Remove dependency on pytest for running tests (#2055)

* Switch to plain unittest for skipping slow tests.

Add a RUN_SLOW environment variable for running them.

* Switch to plain unittest for PyTorch dependency.

* Switch to plain unittest for TensorFlow dependency.

* Avoid leaking open files in the test suite.

This prevents spurious warnings when running tests.

* Fix unicode warning on Python 2 when running tests.

The warning was:

    UnicodeWarning: Unicode equal comparison failed to convert both arguments to Unicode - interpreting them as being unequal

* Support running PyTorch tests on a GPU.

Reverts 27e015bd.

* Tests no longer require pytest.

* Make tests pass on cuda

transformers/tokenization_albert.py

@@ -141,7 +141,7 @@ class AlbertTokenizer(PreTrainedTokenizer):
             pieces = self.sp_model.SampleEncodeAsPieces(text, 64, 0.1)
         new_pieces = []
         for piece in pieces:
-            if len(piece) > 1 and piece[-1] == ',' and piece[-2].isdigit():
+            if len(piece) > 1 and piece[-1] == str(',') and piece[-2].isdigit():
                 cur_pieces = self.sp_model.EncodeAsPieces(
                     piece[:-1].replace(SPIECE_UNDERLINE, ''))
                 if piece[0] != SPIECE_UNDERLINE and cur_pieces[0][0] == SPIECE_UNDERLINE:
@@ -225,9 +225,9 @@ class AlbertTokenizer(PreTrainedTokenizer):
         """
         Creates a mask from the two sequences passed to be used in a sequence-pair classification task.
         An ALBERT sequence pair mask has the following format:
-        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 
-        | first sequence    | second sequence     
-        
+        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1
+        | first sequence    | second sequence
+
         if token_ids_1 is None, only returns the first portion of the mask (0's).
         """
         sep = [self.sep_token_id]