feat: ported aesara to pytensor

.github/workflows/bump_n_publish.yml

@@ -0,0 +1,57 @@
+# This workflow will bump the version of the package and publish it to PyPI
+
+name: Bump and publish
+
+on: 
+  workflow_dispatch:
+  # push:
+  #   branches:
+  #     - master
+
+jobs:
+  bump_version:
+    if: ${{ !startsWith(github.event.head_commit.message, 'bump:') }}
+    strategy:
+      matrix:
+        python-version: ["3.11"]
+        poetry-version: ["1.5.1"]
+    runs-on: "ubuntu-latest"
+    name: "Bump version and create changelog with commitizen"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          token: "${{ secrets.GITHUB_TOKEN }}"
+      - name: Create bump and changelog
+        uses: commitizen-tools/commitizen-action@master
+        with:
+          github_token: ${{ secrets.GITHUB_TOKEN }}
+          changelog_increment_filename: body.md
+      - name: Print Version
+        run: echo "Bumped to version ${{ steps.cz.outputs.version }}"
+      - name: Release
+        uses: softprops/action-gh-release@v1
+        with:
+          body_path: "body.md"
+          tag_name: v${{ env.REVISION }}
+          generate_release_notes: true
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+      # deploy step
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Python Poetry Action
+        uses: abatilo/actions-poetry@v2
+        with:
+          poetry-version: ${{ matrix.poetry-version }}
+      - name: Publish to PyPI
+        run: |
+          poetry config pypi-token.pypi "${{ secrets.PYPI_API_KEY }}"
+          poetry publish --build
+
+
+

.github/workflows/tests.yml

@@ -46,12 +46,12 @@ jobs:
       - name: Test pycmtensor import
         run: |
           python -c "import pycmtensor as cmt; print(cmt.__version__)"
-          cat ~/.aesararc
+          cat ~/.pytensorrc
 
-      - name: Show Aesara config
+      - name: Show pytensor config
         run: |
-          python -c "import aesara; print(aesara.config.blas__ldflags)"
-          python -c "import aesara; print(aesara.config.gcc__cxxflags)"
+          python -c "import pytensor; print(pytensor.config.blas__ldflags)"
+          python -c "import pytensor; print(pytensor.config.gcc__cxxflags)"
 
       - name: Main test
         run: pytest

docs/getting_started/index.md

@@ -13,13 +13,13 @@ While these advanced Python deep learning libraries can be used to evaluate choi
 
 ### What can PyCMTensor do?
 
-PyCMTensor can be used for specifying, estimating, and conducting statistical tests on hybrid discrete choice models. It uses optimized tensor operations through [Aesara](https://aesara.readthedocs.io/en/latest/). 
+PyCMTensor can be used for specifying, estimating, and conducting statistical tests on hybrid discrete choice models. It uses optimized tensor operations through [pytensor](https://pytensor.readthedocs.io/en/latest/). 
 
 The library is built for estimating hybrid neural networks, Logit models, and Mixed Logit models. The models in PyCMTensor are built on computational graphs and are estimated using backpropagation algorithms.
 
 ## Project goals
 
-PyCMTensor aims to combine the interpretability of choice modelling syntaxes and expressions with the computational efficiency of Aesara tensor libraries. It specializes in defining 'hybrid' utility expressions and simplifies the process of integrating a deep neural network into choice models such as TasteNet or ResLogit. Unlike other deep learning libraries, PyCMTensor focuses on econometric modelling and statistical testing over purely predictive or classification models.
+PyCMTensor aims to combine the interpretability of choice modelling syntaxes and expressions with the computational efficiency of pytensor tensor libraries. It specializes in defining 'hybrid' utility expressions and simplifies the process of integrating a deep neural network into choice models such as TasteNet or ResLogit. Unlike other deep learning libraries, PyCMTensor focuses on econometric modelling and statistical testing over purely predictive or classification models.
 
 The objectives of this project include:
 

docs/getting_started/quick-start.md

@@ -124,7 +124,7 @@ The `Beta(name=, value=, lb=, ub=, status=)` object takes the following argument
     ```
 
 !!! info
-    `pycmtensor.expressions.Beta` follows the same syntax as in [Biogeme]() `biogeme.expressions.Beta` for familiarity sake. However, `pycmtensor.expressions.Beta` uses `aesara.tensor` variables to define the mathematical ops. Currently they are not interchangable.
+    `pycmtensor.expressions.Beta` follows the same syntax as in [Biogeme]() `biogeme.expressions.Beta` for familiarity sake. However, `pycmtensor.expressions.Beta` uses `pytensor.tensor` variables to define the mathematical ops. Currently they are not interchangable.
 
