Merge pull request #40 from ResponsiblyAI/dev

v0.1.2

.scrutinizer.yml

@@ -1,6 +1,6 @@
 build:
   environment:
-      python: 3.5.0
+      python: 3.6
   tests:
     override:
       - pylint-run --rcfile=.pylint.ini

.travis.yml

@@ -1,14 +1,11 @@
-dist: xenial   # required for Python >= 3.7
 language: python
 python:
-  - "3.5"
-  - "3.5-dev"  # 3.5 development branch
   - "3.6"
-  - "3.6-dev"  # 3.6 development branch
   - "3.7"
-  - "3.7-dev"  # 3.7 development branch
+  - "3.7"
+  - "3.8"
   - "3.8-dev"  # 3.8 development branch
-  - "nightly"  # nightly build
+  # - "nightly"  # nightly build
 
 cache:
   pip: true
@@ -18,22 +15,24 @@ cache:
 env:
   global:
     - RANDOM_SEED=0
-    - BOTO_CONFIG=/dev/null  # https://github.com/travis-ci/travis-ci/issues/7940
+    # https://github.com/travis-ci/travis-ci/issues/7940
+    - BOTO_CONFIG=/dev/null
 
 before_install:
   - "sudo apt-get install python3-tk"
-  - pip install pipenv
+  # Work around https://github.com/jaraco/zipp/issues/40
+  - python -m pip install -U --upgrade pip setuptools virtualenv
+  - python -m pip install -U pipenv 
   - make doctor
 
 install:
   - make install
 
 script:
-  - make check
-  - make test
+  - make ci
 
 after_success:
-  - pip install coveralls scrutinizer-ocular
+  - python -m pip install -U coveralls scrutinizer-ocular
   - coveralls
   - ocular
 

.verchew.ini

@@ -6,12 +6,12 @@ version = GNU Make
 [Python]
 
 cli = python
-version = Python 3.5.
+version = Python 3.6.
 
 [pipenv]
 
 cli = pipenv
-versions = 10. | 11. | 2018.11 | 2018.10
+versions = 2019. | 2020.
 
 [Graphviz]
 

CHANGELOG.rst

@@ -1,6 +1,13 @@
 Revision History
 ================
 
+0.1.2 (2020/09/15)
+------------------
+
+- Fix Travis CI issues with pipenv
+
+- Fix bugs with word embedding bias
+
 0.1.1 (2019/08/04)
 ------------------
 

Pipfile

@@ -3,9 +3,6 @@ url = "https://pypi.org/simple"
 verify_ssl = true
 name = "pypi"
 
-[requires]
-python_version = "3.5"
-
 [packages]
 responsibly = {editable = true,path = "."}
 
@@ -18,7 +15,11 @@ pep8 = "*"
 pydocstyle = "*"
 restructuredtext_lint = "*"
 # Testing
-pytest = "~= 3.3"
+# pytest >= 5.4.0 doesn't work with pytest-sugar <= 0.9.2
+# once https://github.com/Teemu/pytest-sugar/pull/188 is
+# merged, the upper bound can be removed.
+# Ref: https://github.com/pytest-dev/pytest/issues/6931
+pytest = ">=5.3.5,<5.4.0"
 pytest-describe = "*"
 pytest-expecter = "*"
 pytest-random = "*"
@@ -43,3 +44,6 @@ jupyter = "*"
 notebook = ">=5.7.8"
 requests = ">=2.20.0"
 sniffer = "*"
+# Work around
+zipp = ">=2.2.0"
+

README.rst

@@ -63,7 +63,7 @@ or write us in `Gitter <https://gitter.im/ResponsiblyAI/responsibly>`_.
 Requirements
 ------------
 
--  Python 3.5+
+-  Python 3.6+
 
 Installation
 ------------

pyproject.toml

@@ -0,0 +1,3 @@
+[build-system]
+requires = ["setuptools", "wheel"]
+build-backend = "setuptools.build_meta:__legacy__"

responsibly/__init__.py

@@ -6,7 +6,7 @@
 __project__ = 'responsibly'
 __description__ = 'Toolkit for Auditing and Mitigating Bias and Fairness of Machine Learning Systems 🔎🤖🧰'
 __url__ = 'https://docs.responsibly.ai'
-__version__ = '0.1.1'
+__version__ = '0.1.2'
 __author__ = 'Shlomi Hod'
 __author_email__ = 'shlomi.hod@gmail.com'
 __license__ = 'MIT'

responsibly/fairness/metrics/score.py

@@ -24,7 +24,7 @@ def _get_labels(ys, labels):
         labels = unique_labels(ys)
     else:
         labels = np.asarray(labels)
-        if np.all([l not in ys for l in labels]):
+        if np.all([label not in ys for label in labels]):
             raise ValueError('At least one label specified must be in y.')
 
     return labels
@@ -190,6 +190,7 @@ def roc_curve_by_attr(y_true, y_score, x_sens,
 
     grouped = _groupby_y_x_sens(y_true, y_score, x_sens)
 
