Merge branch 'master' into conditional-generation

transformers/tokenization_bert.py

@@ -44,6 +44,8 @@ PRETRAINED_VOCAB_FILES_MAP = {
         'bert-large-uncased-whole-word-masking-finetuned-squad': "https://s3.amazonaws.com/models.huggingface.co/bert/bert-large-uncased-whole-word-masking-finetuned-squad-vocab.txt",
         'bert-large-cased-whole-word-masking-finetuned-squad': "https://s3.amazonaws.com/models.huggingface.co/bert/bert-large-cased-whole-word-masking-finetuned-squad-vocab.txt",
         'bert-base-cased-finetuned-mrpc': "https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-cased-finetuned-mrpc-vocab.txt",
+        'bert-base-german-dbmdz-cased': "https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-german-dbmdz-cased-vocab.txt",
+        'bert-base-german-dbmdz-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-german-dbmdz-uncased-vocab.txt",
     }
 }
 
@@ -61,6 +63,8 @@ PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
     'bert-large-uncased-whole-word-masking-finetuned-squad': 512,
     'bert-large-cased-whole-word-masking-finetuned-squad': 512,
     'bert-base-cased-finetuned-mrpc': 512,
+    'bert-base-german-dbmdz-cased': 512,
+    'bert-base-german-dbmdz-uncased': 512,
 }
 
 PRETRAINED_INIT_CONFIGURATION = {
@@ -77,6 +81,8 @@ PRETRAINED_INIT_CONFIGURATION = {
     'bert-large-uncased-whole-word-masking-finetuned-squad': {'do_lower_case': True},
     'bert-large-cased-whole-word-masking-finetuned-squad': {'do_lower_case': False},
     'bert-base-cased-finetuned-mrpc': {'do_lower_case': False},
+    'bert-base-german-dbmdz-cased': {'do_lower_case': False},
+    'bert-base-german-dbmdz-uncased': {'do_lower_case': True},
 }
 
 
@@ -187,33 +193,59 @@ class BertTokenizer(PreTrainedTokenizer):
         out_string = ' '.join(tokens).replace(' ##', '').strip()
         return out_string
 
-    def add_special_tokens_single_sequence(self, token_ids):
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
         """
-        Adds special tokens to the a sequence for sequence classification tasks.
-        A BERT sequence has the following format: [CLS] X [SEP]
+        Build model inputs from a sequence or a pair of sequence for sequence classification tasks
+        by concatenating and adding special tokens.
+        A BERT sequence has the following format:
+            single sequence: [CLS] X [SEP]
+            pair of sequences: [CLS] A [SEP] B [SEP]
         """
-        return [self.cls_token_id] + token_ids + [self.sep_token_id]
+        if token_ids_1 is None:
+            return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
+        cls = [self.cls_token_id]
+        sep = [self.sep_token_id]
+        return cls + token_ids_0 + sep + token_ids_1 + sep
 
-    def add_special_tokens_sequence_pair(self, token_ids_0, token_ids_1):
+    def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
         """
-        Adds special tokens to a sequence pair for sequence classification tasks.
-        A BERT sequence pair has the following format: [CLS] A [SEP] B [SEP]
+        Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer ``prepare_for_model`` or ``encode_plus`` methods.
+
+        Args:
+            token_ids_0: list of ids (must not contain special tokens)
+            token_ids_1: Optional list of ids (must not contain special tokens), necessary when fetching sequence ids
+                for sequence pairs
+            already_has_special_tokens: (default False) Set to True if the token list is already formated with
+                special tokens for the model
+
+        Returns:
+            A list of integers in the range [0, 1]: 0 for a special token, 1 for a sequence token.
         """
-        sep = [self.sep_token_id]
-        cls = [self.cls_token_id]
 
-        return cls + token_ids_0 + sep + token_ids_1 + sep
+        if already_has_special_tokens:
+            if token_ids_1 is not None:
+                raise ValueError("You should not supply a second sequence if the provided sequence of "
+                                 "ids is already formated with special tokens for the model.")
+            return list(map(lambda x: 1 if x in [self.sep_token_id, self.cls_token_id] else 0, token_ids_0))
 
