Merge branch 'dev' into main

.github/workflows/python-publish.yml

@@ -0,0 +1,34 @@
+# This workflow will upload a Python Package
+# using Twine when a release is created
+
+name: publish
+
+on:
+  release:
+    types: [created]
+
+jobs:
+  deploy:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: '3.x'
+
+    - name: Install dependencies
+      run: |
+        python3 -m pip install --upgrade pip
+        python3 -m pip install setuptools wheel twine
+
+    - name: Build and publish
+      env:
+        TWINE_USERNAME: __token__
+        TWINE_PASSWORD: ${{ secrets.PYPI_TOKEN }}
+      run: |
+        python3 setup.py sdist bdist_wheel
+        python3 -m twine upload dist/*

.github/workflows/python-test.yml

@@ -0,0 +1,44 @@
+# This workflow will install Python dependencies,
+# then run tests over a variety of Python versions.
+
+name: test
+
+on:
+  push:
+    branches: [ main, dev ]
+    tags: [ '*' ]
+  pull_request:
+    branches: [ main, dev ]
+
+jobs:
+  build:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-latest] # [ubuntu-latest, windows-latest, macos-latest]
+        python-version: [3.8]
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python-version }}
+
+    - name: Install dependencies
+      run: |
+        python3 -m pip install --upgrade pip
+        python3 -m pip install -r requirements-test.txt
+        python3 -m pip install -r requirements.txt
+
+    - name: Test with pytest
+      run: |
+        python3 -m pytest --cov=disent tests/
+
+    - uses: codecov/codecov-action@v1
+      with:
+        token: ${{ secrets.CODECOV_TOKEN }}
+        fail_ci_if_error: true
+      # codecov automatically merges all generated files
+      # if: matrix.os == 'ubuntu-latest' && matrix.python-version == 3.9

MANIFEST.in

@@ -0,0 +1 @@
+include requirements.txt

disent/util/__init__.py

@@ -1,5 +1,6 @@
 import functools
 import logging
+import os
 import time
 from dataclasses import dataclass
 
@@ -19,6 +20,31 @@
 # ========================================================================= #
 
 
+def is_test_run():
+    """
+    This is used internally to test some scripts. There is no need
+    to use this function in your own scripts.
+    """
+    return bool(os.environ.get('DISENT_TEST_RUN', False))
+
+
+def test_run_int(integer: int, div=100):
+    """
+    This is used to test some scripts, by dividing the input number.
+    There is no need to use this function in your own scripts.
+    """
+    return ((integer + div - 1) // div) if is_test_run() else integer
+
+
+def _set_test_run():
+    os.environ['DISENT_TEST_RUN'] = 'True'
+
+
+# ========================================================================= #
+# seeds                                                                     #
+# ========================================================================= #
+
+
 def seed(long=777):
     """
     https://pytorch.org/docs/stable/notes/randomness.html

docs/examples/overview_framework_adagvae.py

@@ -6,6 +6,7 @@
 from disent.frameworks.vae.weaklysupervised import AdaVae
 from disent.model.ae import EncoderConv64, DecoderConv64, GaussianAutoEncoder
 from disent.transform import ToStandardisedTensor
+from disent.util import is_test_run
 
 data: GroundTruthData = XYSquaresData()
 dataset: Dataset = GroundTruthDatasetOrigWeakPairs(data, transform=ToStandardisedTensor())
@@ -20,5 +21,5 @@
     cfg=AdaVae.cfg(beta=4, average_mode='gvae', symmetric_kl=False)
 )
 
-trainer = pl.Trainer(logger=False, checkpoint_callback=False)
+trainer = pl.Trainer(logger=False, checkpoint_callback=False, fast_dev_run=is_test_run())
 trainer.fit(module, dataloader)

docs/examples/overview_framework_ae.py

@@ -6,6 +6,7 @@
 from disent.frameworks.ae.unsupervised import AE
 from disent.model.ae import EncoderConv64, DecoderConv64, AutoEncoder
 from disent.transform import ToStandardisedTensor
+from disent.util import is_test_run
 
 data: GroundTruthData = XYSquaresData()
 dataset: Dataset = GroundTruthDataset(data, transform=ToStandardisedTensor())
@@ -20,5 +21,5 @@
     cfg=AE.cfg()
 )
 
