Enable automatic versioning via setuptools_scm (#2111)

* Enable automatic versioning via setuptools_scm

Based on cornellius-gp/linear_operator#2

* Resolve RTD warnings from example notebooks

.github/workflows/deploy.yml

@@ -23,12 +23,13 @@ jobs:
     - name: Install dependencies
       run: |
         python -m pip install --upgrade pip
-        pip install setuptools wheel twine
+        pip install setuptools_scm setuptools wheel twine
     - name: Build and publish
       env:
         TWINE_USERNAME: ${{ secrets.PYPI_USERNAME }}
         TWINE_PASSWORD: ${{ secrets.PYPI_PASSWORD }}
       run: |
+        python -m setuptools_scm
         python setup.py sdist bdist_wheel
         twine upload dist/*
 
@@ -44,11 +45,13 @@ jobs:
     - name: Install dependencies
       run: |
         conda install -y anaconda-client conda-build
+        pip install setuptools_scm
     - name: Build and publish
       run: |
         conda config --set anaconda_upload yes
         conda config --append channels pytorch
         /usr/share/miniconda/bin/anaconda login --username ${{ secrets.CONDA_USERNAME }} --password ${{ secrets.CONDA_PASSWORD }}
+        python -m setuptools_scm
         cd .conda
         conda build .
         /usr/share/miniconda/bin/anaconda logout

.gitignore

@@ -153,3 +153,6 @@ Temporary Items
 # Data files that might be downloaded
 **/*.tar.gz
 **/*.txt
+
+# version file (auto-generated by setuptools_scm)
+gpytorch/version.py

.readthedocs.yml

@@ -5,6 +5,14 @@
 # Required
 version: 2
 
+build:
+  os: "ubuntu-22.04"
+  tools:
+    python: "3.8"
+  jobs:
+    pre_build:
+      - python -m setuptools_scm
+
 # Build documentation in the docs/ directory with Sphinx
 sphinx:
   configuration: docs/source/conf.py
@@ -16,7 +24,6 @@ sphinx:
 
 # Optionally set the version of Python and requirements required to build your docs
 python:
-  version: 3.8
   install:
     - requirements: requirements.txt
     - requirements: docs/requirements.txt

docs/requirements.txt

@@ -1,3 +1,4 @@
+setuptools_scm
 nbformat
 ipython
 ipykernel

docs/source/conf.py

@@ -28,11 +28,13 @@ def read(*names, **kwargs):
 
 
 def find_version(*file_paths):
-    version_file = read(*file_paths)
-    version_match = re.search(r"^__version__ = ['\"]([^'\"]*)['\"]", version_file, re.M)
-    if version_match:
+    try:
+        with io.open(os.path.join(os.path.dirname(__file__), *file_paths), encoding="utf8") as fp:
+            version_file = fp.read()
+        version_match = re.search(r"^__version__ = version = ['\"]([^'\"]*)['\"]", version_file, re.M)
         return version_match.group(1)
-    raise RuntimeError("Unable to find version string.")
+    except Exception as e:
+        raise RuntimeError("Unable to find version string:\n", e)
 
 
 sys.path.append(os.path.abspath(os.path.join(__file__, "..", "..", "..")))
@@ -63,7 +65,7 @@ def find_version(*file_paths):
 author = "Cornellius GP"
 
 # The short X.Y version
-version = find_version("gpytorch", "__init__.py")
+version = find_version("..", "..", "gpytorch", "version.py")
 # The full version, including alpha/beta/rc tags
 release = version
 

examples/00_Basic_Usage/Metrics.ipynb

@@ -12,10 +12,12 @@
     "\n",
     "We'll be modeling the function\n",
     "\n",
+    "$$\n",
     "\\begin{align}\n",
-    "y &= \\sin(2\\pi x) + \\epsilon \\\\\n",
+    "  y &= \\sin(2\\pi x) + \\epsilon \\\\\n",
     "  \\epsilon &\\sim \\mathcal{N}(0, 0.04) \n",
-    "\\end{align}"
+    "\\end{align}\n",
+    "$$\n"
    ]
   },
   {
@@ -441,7 +443,7 @@
    "hash": "d4d1e4263499bec80672ea0156c357c1ee493ec2b1c70f0acce89fc37c4a6abe"
   },
   "kernelspec": {
-   "display_name": "Python 3.8.12 ('base')",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -455,9 +457,8 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.12"
-  },
-  "orig_nbformat": 4
+   "version": "3.8.0"
+  }
  },
  "nbformat": 4,
  "nbformat_minor": 2