+    # pylint: disable=too-many-function-args
     roc_curves = {x_sens_value: roc_curve(group['y_true'],
                                           group['y_score'],
                                           pos_label, sample_weight,

responsibly/fairness/metrics/utils.py

@@ -1,5 +1,5 @@
 import pandas as pd
-from sklearn.metrics.classification import _check_targets
+from sklearn.metrics._classification import _check_targets
 
 
 def _assert_binary(y1, y2=None):

responsibly/tests/conftest.py

@@ -5,15 +5,18 @@ def pytest_configure(config):
     """Disable verbose output when running tests."""
 
     terminal = config.pluginmanager.getplugin('terminal')
-    base = terminal.TerminalReporter
 
-    class QuietReporter(base):
-        """Reporter that only shows dots when running tests."""
+    class QuietReporter(terminal.TerminalReporter):
+        @property
+        def verbosity(self):
+            return 0
 
-        def __init__(self, *args, **kwargs):
-            super().__init__(*args, **kwargs)
-            self.verbosity = 0
-            self.showlongtestinfo = False
-            self.showfspath = False
+        @property
+        def showlongtestinfo(self):
+            return False
+
+        @property
+        def showfspath(self):
+            return False
 
     terminal.TerminalReporter = QuietReporter

responsibly/tests/test_we.py

@@ -58,6 +58,7 @@ def test_assert_gensim_keyed_vectors():
 
 
 def test_project_params():
+    # pylint: disable=arguments-out-of-order
     v = np.array([1, 2, 3])
     u = np.array([-4, 5, -6])
 
@@ -454,5 +455,5 @@ def test_compute_association(gender_biased_w2v_small):
     """
     (r, pvalue), _ = gender_biased_w2v_small.compute_factual_association()
 
-    assert isclose(r, 0.7070401592764508)
-    assert isclose(pvalue, 1.4324502214459908e-06)
+    assert isclose(r, 0.7070401592764508, abs_tol=ATOL)
+    assert isclose(pvalue, 1.4324502214459908e-06, abs_tol=ATOL)

responsibly/we/bias.py

@@ -83,6 +83,7 @@
 from sklearn.decomposition import PCA
 from sklearn.metrics.pairwise import euclidean_distances
 from sklearn.svm import LinearSVC
+from tabulate import tabulate
 from tqdm import tqdm
 
 from responsibly.consts import RANDOM_STATE
@@ -96,7 +97,6 @@
     project_vector, reject_vector, round_to_extreme,
     take_two_sides_extreme_sorted, update_word_vector,
 )
-from tabulate import tabulate
 
 
 DIRECTION_METHODS = ['single', 'sum', 'pca']
@@ -122,6 +122,8 @@ class BiasWordEmbedding:
 
     def __init__(self, model, only_lower=False, verbose=False,
                  identify_direction=False, to_normalize=True):
+        # pylint: disable=undefined-variable
+
         assert_gensim_keyed_vectors(model)
 
         # TODO: this is bad Python, ask someone about it
@@ -346,10 +348,11 @@ def plot_projection_scores(self, words, n_extreme=10,
         projections_df['color'] = ((projections_df['projection'] + 0.5)
                                    .apply(cmap))
 
-        most_extream_projection = (projections_df['projection']
-                                   .abs()
-                                   .max()
-                                   .round(1))
+        most_extream_projection = np.round(
+            projections_df['projection']
+            .abs()
+            .max(),
+            decimals=1)
 
         sns.barplot(x='projection', y='word', data=projections_df,
                     palette=projections_df['color'])
@@ -522,7 +525,7 @@ def generate_analogies(self, n_analogies=100, seed='ends',
         :return: Data Frame of analogies (x, y), their distances,
                  and their cosine similarity scores
         """
-        # pylint: disable=C0301,R0914
+        # pylint: disable=C0301,R0914,E1136
 
         if not unrestricted:
             warnings.warn('Not Using unrestricted most_similar '

scent.py

@@ -1,5 +1,5 @@
 """Configuration file for sniffer."""
-# pylint: disable=superfluous-parens,bad-continuation,unpacking-non-sequence
+# pylint: disable=superfluous-parens,unpacking-non-sequence
 
 import subprocess
 import time

setup.py

@@ -7,13 +7,6 @@
 
 
 PACKAGE_NAME = 'responsibly'
-MINIMUM_PYTHON_VERSION = '3.5'
-
-
-def check_python_version():
-    """Exit when the Python version is too low."""
-    if sys.version < MINIMUM_PYTHON_VERSION:
-        sys.exit("Python {0}+ is required.".format(MINIMUM_PYTHON_VERSION))
 
 
 def read_package_variable(key, filename='__init__.py'):
@@ -36,8 +29,6 @@ def build_description():
     return readme + '\n' + changelog
 
 
-check_python_version()
-
 setuptools.setup(
     name=read_package_variable('__project__'),
     version=read_package_variable('__version__'),
@@ -69,11 +60,12 @@ def build_description():
         'Programming Language :: Python',
         'Programming Language :: Python :: 3',
         'Programming Language :: Python :: 3 :: Only',
-        'Programming Language :: Python :: 3.5',
         'Programming Language :: Python :: 3.6',
         'Programming Language :: Python :: 3.7',
     ],
 
+    python_requires='>=3.6',
+
     install_requires=[
         "numpy >= 1.15",
         "scipy >= 1.1",