-    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1):
+        if token_ids_1 is not None:
+            return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
+        return [1] + ([0] * len(token_ids_0)) + [1]
+
+    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
         """
         Creates a mask from the two sequences passed to be used in a sequence-pair classification task.
         A BERT sequence pair mask has the following format:
         0 0 0 0 0 0 0 0 0 0 1 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]
         cls = [self.cls_token_id]
-
+        if token_ids_1 is None:
+            return len(cls + token_ids_0 + sep) * [0]
         return len(cls + token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1]
 
     def save_vocabulary(self, vocab_path):

transformers/tokenization_ctrl.py

+# coding=utf-8
+# Copyright 2018 Salesforce and The HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Tokenization classes for Salesforce CTRL."""
+from __future__ import (absolute_import, division, print_function,
+                        unicode_literals)
+
+import json
+import logging
+import os
+import regex as re
+from io import open
+
+from .tokenization_utils import PreTrainedTokenizer
+
+logger = logging.getLogger(__name__)
+
+VOCAB_FILES_NAMES = {
+    'vocab_file': 'vocab.json',
+    'merges_file': 'merges.txt',
+}
+
+PRETRAINED_VOCAB_FILES_MAP = {
+    'vocab_file':
+    {
+        'ctrl': "https://raw.githubusercontent.com/salesforce/ctrl/master/ctrl-vocab.json",
+    },
+    'merges_file':
+    {
+        'ctrl': "https://raw.githubusercontent.com/salesforce/ctrl/master/ctrl-merges.txt",
+    },
+}
+
+PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
+    'ctrl': 256,
+}
+
+CONTROL_CODES = {
+    "Pregnancy": 168629,
+    "Christianity": 7675,
+    "Explain": 106423,
+    "Fitness": 63440,
+    "Saving": 63163,
+    "Ask": 27171,
+    "Ass": 95985,
+    "Joke": 163509,
+    "Questions": 45622,
+    "Thoughts": 49605,
+    "Retail": 52342,
+    "Feminism": 164338,
+    "Writing": 11992,
+    "Atheism": 192263,
+    "Netflix": 48616,
+    "Computing": 39639,
+    "Opinion": 43213,
+    "Alone": 44967,
+    "Funny": 58917,
+    "Gaming": 40358,
+    "Human": 4088,
+    "India": 1331,
+    "Joker": 77138,
+    "Diet": 36206,
+    "Legal": 11859,
+    "Norman": 4939,
+    "Tip": 72689,
+    "Weight": 52343,
+    "Movies": 46273,
+    "Running": 23425,
+    "Science": 2090,
+    "Horror": 37793,
+    "Confession": 60572,
+    "Finance": 12250,
+    "Politics": 16360,
+    "Scary": 191985,
+    "Support": 12654,
+    "Technologies": 32516,
+    "Teenage": 66160,
+    "Event": 32769,
+    "Learned": 67460,
+    "Notion": 182770,
+    "Wikipedia": 37583,
+    "Books": 6665,
+    "Extract": 76050,
+    "Confessions": 102701,
+    "Conspiracy": 75932,
+    "Links": 63674,
+    "Narcissus": 150425,
+    "Relationship": 54766,
+    "Relationships": 134796,
+    "Reviews": 41671,
+    "News": 4256,
+    "Translation": 26820,
+    "multilingual": 128406,
+}
+
+def get_pairs(word):
+    """Return set of symbol pairs in a word.
+
+    Word is represented as tuple of symbols (symbols being variable-length strings).
+    """
+    pairs = set()
+    prev_char = word[0]
+    for char in word[1:]:
+        pairs.add((prev_char, char))
+        prev_char = char
+
+    pairs = set(pairs)
+    return pairs
+
+class CTRLTokenizer(PreTrainedTokenizer):
+    """
+    CTRL BPE tokenizer. Peculiarities:
+        - Byte-Pair-Encoding
+    """
+    vocab_files_names = VOCAB_FILES_NAMES
+    pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
+    max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
+    control_codes = CONTROL_CODES
+
+    def __init__(self, vocab_file, merges_file, unk_token="<unk>", **kwargs):
+        super(CTRLTokenizer, self).__init__(unk_token=unk_token, **kwargs)
+        self.max_len_single_sentence = self.max_len # no default special tokens - you can update this value if you add special tokens
+        self.max_len_sentences_pair = self.max_len # no default special tokens - you can update this value if you add special tokens
+
+        self.encoder = json.load(open(vocab_file, encoding="utf-8"))
+        self.decoder = {v:k for k,v in self.encoder.items()}
