Update to new test infra and only run conditionally

.circleci/config.yml

@@ -13,8 +13,6 @@ jobs:
             - run: sudo pip install --progress-bar off .
             - run: sudo pip install pytest codecov pytest-cov
             - run: sudo pip install tensorboardX scikit-learn
-            - run: sudo apt-get -y install libmecab-dev mecab mecab-ipadic-utf8 swig
-            - run: sudo pip install mecab-python3
             - run: python -m pytest -sv ./transformers/tests/ --cov
             - run: codecov
     build_py3_torch:
@@ -29,8 +27,6 @@ jobs:
             - run: sudo pip install --progress-bar off .
             - run: sudo pip install pytest codecov pytest-cov
             - run: sudo pip install tensorboardX scikit-learn
-            - run: sudo apt-get -y install libmecab-dev mecab mecab-ipadic-utf8 swig
-            - run: sudo pip install mecab-python3
             - run: python -m pytest -sv ./transformers/tests/ --cov
             - run: python -m pytest -sv ./examples/
             - run: codecov
@@ -46,8 +42,6 @@ jobs:
             - run: sudo pip install --progress-bar off .
             - run: sudo pip install pytest codecov pytest-cov
             - run: sudo pip install tensorboardX scikit-learn
-            - run: sudo apt-get -y install libmecab-dev mecab mecab-ipadic-utf8 swig
-            - run: sudo pip install mecab-python3
             - run: python -m pytest -sv ./transformers/tests/ --cov
             - run: codecov
     build_py2_torch:
@@ -61,8 +55,6 @@ jobs:
             - run: sudo pip install torch
             - run: sudo pip install --progress-bar off .
             - run: sudo pip install pytest codecov pytest-cov
-            - run: sudo apt-get -y install libmecab-dev mecab mecab-ipadic-utf8 swig
-            - run: sudo pip install mecab-python
             - run: python -m pytest -sv ./transformers/tests/ --cov
             - run: codecov
     build_py2_tf:
@@ -76,10 +68,18 @@ jobs:
             - run: sudo pip install tensorflow
             - run: sudo pip install --progress-bar off .
             - run: sudo pip install pytest codecov pytest-cov
-            - run: sudo apt-get -y install libmecab-dev mecab mecab-ipadic-utf8 swig
-            - run: sudo pip install mecab-python
             - run: python -m pytest -sv ./transformers/tests/ --cov
             - run: codecov
+    build_py3_custom_tokenizers:
+        working_directory: ~/transformers
+        docker:
+            - image: circleci/python:3.5
+        steps:
+            - checkout
+            - run: sudo pip install --progress-bar off .
+            - run: sudo pip install pytest
+            - run: sudo pip install mecab-python3
+            - run: python -m pytest -sv ./transformers/tests/tokenization_bert_japanese_test.py
     deploy_doc:
         working_directory: ~/transformers
         docker:

transformers/tests/tokenization_bert_japanese_test.py

@@ -16,7 +16,6 @@ from __future__ import absolute_import, division, print_function, unicode_litera
 
 import os
 import unittest
-import pytest
 from io import open
 
 from transformers.tokenization_bert import WordpieceTokenizer
@@ -25,8 +24,10 @@ from transformers.tokenization_bert_japanese import (BertJapaneseTokenizer,
                                                      VOCAB_FILES_NAMES)
 
 from .tokenization_tests_commons import CommonTestCases
+from .utils import slow, custom_tokenizers
 
 
+@custom_tokenizers
 class BertJapaneseTokenizationTest(CommonTestCases.CommonTokenizerTester):
 
     tokenizer_class = BertJapaneseTokenizer
@@ -104,7 +105,7 @@ class BertJapaneseTokenizationTest(CommonTestCases.CommonTokenizerTester):
         self.assertListEqual(tokenizer.tokenize(u"こんばんは こんばんにちは こんにちは"),
                              [u"こん", u"##ばんは", u"[UNK]", u"こんにちは"])
 
-    @pytest.mark.slow
+    @slow
     def test_sequence_builders(self):
         tokenizer = self.tokenizer_class.from_pretrained("bert-base-japanese")
 
@@ -172,7 +173,7 @@ class BertJapaneseCharacterTokenizationTest(CommonTestCases.CommonTokenizerTeste
         self.assertListEqual(tokenizer.tokenize(u"こんにちほ"),
                              [u"こ", u"ん", u"に", u"ち", u"[UNK]"])
 
-    @pytest.mark.slow
+    @slow
     def test_sequence_builders(self):
         tokenizer = self.tokenizer_class.from_pretrained("bert-base-japanese-char")
 
@@ -188,5 +189,3 @@ class BertJapaneseCharacterTokenizationTest(CommonTestCases.CommonTokenizerTeste
 
 
 
-if __name__ == '__main__':
-    unittest.main()

transformers/tests/utils.py

@@ -6,18 +6,23 @@ from distutils.util import strtobool
 from transformers.file_utils import _tf_available, _torch_available
 
 
-try:
-    run_slow = os.environ["RUN_SLOW"]
-except KeyError:
-    # RUN_SLOW isn't set, default to skipping slow tests.
-    _run_slow_tests = False
-else:
-    # RUN_SLOW is set, convert it to True or False.
+def parse_flag_from_env(key, default=False):
     try:
-        _run_slow_tests = strtobool(run_slow)
-    except ValueError:
-        # More values are supported, but let's keep the message simple.
-        raise ValueError("If set, RUN_SLOW must be yes or no.")
+        value = os.environ[key]
+    except KeyError:
+        # KEY isn't set, default to `default`.
+        _value = default
+    else:
+        # KEY is set, convert it to True or False.
+        try:
+            _value = strtobool(value)
+        except ValueError:
+            # More values are supported, but let's keep the message simple.
+            raise ValueError("If set, {} must be yes or no.".format(key))
+    return _value
+
+_run_slow_tests = parse_flag_from_env("RUN_SLOW", default=False)
+_run_custom_tokenizers = parse_flag_from_env("RUN_CUSTOM_TOKENIZERS", default=False)
 
 
 def slow(test_case):
@@ -33,6 +38,19 @@ def slow(test_case):
     return test_case
 
 
+def custom_tokenizers(test_case):
+    """
+    Decorator marking a test for a custom tokenizer.
+
+    Custom tokenizers require additional dependencies, and are skipped
+    by default. Set the RUN_CUSTOM_TOKENIZERS environment variable
+    to a truthy value to run them.
+    """
+    if not _run_custom_tokenizers:
+        test_case = unittest.skip("test of custom tokenizers")(test_case)
+    return test_case
+
+
 def require_torch(test_case):
     """
     Decorator marking a test that requires PyTorch.
@@ -59,6 +77,6 @@ def require_tf(test_case):
 
 if _torch_available:
     # Set the USE_CUDA environment variable to select a GPU.
-    torch_device = "cuda" if os.environ.get("USE_CUDA") else "cpu"
+    torch_device = "cuda" if parse_flag_from_env("USE_CUDA") else "cpu"
 else:
     torch_device = None