Bump minimum PyMC dependency

Do not use custom `SolveTriangular` `Op` in `kalman_filter.py`

Author: ricardoV94

.github/workflows/pypi.yml

@@ -17,7 +17,7 @@ jobs:
           fetch-depth: 0
       - uses: actions/setup-python@v4
         with:
-          python-version: 3.8
+          python-version: "3.10"
       - name: Build the sdist and the wheel
         run: |
           pip install build
@@ -63,7 +63,7 @@ jobs:
           repository_url: https://test.pypi.org/legacy/
       - uses: actions/setup-python@v4
         with:
-          python-version: 3.8
+          python-version: "3.10"
       - name: Test pip install from test.pypi
         run: |
           python -m venv venv-test-pypi

.github/workflows/test.yml

@@ -19,7 +19,7 @@ jobs:
       matrix:
         os: [ubuntu-latest]
         floatx: [float64]
-        python-version: ["3.8"]
+        python-version: ["3.9"]
         test-subset:
           - pymc_experimental/tests
       fail-fast: false
@@ -84,7 +84,7 @@ jobs:
       matrix:
         os: [windows-latest]
         floatx: [float32]
-        python-version: ["3.10"]
+        python-version: ["3.11"]
         test-subset:
           - pymc_experimental/tests
       fail-fast: false

conda-envs/environment-test.yml

@@ -11,6 +11,6 @@ dependencies:
 - xhistogram
 - statsmodels
 - pip:
-  - pymc>=5.6.0  # CI was failing to resolve
+  - pymc>=5.8.1  # CI was failing to resolve
   - blackjax
   - scikit-learn

conda-envs/windows-environment-test.yml

@@ -10,5 +10,5 @@ dependencies:
 - xhistogram
 - statsmodels
 - pip:
-  - pymc>=5.6.0  # CI was failing to resolve
+  - pymc>=5.8.1  # CI was failing to resolve
   - scikit-learn

pymc_experimental/distributions/continuous.py

@@ -28,7 +28,7 @@
 from pymc.distributions.shape_utils import rv_size_is_none
 from pymc.pytensorf import floatX
 from pytensor.tensor.random.op import RandomVariable
-from pytensor.tensor.var import TensorVariable
+from pytensor.tensor.variable import TensorVariable
 from scipy import stats
 
 

pymc_experimental/distributions/histogram_utils.py

@@ -100,15 +100,15 @@ def histogram_approximation(name, dist, *, observed, **h_kwargs):
     ----------
     name : str
         Name for the Potential
-    dist : pytensor.tensor.var.TensorVariable
+    dist : TensorVariable
         The output of pm.Distribution.dist()
     observed : ArrayLike
         Observed value to construct a histogram. Histogram is computed over 0th axis.
         Dask is supported.
 
     Returns
     -------
-    pytensor.tensor.var.TensorVariable
+    TensorVariable
         Potential
 
     Examples

pymc_experimental/gp/latent_approx.py

@@ -11,12 +11,16 @@
 #   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.
-
+from functools import partial
 
 import numpy as np
 import pymc as pm
 import pytensor.tensor as pt
-from pymc.gp.util import JITTER_DEFAULT, cholesky, solve_lower, solve_upper, stabilize
+from pymc.gp.util import JITTER_DEFAULT, stabilize
+from pytensor.tensor.linalg import cholesky, solve_triangular
+
+solve_lower = partial(solve_triangular, lower=True)
+solve_upper = partial(solve_triangular, lower=False)
 
 
 class LatentApprox(pm.gp.Latent):

pymc_experimental/statespace/filters/kalman_filter.py

@@ -9,7 +9,7 @@
 from pytensor.raise_op import Assert
 from pytensor.tensor import TensorVariable
 from pytensor.tensor.nlinalg import matrix_dot
-from pytensor.tensor.slinalg import SolveTriangular
+from pytensor.tensor.slinalg import solve_triangular
 
 from pymc_experimental.statespace.filters.utilities import (
     quad_form_sym,
@@ -22,7 +22,6 @@
 MVN_CONST = pt.log(2 * pt.constant(np.pi, dtype="float64"))
 PARAM_NAMES = ["c", "d", "T", "Z", "R", "H", "Q"]
 
-solve_lower_triangular = SolveTriangular(lower=True)
 assert_data_is_1d = Assert("UnivariateTimeSeries filter requires data be at most 1-dimensional")
 assert_time_varying_dim_correct = Assert(
     "The first dimension of a time varying matrix (the time dimension) must be "
@@ -684,13 +683,13 @@ def update(self, a, P, y, c, d, Z, H, all_nan_flag):
         F_chol = pt.linalg.cholesky(F)
 
         # If everything is missing, K = 0, IKZ = I
-        K = solve_lower_triangular(F_chol.T, solve_lower_triangular(F_chol, PZT.T)).T
+        K = solve_triangular(F_chol.T, solve_triangular(F_chol, PZT.T)).T
         I_KZ = self.eye_states - K.dot(Z)
 
         a_filtered = a + K.dot(v)
         P_filtered = quad_form_sym(I_KZ, P) + quad_form_sym(K, H)
 
-        inner_term = solve_lower_triangular(F_chol.T, solve_lower_triangular(F_chol, v))
+        inner_term = solve_triangular(F_chol.T, solve_triangular(F_chol, v))
         n = y.shape[0]
 
         ll = pt.switch(

requirements.txt

@@ -1,2 +1,2 @@
-pymc>=5.6.0
+pymc>=5.8.1
 scikit-learn