+        merges = open(merges_file, encoding='utf-8').read().split('\n')[1:-1]
+        merges = [tuple(merge.split()) for merge in merges]
+        self.bpe_ranks = dict(zip(merges, range(len(merges))))
+        self.cache = {}
+
+    @property
+    def vocab_size(self):
+        return len(self.encoder)
+
+    def bpe(self, token):
+        if token in self.cache:
+            return self.cache[token]
+        word = tuple(token)
+        word = tuple(list(word[:-1]) + [word[-1]+'</w>'])
+        pairs = get_pairs(word)
+
+        if not pairs:
+            return token
+
+        while True:
+            bigram = min(pairs, key = lambda pair: self.bpe_ranks.get(pair, float('inf')))
+            if bigram not in self.bpe_ranks:
+                break
+            first, second = bigram
+            new_word = []
+            i = 0
+            while i < len(word):
+                try:
+                    j = word.index(first, i)
+                    new_word.extend(word[i:j])
+                    i = j
+                except:
+                    new_word.extend(word[i:])
+                    break
+
+                if word[i] == first and i < len(word)-1 and word[i+1] == second:
+                    new_word.append(first+second)
+                    i += 2
+                else:
+                    new_word.append(word[i])
+                    i += 1
+            new_word = tuple(new_word)
+            word = new_word
+            if len(word) == 1:
+                break
+            else:
+                pairs = get_pairs(word)
+        word = '@@ '.join(word)
+        word = word[:-4]
+        self.cache[token] = word
+        return word
+
+    def _tokenize(self, text):
+        """ Tokenize a string.
+        """
+        split_tokens = []
+
+        text = text.split(' ')
+
+        for token in text:
+            split_tokens.extend([t for t in self.bpe(token).split(' ')])
+        return split_tokens
+
+    def _convert_token_to_id(self, token):
+        """ Converts a token (str/unicode) in an id using the vocab. """
+        return self.encoder.get(token, self.encoder.get(self.unk_token))
+
+    def _convert_id_to_token(self, index):
+        """Converts an index (integer) in a token (string/unicode) using the vocab."""
+        return self.decoder.get(index, self.unk_token)
+
+    def convert_tokens_to_string(self, tokens):
+        """ Converts a sequence of tokens (string) in a single string. """
+        out_string = ' '.join(tokens).replace('@@ ', '').strip()
+        return out_string
+
+    def save_vocabulary(self, save_directory):
+        """Save the tokenizer vocabulary and merge files to a directory."""
+        if not os.path.isdir(save_directory):
+            logger.error("Vocabulary path ({}) should be a directory".format(save_directory))
+            return
+        vocab_file = os.path.join(save_directory, VOCAB_FILES_NAMES['vocab_file'])
+        merge_file = os.path.join(save_directory, VOCAB_FILES_NAMES['merges_file'])
+
+        with open(vocab_file, 'w', encoding='utf-8') as f:
+            f.write(json.dumps(self.encoder, ensure_ascii=False))
+
+        index = 0
+        with open(merge_file, "w", encoding="utf-8") as writer:
+            writer.write(u'#version: 0.2\n')
+            for bpe_tokens, token_index in sorted(self.bpe_ranks.items(), key=lambda kv: kv[1]):
+                if index != token_index:
+                    logger.warning("Saving vocabulary to {}: BPE merge indices are not consecutive."
+                                   " Please check that the tokenizer is not corrupted!".format(merge_file))
+                    index = token_index
+                writer.write(' '.join(bpe_tokens) + u'\n')
+                index += 1
+
+        return vocab_file, merge_file
+
+    # def decode(self, token_ids, skip_special_tokens=False, clean_up_tokenization_spaces=True):
+    #     filtered_tokens = ' '.join(self.convert_ids_to_tokens(token_ids, skip_special_tokens=skip_special_tokens))
+    #     tokens_generated_so_far = re.sub('(@@ )', '', string=filtered_tokens)
+    #     tokens_generated_so_far = re.sub('(@@ ?$)', '', string=tokens_generated_so_far)
+    #     return ''.join(tokens_generated_so_far)