-trainer = pl.Trainer(logger=False, checkpoint_callback=False)
+trainer = pl.Trainer(logger=False, checkpoint_callback=False, max_steps=1, fast_dev_run=is_test_run())
 trainer.fit(module, dataloader)

docs/examples/overview_framework_betavae.py

@@ -6,6 +6,7 @@
 from disent.frameworks.vae.unsupervised import BetaVae
 from disent.model.ae import EncoderConv64, DecoderConv64, GaussianAutoEncoder
 from disent.transform import ToStandardisedTensor
+from disent.util import is_test_run
 
 data: GroundTruthData = XYSquaresData()
 dataset: Dataset = GroundTruthDataset(data, transform=ToStandardisedTensor())
@@ -20,5 +21,5 @@
     cfg=BetaVae.cfg(beta=4)
 )
 
-trainer = pl.Trainer(logger=False, checkpoint_callback=False)
+trainer = pl.Trainer(logger=False, checkpoint_callback=False, max_steps=1, fast_dev_run=is_test_run())
 trainer.fit(module, dataloader)

docs/examples/overview_metrics.py

@@ -7,6 +7,7 @@
 from disent.metrics import metric_dci, metric_mig
 from disent.model.ae import EncoderConv64, DecoderConv64, GaussianAutoEncoder
 from disent.transform import ToStandardisedTensor
+from disent.util import is_test_run, test_run_int
 
 data = XYObjectData()
 dataset = GroundTruthDataset(data, transform=ToStandardisedTensor())
@@ -23,16 +24,16 @@ def make_vae(beta):
     )
 
 def train(module):
-    trainer = pl.Trainer(logger=False, checkpoint_callback=False, max_steps=256)
+    trainer = pl.Trainer(logger=False, checkpoint_callback=False, max_steps=256, fast_dev_run=is_test_run())
     trainer.fit(module, dataloader)
 
     # we cannot guarantee which device the representation is on
     get_repr = lambda x: module.encode(x.to(module.device))
 
     # evaluate
     return {
-        **metric_dci(dataset, get_repr, num_train=1000, num_test=500, boost_mode='sklearn'),
-        **metric_mig(dataset, get_repr, num_train=2000),
+        **metric_dci(dataset, get_repr, num_train=10 if is_test_run() else 1000, num_test=5 if is_test_run() else 500, boost_mode='sklearn'),
+        **metric_mig(dataset, get_repr, num_train=20 if is_test_run() else 2000),
     }
 
 a_results = train(make_vae(beta=4))

requirements-test.txt

@@ -0,0 +1,3 @@
+
+pytest == 6.2.2
+pytest-cov == 2.11.1

requirements.txt

@@ -1,58 +1,58 @@
-pip == 21.0
+pip>=21.0
 
 # DATA SCIENCE & ML
 # =================
-numpy == 1.19.5
-# pandas == 1.2.1
+numpy==1.19.5
+# pandas==1.2.1
 
-# opencv-python == 4.5.1.48
+# opencv-python==4.5.1.48
 
-torch == 1.7.1
-torchvision == 0.8.2
-torch-optimizer == 0.1.0
-pytorch-lightning == 1.1.5
-# pytorch-lightning-bolts == 0.3.0
+torch==1.7.1
+torchvision==0.8.2
+torch-optimizer==0.1.0
+pytorch-lightning==1.1.7
+# pytorch-lightning-bolts==0.3.0
 
-# tensorflow == 2.4.1
-# tensorflow-gpu == 2.4.1
-# tensorboard == 2.4.1
+# tensorflow==2.4.1
+# tensorflow-gpu==2.4.1
+# tensorboard==2.4.1
 
-scipy == 1.6.0
-scikit-learn == 0.24.1
-# xgboost == 1.3.3
-# lightgbm == 3.1.1
+scipy==1.6.0
+scikit-learn==0.24.1
+# xgboost==1.3.3
+# lightgbm==3.1.1
 
 # IMAGE AUGMENTATION
 # ==================
-kornia == 0.4.1  # $ pip install git+https://github.com/kornia/kornia
-# imgaug == 0.4.0
-# albumentations == 0.5.2
+kornia==0.4.1  # $ pip install git+https://github.com/kornia/kornia
+# imgaug==0.4.0
+# albumentations==0.5.2
 
