
                pymc-experimental.readthedocs.io|
                docs.pymc.io|
                www.pymc.io|

conda-envs/environment-dev.yml

@@ -5,7 +5,6 @@ channels:
 - defaults
 dependencies:
 # Base dependencies
-- aeppl=0.0.38
 - aesara=2.8.8
 - arviz>=0.13.0
 - blas
@@ -41,3 +40,4 @@ dependencies:
 - types-cachetools
 - pip:
   - git+https://github.com/pymc-devs/pymc-sphinx-theme
+  - numdifftools>=0.9.40

conda-envs/environment-test.yml

@@ -5,7 +5,6 @@ channels:
 - defaults
 dependencies:
 # Base dependencies
-- aeppl=0.0.38
 - aesara=2.8.8
 - arviz>=0.13.0
 - blas
@@ -29,3 +28,5 @@ dependencies:
 - pytest>=3.0
 - mypy=0.990
 - types-cachetools
+- pip:
+  - numdifftools>=0.9.40

conda-envs/windows-environment-dev.yml

@@ -5,7 +5,6 @@ channels:
 - defaults
 dependencies:
 # Base dependencies (see install guide for Windows)
-- aeppl=0.0.38
 - aesara=2.8.8
 - arviz>=0.13.0
 - blas
@@ -38,3 +37,4 @@ dependencies:
 - types-cachetools
 - pip:
   - git+https://github.com/pymc-devs/pymc-sphinx-theme
+  - numdifftools>=0.9.40

conda-envs/windows-environment-test.yml

@@ -5,7 +5,6 @@ channels:
 - defaults
 dependencies:
 # Base dependencies (see install guide for Windows)
-- aeppl=0.0.38
 - aesara=2.8.8
 - arviz>=0.13.0
 - blas
@@ -30,3 +29,5 @@ dependencies:
 - pytest>=3.0
 - mypy=0.990
 - types-cachetools
+- pip:
+  - numdifftools>=0.9.40

docs/source/conf.py

@@ -188,7 +188,6 @@
 intersphinx_mapping = {
     "arviz": ("https://python.arviz.org/en/latest/", None),
     "aesara": ("https://aesara.readthedocs.io/en/latest/", None),
-    "aeppl": ("https://aeppl.readthedocs.io/en/latest/", None),
     "home": ("https://www.pymc.io", None),
     "pmx": ("https://www.pymc.io/projects/experimental/en/latest", None),
     "numpy": ("https://numpy.org/doc/stable/", None),

docs/source/learn/core_notebooks/GLM_linear.ipynb

@@ -522,14 +522,15 @@
    "source": [
     "%load_ext watermark\n",
     "\n",
-    "%watermark -n -u -v -iv -w -p aesara,aeppl"
+    "%watermark -n -u -v -iv -w -p aesara"
    ]
   }
  ],
  "metadata": {
   "anaconda-cloud": {},
+  "hide_input": false,
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "Python 3",
    "language": "python",
    "name": "python3"
   },
@@ -543,14 +544,27 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.7"
+   "version": "3.8.10"
   },
   "latex_envs": {
    "bibliofile": "biblio.bib",
    "cite_by": "apalike",
    "current_citInitial": 1,
    "eqLabelWithNumbers": true,
    "eqNumInitial": 0
+  },
+  "toc": {
+   "base_numbering": 1,
+   "nav_menu": {},
+   "number_sections": true,
+   "sideBar": true,
+   "skip_h1_title": false,
+   "title_cell": "Table of Contents",
+   "title_sidebar": "Contents",
+   "toc_cell": false,
+   "toc_position": {},
+   "toc_section_display": true,
+   "toc_window_display": false
   }
  },
  "nbformat": 4,