transformers/tokenization_roberta.py

@@ -46,12 +46,14 @@ PRETRAINED_VOCAB_FILES_MAP = {
         'roberta-base': "https://s3.amazonaws.com/models.huggingface.co/bert/roberta-base-vocab.json",
         'roberta-large': "https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-vocab.json",
         'roberta-large-mnli': "https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-mnli-vocab.json",
+        'distilroberta-base': "https://s3.amazonaws.com/models.huggingface.co/bert/distilroberta-base-vocab.json",
     },
     'merges_file':
     {
         'roberta-base': "https://s3.amazonaws.com/models.huggingface.co/bert/roberta-base-merges.txt",
         'roberta-large': "https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-merges.txt",
         'roberta-large-mnli': "https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-mnli-merges.txt",
+        'distilroberta-base': "https://s3.amazonaws.com/models.huggingface.co/bert/distilroberta-base-merges.txt",
     },
 }
 
@@ -59,6 +61,7 @@ PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
     'roberta-base': 512,
     'roberta-large': 512,
     'roberta-large-mnli': 512,
+    'distilroberta-base': 512,
 }
 
 
@@ -84,30 +87,57 @@ class RobertaTokenizer(GPT2Tokenizer):
         self.max_len_single_sentence = self.max_len - 2  # take into account special tokens
         self.max_len_sentences_pair = self.max_len - 4  # take into account special tokens
 
-    def add_special_tokens_single_sequence(self, token_ids):
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
         """
-        Adds special tokens to a sequence for sequence classification tasks.
-        A RoBERTa sequence has the following format: <s> X </s>
+        Build model inputs from a sequence or a pair of sequence for sequence classification tasks
+        by concatenating and adding special tokens.
+        A RoBERTa sequence has the following format:
+            single sequence: <s> X </s>
+            pair of sequences: <s> A </s></s> B </s>
         """
-        return [self.cls_token_id] + token_ids + [self.sep_token_id]
+        if token_ids_1 is None:
+            return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
+        cls = [self.cls_token_id]
+        sep = [self.sep_token_id]
+        return cls + token_ids_0 + sep + sep + token_ids_1 + sep
 
-    def add_special_tokens_sequence_pair(self, token_ids_0, token_ids_1):
+    def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
         """
-        Adds special tokens to a sequence pair for sequence classification tasks.
-        A RoBERTa sequence pair has the following format: <s> A </s></s> B </s>
+        Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer ``prepare_for_model`` or ``encode_plus`` methods.
+
+        Args:
+            token_ids_0: list of ids (must not contain special tokens)
+            token_ids_1: Optional list of ids (must not contain special tokens), necessary when fetching sequence ids
+                for sequence pairs
+            already_has_special_tokens: (default False) Set to True if the token list is already formated with
+                special tokens for the model
+
+        Returns:
+            A list of integers in the range [0, 1]: 0 for a special token, 1 for a sequence token.
         """
-        sep = [self.sep_token_id]
-        cls = [self.cls_token_id]
-        return cls + token_ids_0 + sep + sep + token_ids_1 + sep
+        if already_has_special_tokens:
+            if token_ids_1 is not None:
+                raise ValueError("You should not supply a second sequence if the provided sequence of "
+                                 "ids is already formated with special tokens for the model.")
+            return list(map(lambda x: 1 if x in [self.sep_token_id, self.cls_token_id] else 0, token_ids_0))
 
-    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1):
+        if token_ids_1 is None:
+            return [1] + ([0] * len(token_ids_0)) + [1]
+        return [1] + ([0] * len(token_ids_0)) + [1, 1] + ([0] * len(token_ids_1)) + [1]
+
+    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
         """
         Creates a mask from the two sequences passed to be used in a sequence-pair classification task.
         A RoBERTa sequence pair mask has the following format:
         0 0 0 0 0 0 0 0 0 0 1 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]
         cls = [self.cls_token_id]
 