 # INPUT / OUTPUT
 # ================
-# psutil == 5.8.0
-# pillow == 8.1.0
-# imageio == 2.9.0
-# moviepy == 1.0.3
-h5py >= 2.10.0  # as of tensorflow 2.4 it does not support h5py 3+
+# psutil==5.8.0
+# pillow==8.1.0
+# imageio==2.9.0
+# moviepy==1.0.3
+h5py>=2.10.0  # as of tensorflow 2.4 it does not support h5py 3+
 
 # GRAPHING & DISPLAY
 # ==================
-matplotlib == 3.3.3
-# plotly == 4.14.3
-# seaborn == 0.11.1
-# tqdm == 4.56.0
+matplotlib==3.3.4
+# plotly==4.14.3
+# seaborn==0.11.1
+# tqdm==4.56.0
 
-# streamlit == 0.75.0
-wandb == 0.10.15
+# streamlit==0.75.0
+wandb==0.10.17
 
 # UTILITY
 # =======
-# pytest == 6.2.1
+# pytest==6.2.1
 
-hydra-core == 1.0.5
-hydra-colorlog == 1.0.0
-# hydra-joblib-launcher == 1.1.1
-hydra-submitit-launcher == 1.0.1
+hydra-core==1.0.6
+hydra-colorlog==1.0.0
+# hydra-joblib-launcher==1.1.1
+hydra-submitit-launcher==1.1.0
 
-# submitit == 1.1.5
+# submitit==1.1.5

setup.py

@@ -0,0 +1,33 @@
+import setuptools
+
+
+with open("README.md", "r", encoding="utf-8") as file:
+    long_description = file.read()
+
+with open('requirements.txt', 'r') as f:
+    install_requires = (req[0] for req in map(lambda x: x.split('#'), f.readlines()))
+    install_requires = [req for req in map(str.strip, install_requires) if req]
+
+
+setuptools.setup(
+    name="disent",
+    author="Nathan Juraj Michlo",
+    author_email="[email protected]",
+
+    version="0.0.1.dev1",
+    python_requires="==3.8",
+    packages=setuptools.find_packages(),
+
+    install_requires=install_requires,
+
+    url="https://github.com/nmichlo/eunomia",
+    description="Vae disentanglement framework built with pytorch lightning.",
+    long_description=long_description,
+    long_description_content_type="text/markdown",
+
+    classifiers=[
+        "Operating System :: OS Independent",
+        "Programming Language :: Python :: 3.8",
+        "Intended Audience :: Science/Research",
+    ],
+)

tests/test_docs_examples.py

@@ -0,0 +1,30 @@
+import sys
+from contextlib import contextmanager
+import pytest
+import os
+from glob import glob
+from disent.util import _set_test_run
+
+
+@contextmanager
+def no_stdout():
+    old_stdout = sys.stdout
+    sys.stdout = open(os.devnull, 'w')
+    yield
+    sys.stdout = old_stdout
+
+
+ROOT_DIR = os.path.abspath(__file__ + '/../..')
+
+
+@pytest.mark.parametrize("module", [
+    os.path.relpath(path, ROOT_DIR).replace('/', '.')[:-3]
+    for path in glob(os.path.join(ROOT_DIR, 'docs/examples/**.py'))
+])
+def test_docs_examples(capsys, module):
+    # make sure everything doesnt take 5 years to run
+    _set_test_run()
+    # run all the files in the examples folder
+    import importlib
+    with no_stdout():
+        importlib.import_module(module)

disent/data/util/__test__state_space.py → tests/test_state_space.py

@@ -45,6 +45,6 @@ def test_new_functions():
     s = StateSpace([2, 4, 6])
     print(np.max([s.sample_factors((2, 2), factor_indices=[2, 1, 2, 2]) for i in range(100)], axis=0))
     print(np.max([s.sample_missing_factors([[1, 1], [2, 2]], known_factor_indices=[0, 2]) for i in range(100)], axis=0))
-    print(np.min([s.resample_radius([[0, 1, 2], [0, 0, 0]], resample_radius=1, distinct=True) for i in range(1000)], axis=0).tolist())
-    print(np.max([s.resample_radius([[0, 1, 2], [0, 0, 0]], resample_radius=1, distinct=True) for i in range(1000)], axis=0).tolist())
+    # print(np.min([s.resample_radius([[0, 1, 2], [0, 0, 0]], resample_radius=1, distinct=True) for i in range(1000)], axis=0).tolist())
+    # print(np.max([s.resample_radius([[0, 1, 2], [0, 0, 0]], resample_radius=1, distinct=True) for i in range(1000)], axis=0).tolist())