examples/01_Exact_GPs/GP_Regression_DistributionalKernel.ipynb

@@ -10,10 +10,12 @@
     "\n",
     "In this notebook, we're going to demonstrate one way of dealing with uncertainty in our training data. Let's say that we're collecting training data that models the following function.\n",
     "\n",
+    "$$\n",
     "\\begin{align}\n",
     "y &= \\sin(2\\pi x) + \\epsilon \\\\\n",
     "  \\epsilon &\\sim \\mathcal{N}(0, 0.2) \n",
     "\\end{align}\n",
+    "$$\n",
     "\n",
     "However, now assume that we're a bit uncertain about our features. In particular, we're going to assume that every `x_i` value is not a point but a distribution instead. E.g.\n",
     "\n",
@@ -780,9 +782,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3.7.4 64-bit ('base': conda)",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
-   "name": "python37464bitbaseconda52eab690427c4f7ea56588deee120c46"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {
@@ -794,7 +796,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.7.4"
+   "version": "3.8.0"
   }
  },
  "nbformat": 4,

examples/01_Exact_GPs/Simple_GP_Regression.ipynb

@@ -10,10 +10,12 @@
     "\n",
     "In this notebook, we demonstrate many of the design features of GPyTorch using the simplest example, training an RBF kernel Gaussian process on a simple function. We'll be modeling the function\n",
     "\n",
+    "$$\n",
     "\\begin{align}\n",
     "y &= \\sin(2\\pi x) + \\epsilon \\\\\n",
     "  \\epsilon &\\sim \\mathcal{N}(0, 0.04) \n",
     "\\end{align}\n",
+    "$$\n",
     "\n",
     "with 100 training examples, and testing on 51 test examples.\n",
     "\n",
@@ -361,7 +363,7 @@
  "metadata": {
   "anaconda-cloud": {},
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -375,7 +377,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.7.0"
+   "version": "3.8.0"
   }
  },
  "nbformat": 4,

examples/02_Scalable_Exact_GPs/Simple_GP_Regression_CUDA.ipynb

@@ -18,10 +18,12 @@
     "\n",
     "In this notebook, we demonstrate many of the design features of GPyTorch using the simplest example, training an RBF kernel Gaussian process on a simple function. We'll be modeling the function\n",
     "\n",
+    "$$\n",
     "\\begin{align}\n",
     "  y &= \\sin(2\\pi x) + \\epsilon \\\\ \n",
     "  \\epsilon &\\sim \\mathcal{N}(0, 0.2) \n",
     "\\end{align}\n",
+    "$$\n",
     "\n",
     "with 11 training examples, and testing on 51 test examples."
    ]
@@ -326,7 +328,7 @@
  "metadata": {
   "anaconda-cloud": {},
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -340,7 +342,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.7.3"
+   "version": "3.8.0"
   }
  },
  "nbformat": 4,

examples/04_Variational_and_Approximate_GPs/Approximate_GP_Objective_Functions.ipynb

@@ -214,10 +214,12 @@
     "The **predictive log likelihood** is an alternative to the variational ELBO that was proposed in [Jankowiak et al., 2020](https://arxiv.org/abs/1910.07123).\n",
     "It typically produces predictive variances than the `gpytorch.mlls.VariationalELBO` objective.\n",
     "\n",
+    "$$\n",
     "\\begin{align}\n",
     "\\mathcal{L}_\\text{PLL} &= \\mathbb{E}_{p_\\text{data}( y, \\mathbf x )} \\left[ \\log p( y \\! \\mid \\! \\mathbf x)  \\right] - \\beta \\: \\text{KL} \\left[ q( \\mathbf u) \\Vert p( \\mathbf u) \\right] \\\\ \n",
     "  &\\approx \\sum_{i=1}^N \\log \\mathbb{E}_{q(\\mathbf u)} \\left[ \\int p( y_i \\! \\mid \\! f_i) p(f_i \\! \\mid \\! \\mathbf u) \\: d f_i \\right] - \\beta \\: \\text{KL} \\left[ q( \\mathbf u) \\Vert p( \\mathbf u) \\right] \n",
     "\\end{align}\n",
+    "$$\n",
     "\n",
     "Note that this objective is *very similar* to the variational ELBO.\n",
     "The only difference is that the $\\log$ occurs *outside* the expectation $\\mathbb E_{q(\\mathbf u)}$.\n",

examples/04_Variational_and_Approximate_GPs/GP_Regression_with_Uncertain_Inputs.ipynb