-        return len(cls + token_ids_0 + sep + sep) * [0] + len(token_ids_1 + sep) * [1]
\ No newline at end of file
+        if token_ids_1 is None:
+            return len(cls + token_ids_0 + sep) * [0]
+        return len(cls + token_ids_0 + sep + sep) * [0] + len(token_ids_1 + sep) * [1]

transformers/tokenization_utils.py

@@ -337,13 +337,13 @@ class PreTrainedTokenizer(object):
                 vocab_files[file_id] = full_file_name
 
             if all(full_file_name is None for full_file_name in vocab_files.values()):
-                logger.error(
-                    "Model name '{}' was not found in model name list ({}). "
-                    "We assumed '{}' was a path or url but couldn't find tokenizer files"
-                    "at this path or url.".format(
+                raise EnvironmentError(
+                    "Model name '{}' was not found in tokenizers model name list ({}). "
+                    "We assumed '{}' was a path or url to a directory containing vocabulary files "
+                    "named {} but couldn't find such vocabulary files at this path or url.".format(
                         pretrained_model_name_or_path, ', '.join(s3_models),
-                        pretrained_model_name_or_path, ))
-                return None
+                        pretrained_model_name_or_path, 
+                        list(cls.vocab_files_names.values())))
 
         # Get files from url, cache, or disk depending on the case
         try:
@@ -353,17 +353,18 @@ class PreTrainedTokenizer(object):
                     resolved_vocab_files[file_id] = None
                 else:
                     resolved_vocab_files[file_id] = cached_path(file_path, cache_dir=cache_dir, force_download=force_download, proxies=proxies)
-        except EnvironmentError as e:
+        except EnvironmentError:
             if pretrained_model_name_or_path in s3_models:
-                logger.error("Couldn't reach server to download vocabulary.")
+                msg = "Couldn't reach server at '{}' to download vocabulary files."
             else:
-                logger.error(
-                    "Model name '{}' was not found in model name list ({}). "
-                    "We assumed '{}' was a path or url but couldn't find files {} "
-                    "at this path or url.".format(
+                msg = "Model name '{}' was not found in tokenizers model name list ({}). " \
+                    "We assumed '{}' was a path or url to a directory containing vocabulary files " \
+                    "named {}, but couldn't find such vocabulary files at this path or url.".format(
                         pretrained_model_name_or_path, ', '.join(s3_models),
-                        pretrained_model_name_or_path, str(vocab_files.keys())))
-            raise e
+                        pretrained_model_name_or_path,
+                        list(cls.vocab_files_names.values()))
+
+            raise EnvironmentError(msg)
 
         for file_id, file_path in vocab_files.items():
             if file_path == resolved_vocab_files[file_id]:
@@ -539,15 +540,9 @@ class PreTrainedTokenizer(object):
         Returns:
             Number of tokens added to sequences
         """
-
-        if pair:
-            initial_tokens_len = len(self.encode("This is a sequence") + self.encode("This is another"))
-            final_tokens_len = len(self.encode("This is a sequence", "This is another", add_special_tokens=True))
-        else:
-            initial_tokens_len = len(self.encode("This is a sequence"))
-            final_tokens_len = len(self.encode("This is a sequence", add_special_tokens=True))
-
-        return final_tokens_len - initial_tokens_len
+        token_ids_0 = []
+        token_ids_1 = []
+        return len(self.build_inputs_with_special_tokens(token_ids_0, token_ids_1 if pair else None))
 