docs/source/learn/core_notebooks/model_comparison.ipynb

@@ -536,16 +536,17 @@
     }
    ],
    "source": [
-    "%watermark -n -u -v -iv -w -p xarray,aesara,aeppl"
+    "%watermark -n -u -v -iv -w -p xarray,aesara"
    ]
   }
  ],
  "metadata": {
+  "hide_input": false,
   "interpreter": {
    "hash": "baf205d70af30bf8b721a304f5a44beb31bf8af014f6b7340f1a7ae004926653"
   },
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "Python 3",
    "language": "python",
    "name": "python3"
   },
@@ -559,7 +560,20 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.7"
+   "version": "3.8.10"
+  },
+  "toc": {
+   "base_numbering": 1,
+   "nav_menu": {},
+   "number_sections": true,
+   "sideBar": true,
+   "skip_h1_title": false,
+   "title_cell": "Table of Contents",
+   "title_sidebar": "Contents",
+   "toc_cell": false,
+   "toc_position": {},
+   "toc_section_display": true,
+   "toc_window_display": false
   }
  },
  "nbformat": 4,

docs/source/learn/core_notebooks/posterior_predictive.ipynb

@@ -4649,14 +4649,15 @@
    ],
    "source": [
     "%load_ext watermark\n",
-    "%watermark -n -u -v -iv -w -p aesara,aeppl"
+    "%watermark -n -u -v -iv -w -p aesara"
    ]
   }
  ],
  "metadata": {
   "anaconda-cloud": {},
+  "hide_input": false,
   "kernelspec": {
-   "display_name": "Python 3.9.13 ('pymc-dev-py39')",
+   "display_name": "Python 3",
    "language": "python",
    "name": "python3"
   },
@@ -4670,7 +4671,20 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.13"
+   "version": "3.8.10"
+  },
+  "toc": {
+   "base_numbering": 1,
+   "nav_menu": {},
+   "number_sections": true,
+   "sideBar": true,
+   "skip_h1_title": false,
+   "title_cell": "Table of Contents",
+   "title_sidebar": "Contents",
+   "toc_cell": false,
+   "toc_position": {},
+   "toc_section_display": true,
+   "toc_window_display": false
   },
   "vscode": {
    "interpreter": {

docs/source/learn/core_notebooks/pymc_aesara.ipynb

@@ -4340,14 +4340,15 @@
    ],
    "source": [
     "%load_ext watermark\n",
-    "%watermark -n -u -v -iv -w -p xarray,aeppl"
+    "%watermark -n -u -v -iv -w -p xarray"
    ]
   }
  ],
  "metadata": {
   "anaconda-cloud": {},
+  "hide_input": false,
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "Python 3",
    "language": "python",
    "name": "python3"
   },
@@ -4361,7 +4362,20 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.5"
+   "version": "3.8.10"
+  },
+  "toc": {
+   "base_numbering": 1,
+   "nav_menu": {},
+   "number_sections": true,
+   "sideBar": true,
+   "skip_h1_title": false,
+   "title_cell": "Table of Contents",
+   "title_sidebar": "Contents",
+   "toc_cell": false,
+   "toc_position": {},
+   "toc_section_display": true,
+   "toc_window_display": false
   }
  },
  "nbformat": 4,

docs/source/learn/core_notebooks/pymc_overview.ipynb

@@ -54,6 +54,7 @@ def __set_compiler_flags():
 from pymc.distributions import *
 from pymc.exceptions import *
 from pymc.func_utils import find_constrained_prior
+from pymc.logprob import *
 from pymc.math import (
     expand_packed_triangular,
     invlogit,

pymc/__init__.py

@@ -32,8 +32,6 @@
 import pandas as pd
 import scipy.sparse as sps
 
-from aeppl.logprob import CheckParameterValue
-from aeppl.transforms import RVTransform
 from aesara import scalar
 from aesara.compile import Function, Mode, get_mode
 from aesara.gradient import grad
@@ -65,6 +63,8 @@
 from aesara.tensor.var import TensorConstant, TensorVariable
 
 from pymc.exceptions import NotConstantValueError
+from pymc.logprob.transforms import RVTransform
+from pymc.logprob.utils import CheckParameterValue
 from pymc.vartypes import continuous_types, isgenerator, typefilter
 
 PotentialShapeType = Union[int, np.ndarray, Sequence[Union[int, Variable]], TensorVariable]
@@ -947,7 +947,7 @@ def largest_common_dtype(tensors):
 
 @node_rewriter(tracks=[CheckParameterValue])
 def local_remove_check_parameter(fgraph, node):
-    """Rewrite that removes Aeppl's CheckParameterValue
+    """Rewrite that removes CheckParameterValue
 
     This is used when compile_rv_inplace
     """
@@ -1071,13 +1071,13 @@ def compile_pymc(
         Ensures that compiled functions containing random variables will produce new
         samples on each call.
     local_check_parameter_to_ninf_switch
-        Replaces Aeppl's CheckParameterValue assertions is logp expressions with Switches
+        Replaces CheckParameterValue assertions is logp expressions with Switches
         that return -inf in place of the assert.
 