@@ -10,10 +10,12 @@
     "\n",
     "In this notebook, we're going to demonstrate one way of dealing with uncertainty in our training data. Let's say that we're collecting training data that models the following function.\n",
     "\n",
+    "$$\n",
     "\\begin{align}\n",
     "y &= \\sin(2\\pi x) + \\epsilon \\\\\n",
     "  \\epsilon &\\sim \\mathcal{N}(0, 0.2) \n",
     "\\end{align}\n",
+    "$$\n",
     "\n",
     "However, now assume that we're a bit uncertain about our features. In particular, we're going to assume that every `x_i` value is not a point but a distribution instead. E.g.\n",
     "\n",

examples/08_Advanced_Usage/Simple_GP_Regression_Derivative_Information_1d.ipynb

@@ -9,11 +9,13 @@
     "## Introduction\n",
     "In this notebook, we show how to train a GP regression model in GPyTorch of an unknown function given function value and derivative observations. We consider modeling the function:\n",
     "\n",
+    "$$\n",
     "\\begin{align}\n",
     "              y &= \\sin(2x) + cos(x) + \\epsilon \\\\\n",
     "  \\frac{dy}{dx} &= 2\\cos(2x) - \\sin(x) + \\epsilon \\\\  \n",
     "       \\epsilon &\\sim \\mathcal{N}(0, 0.5)\n",
     "\\end{align}\n",
+    "$$\n",
     "\n",
     "using 50 value and derivative observations."
    ]

gpytorch/__init__.py

@@ -266,7 +266,12 @@ def sqrt_inv_matmul(input: Anysor, rhs: Tensor, lhs: Optional[Tensor] = None) ->
     return linear_operator.sqrt_inv_matmul(input=input, rhs=rhs, lhs=lhs)
 
 
-__version__ = "1.8.1"
+# Read version number as written by setuptools_scm
+try:
+    from gpytorch.version import version as __version__
+except Exception:  # pragma: no cover
+    __version__ = "Unknown"  # pragma: no cover
+
 
 __all__ = [
     # Submodules

pyproject.toml

@@ -0,0 +1,7 @@
+[build-system]
+requires = ["setuptools", "setuptools-scm", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[tool.setuptools_scm]
+local_scheme = "node-and-date"
+write_to = "./gpytorch/version.py"

setup.cfg

@@ -1,6 +1,3 @@
-[bdist_wheel]
-universal=1
-
 [pep8]
 max-line-length = 120
 

setup.py

@@ -3,26 +3,38 @@
 import io
 import os
 import re
+import sys
 
 from setuptools import find_packages, setup
 
+REQUIRED_MAJOR = 3
+REQUIRED_MINOR = 8
 
-# Get version
-def read(*names, **kwargs):
-    with io.open(os.path.join(os.path.dirname(__file__), *names), encoding=kwargs.get("encoding", "utf8")) as fp:
-        return fp.read()
+# Check for python version
+if sys.version_info < (REQUIRED_MAJOR, REQUIRED_MINOR):
+    error = (
+        "Your version of python ({major}.{minor}) is too old. You need python >= {required_major}.{required_minor}."
+    ).format(
+        major=sys.version_info.major,
+        minor=sys.version_info.minor,
+        required_minor=REQUIRED_MINOR,
+        required_major=REQUIRED_MAJOR,
+    )
+    sys.exit(error)
 
 
+# Get version
 def find_version(*file_paths):
-    version_file = read(*file_paths)
-    version_match = re.search(r"^__version__ = ['\"]([^'\"]*)['\"]", version_file, re.M)
-    if version_match:
+    try:
+        with io.open(os.path.join(os.path.dirname(__file__), *file_paths), encoding="utf8") as fp:
+            version_file = fp.read()
+        version_match = re.search(r"^__version__ = version = ['\"]([^'\"]*)['\"]", version_file, re.M)
         return version_match.group(1)
-    raise RuntimeError("Unable to find version string.")
+    except Exception:
+        return None
 
 
 readme = open("README.md").read()
-version = find_version("gpytorch", "__init__.py")
 
 
 torch_min = "1.11"
@@ -43,7 +55,7 @@ def find_version(*file_paths):
 # Run the setup
 setup(
     name="gpytorch",
-    version=version,
+    version=find_version("gpytorch", "version.py"),
     description="An implementation of Gaussian Processes in Pytorch",
     long_description=readme,
     long_description_content_type="text/markdown",