     def add_special_tokens(self, special_tokens_dict):
         """
@@ -699,7 +694,7 @@ class PreTrainedTokenizer(object):
                 add_special_tokens=False,
                 max_length=None,
                 stride=0,
-                truncate_first_sequence=True,
+                truncation_strategy='longest_first',
                 return_tensors=None,
                 **kwargs):
         """
@@ -719,9 +714,13 @@ class PreTrainedTokenizer(object):
             max_length: if set to a number, will limit the total sequence returned so that it has a maximum length.
                 If there are overflowing tokens, those will be added to the returned dictionary
             stride: if set to a number along with max_length, the overflowing tokens returned will contain some tokens
-                from the main sequence returned. The value of this argument defined the number of additional tokens.
-            truncate_first_sequence: if there is a specified max_length, this flag will choose which sequence
-                will be truncated.
+                from the main sequence returned. The value of this argument defines the number of additional tokens.
+            truncation_strategy: string selected in the following options:
+                - 'longest_first' (default) Iteratively reduce the inputs sequence until the input is under max_length
+                    starting from the longest one at each token (when there is a pair of input sequences)
+                - 'only_first': Only truncate the first sequence
+                - 'only_second': Only truncate the second sequence
+                - 'do_not_truncate': Does not truncate (raise an error if the input sequence is longer than max_length)
             return_tensors: (optional) can be set to 'tf' or 'pt' to return respectively TensorFlow tf.constant
                 or PyTorch torch.Tensor instead of a list of python integers.
             **kwargs: passed to the `self.tokenize()` method
@@ -731,7 +730,7 @@ class PreTrainedTokenizer(object):
                                           max_length=max_length,
                                           add_special_tokens=add_special_tokens,
                                           stride=stride,
-                                          truncate_first_sequence=truncate_first_sequence,
+                                          truncation_strategy=truncation_strategy,
                                           return_tensors=return_tensors,
                                           **kwargs)
 
@@ -743,7 +742,7 @@ class PreTrainedTokenizer(object):
                     add_special_tokens=False,
                     max_length=None,
                     stride=0,
-                    truncate_first_sequence=True,
+                    truncation_strategy='longest_first',
                     return_tensors=None,
                     **kwargs):
         """
@@ -762,9 +761,13 @@ class PreTrainedTokenizer(object):
             max_length: if set to a number, will limit the total sequence returned so that it has a maximum length.
                 If there are overflowing tokens, those will be added to the returned dictionary
             stride: if set to a number along with max_length, the overflowing tokens returned will contain some tokens
-                from the main sequence returned. The value of this argument defined the number of additional tokens.
-            truncate_first_sequence: if there is a specified max_length, this flag will choose which sequence
-                will be truncated.
+                from the main sequence returned. The value of this argument defines the number of additional tokens.
+            truncation_strategy: string selected in the following options:
+                - 'longest_first' (default) Iteratively reduce the inputs sequence until the input is under max_length
+                    starting from the longest one at each token (when there is a pair of input sequences)
+                - 'only_first': Only truncate the first sequence
+                - 'only_second': Only truncate the second sequence
+                - 'do_not_truncate': Does not truncate (raise an error if the input sequence is longer than max_length)
             return_tensors: (optional) can be set to 'tf' or 'pt' to return respectively TensorFlow tf.constant
                 or PyTorch torch.Tensor instead of a list of python integers.
             **kwargs: passed to the `self.tokenize()` method
@@ -788,12 +791,11 @@ class PreTrainedTokenizer(object):
                                       max_length=max_length,
                                       add_special_tokens=add_special_tokens,
                                       stride=stride,
-                                      truncate_first_sequence=truncate_first_sequence,
+                                      truncation_strategy=truncation_strategy,
                                       return_tensors=return_tensors)
 
-
     def prepare_for_model(self, ids, pair_ids=None, max_length=None, add_special_tokens=False, stride=0,
-                          truncate_first_sequence=True, return_tensors=None):
+                          truncation_strategy='longest_first', return_tensors=None):
         """
         Prepares a sequence of input id, or a pair of sequences of inputs ids so that it can be used by the model.
         It adds special tokens, truncates
@@ -810,41 +812,50 @@ class PreTrainedTokenizer(object):
                 to their model.
             stride: window stride for overflowing tokens. Can be useful for edge effect removal when using sequential
                 list of inputs.
-            truncate_first_sequence: if set to `True` and an optional second list of input ids is provided,
-                alongside a specified `max_length`, will truncate the first sequence if the total size is superior
-                than the specified `max_length`. If set to `False`, will truncate the second sequence instead.
+            truncation_strategy: string selected in the following options:
+                - 'longest_first' (default) Iteratively reduce the inputs sequence until the input is under max_length
+                    starting from the longest one at each token (when there is a pair of input sequences)
+                - 'only_first': Only truncate the first sequence
+                - 'only_second': Only truncate the second sequence
+                - 'do_not_truncate': Does not truncate (raise an error if the input sequence is longer than max_length)
             return_tensors: (optional) can be set to 'tf' or 'pt' to return respectively TensorFlow tf.constant
                 or PyTorch torch.Tensor instead of a list of python integers.
 