     Optional rewrites
     -----------------
     local_remove_check_parameter
-        Replaces Aeppl's CheckParameterValue assertions is logp expressions. This is used
+        Replaces CheckParameterValue assertions is logp expressions. This is used
         as an alteranative to the default local_check_parameter_to_ninf_switch whenenver
         this function is called within a model context and the model `check_bounds` flag
         is set to False.

pymc/aesaraf.py

@@ -15,7 +15,6 @@
 from pymc.distributions.logprob import (  # isort:skip
     logcdf,
     logp,
-    joint_logp,
 )
 
 from pymc.distributions.bound import Bound
@@ -198,7 +197,6 @@
     "Censored",
     "CAR",
     "PolyaGamma",
-    "joint_logp",
     "logp",
     "logcdf",
 ]

pymc/distributions/__init__.py

@@ -29,7 +29,7 @@
 class CensoredRV(SymbolicRandomVariable):
     """Censored random variable"""
 
-    inline_aeppl = True
+    inline_logprob = True
     _print_name = ("Censored", "\\operatorname{Censored}")
 
 

pymc/distributions/censored.py

@@ -28,7 +28,6 @@
 import aesara.tensor as at
 import numpy as np
 
-from aeppl.logprob import _logprob, logcdf, logprob
 from aesara.graph.basic import Apply, Variable
 from aesara.graph.op import Op
 from aesara.raise_op import Assert
@@ -57,6 +56,8 @@
 from aesara.tensor.random.op import RandomVariable
 from aesara.tensor.var import TensorConstant
 
+from pymc.logprob.abstract import _logprob, logcdf, logprob
+
 try:
     from polyagamma import polyagamma_cdf, polyagamma_pdf, random_polyagamma
 except ImportError:  # pragma: no cover
@@ -531,6 +532,9 @@ def logcdf(value, mu, sigma):
             msg="sigma > 0",
         )
 
+    def icdf(value, mu, sigma):
+        return mu + sigma * -np.sqrt(2.0) * at.erfcinv(2 * value)
+
 
 class TruncatedNormalRV(RandomVariable):
     name = "truncated_normal"
@@ -1290,10 +1294,6 @@ class Exponential(PositiveContinuous):
     Variance  :math:`\dfrac{1}{\lambda^2}`
     ========  ============================
 
-    Notes
-    -----
-    Logp calculation is defined in `aeppl.logprob <https://github.com/aesara-devs/aeppl/blob/main/aeppl/logprob.py/>`_.
-
     Parameters
     ----------
     lam : tensor_like of float

pymc/distributions/continuous.py

@@ -819,6 +819,9 @@ def logcdf(value, p):
             msg="0 <= p <= 1",
         )
 
+    def icdf(value, p):
+        return at.ceil(at.log1p(-value) / at.log1p(-p)).astype("int64")
+
 
 class HyperGeometric(Discrete):
     R"""

pymc/distributions/discrete.py

@@ -27,7 +27,6 @@
 import scipy.linalg
 import scipy.stats
 
-from aeppl.logprob import CheckParameterValue
 from aesara.compile.builders import OpFromGraph
 from aesara.graph.basic import Apply, Variable
 from aesara.graph.op import Op
@@ -38,6 +37,7 @@
 
 from pymc.aesaraf import floatX
 from pymc.distributions.shape_utils import to_tuple
+from pymc.logprob.utils import CheckParameterValue
 
 solve_lower = SolveTriangular(lower=True)
 solve_upper = SolveTriangular(lower=False)