 
 ## Specifying utility equations

docs/user_guide/configuration.md

@@ -150,14 +150,14 @@ pycmtensor.config.add('seed', 100)
 
 	Default: `True`
 
-## Aesara config
+## pytensor config
 
-PyCMTensor uses the `aesara` library, which has its own set of configurations. We use the following by default:
+PyCMTensor uses the `pytensor` library, which has its own set of configurations. We use the following by default:
 
-`aesara.config.on_unused_input = "ignore"`
+`pytensor.config.on_unused_input = "ignore"`
 
-`aesara.config.mode = "Mode"`
+`pytensor.config.mode = "Mode"`
 
-`aesara.config.allow_gc = False`
+`pytensor.config.allow_gc = False`
 
-Refer to https://aesara.readthedocs.io/en/latest/config.html for other options. 
+Refer to https://pytensor.readthedocs.io/en/latest/config.html for other options. 

docs/user_guide/index.md

@@ -10,9 +10,9 @@ The entire model is build as a computational graph with nodes representing inter
 The graph is then manipulated internally and optimized during compilation.
 When an input is provided to certain tensors, the graph computes the values for other symbols in the graph. 
 
-The core API library, Aesara, gives us a set of tools that takes the defined symbolic tensors, constructs a computational graph and optimizes the mathematical operations around it.
+The core API library, pytensor, gives us a set of tools that takes the defined symbolic tensors, constructs a computational graph and optimizes the mathematical operations around it.
 
-As a developer, you define or extend a new model by creating new tensor variable objects from primitive tensors. For instance, take a basic linear equation, $y = bx +c$. The variables $y$, $x$, and $c$ are symbolic tensors making up a computational graph, and $b$ is a shared variable consisting of a mutable defined value. When we call `train(y)`, an Aesara function is executed on the equation and an update value is passed to $b$.
+As a developer, you define or extend a new model by creating new tensor variable objects from primitive tensors. For instance, take a basic linear equation, $y = bx +c$. The variables $y$, $x$, and $c$ are symbolic tensors making up a computational graph, and $b$ is a shared variable consisting of a mutable defined value. When we call `train(y)`, an pytensor function is executed on the equation and an update value is passed to $b$.
 
 A PyCMTensor model is composed of 
 
@@ -28,7 +28,7 @@ This symbolic approach does add complexity, however, the tradeoffs are that a. m
 
 ### `Beta`
 
-A `Beta` variable is the main data object when estimate a PyCMTensor model. `Beta` inherits the `TensorExpression` class which adds Aesara tensor operations to the data object. A `Beta` is defined as such:
+A `Beta` variable is the main data object when estimate a PyCMTensor model. `Beta` inherits the `TensorExpression` class which adds pytensor tensor operations to the data object. A `Beta` is defined as such:
 
 ```python
 Beta(name, value, lb, ub, status)
@@ -40,7 +40,7 @@ In mathematical operations, for instance:
 
 ```python
 from pycmtensor.expressions import Beta
-import aesara.tensor as aet
+import pytensor.tensor as aet
 import numpy as np 
 beta_cost = Beta("beta_cost", 2.)
 train_cost = aet.vector("train_cost") 

environment.yml

@@ -1,14 +1,13 @@
-name: pycmtensor-dev
+name: pytensor-dev
 channels:
     - conda-forge
     - defaults
 dependencies:
     # core dependencies
     - python>=3.9,<3.13
-    - aesara=2.9.3
-    - numpy<1.26.0
-    - scipy<1.12.0
-    - pandas<2.1.0
+    - pytensor<2.27.0
+    - scipy<1.15.0
+    - pandas<2.3.0
     - seaborn<0.14.0
     - scikit-learn<1.6.0
 
@@ -17,10 +16,10 @@ dependencies:
 
 
 # remove any old environment
-# conda env remove -n pycmtensor-dev
+# conda env remove -n pytensor-dev -y
 
-# install the environment
-# conda env create -f environment.yml && conda activate pycmtensor-dev
+# navigate to this directory and install the environment
+# conda env create -n pytensor-dev && conda activate pytensor-dev
 
 # development
 # conda install git ipykernel poetry