Fix CI (#140)

* Drop support for Pythons older than 3.8

* Remove 2020-era Windows patch from CI

* Remove baroque check_version step from CI

* CI: renovate, add pip cache

* CI: split linting to separate step

* CI: don't allow compiling libraries from scratch, use correct Torch repo

* CI: print pytest progress

* Drop Boltons dependency for small custom LRU dict

.github/workflows/run_tests.yml

@@ -9,78 +9,45 @@ on:
     - cron: "0 2 * * 6"
 
 jobs:
-  check_version:
+  build:
     strategy:
       matrix:
         python-version: [ 3.8 ]
-        os: [ ubuntu-latest ]
+        os: [ ubuntu-latest, macOS-latest, windows-latest ]
+      fail-fast: false
     runs-on: ${{ matrix.os }}
     steps:
       - name: Checkout code
-        uses: actions/checkout@v2
+        uses: actions/checkout@v4
 
       - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v2
+        uses: actions/setup-python@v4
         with:
           python-version: ${{ matrix.python-version }}
+          cache: pip
+          cache-dependency-path: setup.py
 
-      - name: Check version
-        if: (github.event_name == 'pull_request' && github.base_ref == 'master')
-        run: |
-          python -m pip install --upgrade pip
-
-          python -m pip install git+https://github.com/google-research/torchsde.git
-          master_info=$(pip list | grep torchsde)
-          master_version=$(echo ${master_info} | cut -d " " -f2)
-          python -m pip uninstall -y torchsde
-
-          python setup.py install
-          pr_info=$(pip list | grep torchsde)
-          pr_version=$(echo ${pr_info} | cut -d " " -f2)
+      - name: Install
+        run: pip install pytest -e . --only-binary=numpy,scipy,matplotlib,torch
+        env:
+          PIP_EXTRA_INDEX_URL: https://download.pytorch.org/whl/cpu
 
-          python -c "import itertools as it
-          import sys
-
-          _, master_version, pr_version = sys.argv
-
-          master_version_ = [int(i) for i in master_version.split('.')]
-          pr_version_ = [int(i) for i in pr_version.split('.')]
-
-          master_version__ = tuple(m for p, m in it.zip_longest(pr_version_, master_version_, fillvalue=0))
-          pr_version__ = tuple(p for p, m in it.zip_longest(pr_version_, master_version_, fillvalue=0))
-          sys.exit(pr_version__ < master_version__)" ${master_version} ${pr_version}
+      - name: Test with pytest
+        run: python -m pytest -v
 
-  build:
-    needs: [ check_version ]
-    strategy:
-      matrix:
-        python-version: [ 3.6, 3.8 ]
-        os: [ ubuntu-latest, macOS-latest, windows-latest ]
-      fail-fast: false
-    runs-on: ${{ matrix.os }}
+  lint:
+    runs-on: ubuntu-latest
     steps:
       - name: Checkout code
-        uses: actions/checkout@v2
+        uses: actions/checkout@v4
 
-      - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v2
+      - uses: actions/setup-python@v4
         with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          python -m pip install flake8 pytest
-
-      - name: Windows patch  # Specifically for windows, since pip fails to fetch torch 1.6.0 as of Oct 2020.
-        if: runner.os == 'Windows'
-        run: python -m pip install torch==1.6.0+cpu torchvision==0.7.0+cpu -f https://download.pytorch.org/whl/torch_stable.html
+          python-version: "3.11"
+          cache: pip
+          cache-dependency-path: setup.py
 
       - name: Lint with flake8
         run: |
+          python -m pip install flake8
           python -m flake8 .
-
-      - name: Test with pytest
-        run: |
-          python setup.py install
-          python -m pytest

README.md

@@ -11,7 +11,7 @@ This library provides [stochastic differential equation (SDE)](https://en.wikipe
 pip install torchsde
 ```
 
-**Requirements:** Python >=3.6 and PyTorch >=1.6.0.
+**Requirements:** Python >=3.8 and PyTorch >=1.6.0.
 
 ## Documentation
 Available [here](./DOCUMENTATION.md).

setup.py

@@ -40,15 +40,12 @@
     url="https://github.com/google-research/torchsde",
     packages=setuptools.find_packages(exclude=['benchmarks', 'diagnostics', 'examples', 'tests']),
     install_requires=[
-        "boltons>=20.2.1",
-        "numpy==1.19;python_version<'3.7'",
-        "numpy>=1.19;python_version>='3.7'",
-        "scipy==1.5;python_version<'3.7'",
-        "scipy>=1.5;python_version>='3.7'",
+        "numpy>=1.19",
+        "scipy>=1.5",
         "torch>=1.6.0",
         "trampoline>=0.1.2",
     ],
-    python_requires='~=3.6',
+    python_requires='>=3.8',
     classifiers=[
         "Programming Language :: Python :: 3",
         "License :: OSI Approved :: Apache Software License",

torchsde/_brownian/brownian_interval.py

@@ -16,7 +16,6 @@
 import trampoline
 import warnings
 
-import boltons.cacheutils
 import numpy as np
 import torch
 
@@ -112,6 +111,21 @@ def __getitem__(self, item):
         raise KeyError
 
 
+class _LRUDict(dict):
+    def __init__(self, max_size):
+        super().__init__()
+        self._max_size = max_size
+        self._keys = []
+
+    def __setitem__(self, key, value):
+        if key in self:
+            self._keys.remove(key)
+        elif len(self) >= self._max_size:
+            del self[self._keys.pop(0)]
+        super().__setitem__(key, value)
+        self._keys.append(key)
+
+
 class _Interval:
     # Intervals correspond to some subinterval of the overall interval [t0, t1].
     # They are arranged as a binary tree: each node corresponds to an interval. If a node has children, they are left
@@ -505,7 +519,7 @@ def __init__(self,
         elif cache_size == 0:
             self._increment_and_space_time_levy_area_cache = _EmptyDict()
         else:
-            self._increment_and_space_time_levy_area_cache = boltons.cacheutils.LRU(max_size=cache_size)
+            self._increment_and_space_time_levy_area_cache = _LRUDict(max_size=cache_size)
 
         # We keep track of the most recently queried interval, and start searching for the next interval from that
         # element of the binary tree. This is because subsequent queries are likely to be near the most recent query.