         Return:
-            a dictionary containing the `input_ids` as well as the `overflowing_tokens` if a `max_length` was given.
+            A Dictionary of shape::
+
+                {
+                    input_ids: list[int],
+                    overflowing_tokens: list[int] if a ``max_length`` is specified, else None
+                    special_tokens_mask: list[int] if ``add_special_tokens`` if set to ``True``
+                }
+
+            With the fields:
+                ``input_ids``: list of tokens to be fed to a model
+
+                ``overflowing_tokens``: list of overflowing tokens if a max length is specified.
+
+                ``special_tokens_mask``: if adding special tokens, this is a list of [0, 1], with 0 specifying special added
+                tokens and 1 specifying sequence tokens.
         """
         pair = bool(pair_ids is not None)
         len_ids = len(ids)
         len_pair_ids = len(pair_ids) if pair else 0
 
         encoded_inputs = {}
-        if max_length:
-            n_added_tokens = self.num_added_tokens(pair=pair) if add_special_tokens else 0
-            if pair and n_added_tokens + (len_pair_ids if truncate_first_sequence else len_ids) >= max_length:
-                logger.warning(
-                    "You supplied a pair of sequence in which the sequence that will not be truncated is longer than the maximum specified length."
-                    "This pair of sequences will not be truncated.")
-            else:
-                if n_added_tokens + len_ids + len_pair_ids > max_length:
-                    if truncate_first_sequence or not pair:
-                        encoded_inputs["overflowing_tokens"] = ids[max_length - len_pair_ids - n_added_tokens - stride:]
-                        ids = ids[:max_length - len_pair_ids - n_added_tokens]
-                    elif not truncate_first_sequence and pair:
-                        encoded_inputs["overflowing_tokens"] = pair_ids[max_length - len_ids - n_added_tokens - stride:]
-                        pair_ids = pair_ids[:max_length - len_ids - n_added_tokens]
-                    else:
-                        logger.warning(
-                            "Cannot truncate second sequence as it is not provided. No truncation.")
+        total_len = len_ids + len_pair_ids + (self.num_added_tokens(pair=pair) if add_special_tokens else 0)
+        if max_length and total_len > max_length:
+            ids, pair_ids, overflowing_tokens = self.truncate_sequences(ids, pair_ids=pair_ids,
+                                                                        num_tokens_to_remove=total_len-max_length,
+                                                                        truncation_strategy=truncation_strategy,
+                                                                        stride=stride)
+            encoded_inputs["overflowing_tokens"] = overflowing_tokens
+            encoded_inputs["num_truncated_tokens"] = total_len - max_length
 
         if add_special_tokens:
-            sequence = self.add_special_tokens_sequence_pair(ids, pair_ids) if pair else self.add_special_tokens_single_sequence(ids)
-            token_type_ids = self.create_token_type_ids_from_sequences(ids, pair_ids) if pair else [0] * len(sequence)
+            sequence = self.build_inputs_with_special_tokens(ids, pair_ids)
+            token_type_ids = self.create_token_type_ids_from_sequences(ids, pair_ids)
+            encoded_inputs["special_tokens_mask"] = self.get_special_tokens_mask(ids, pair_ids)
         else:
             sequence = ids + pair_ids if pair else ids
             token_type_ids = [0] * len(ids) + ([1] * len(pair_ids) if pair else [])
@@ -861,20 +872,89 @@ class PreTrainedTokenizer(object):
         encoded_inputs["input_ids"] = sequence
         encoded_inputs["token_type_ids"] = token_type_ids
 
+        if max_length and len(encoded_inputs["input_ids"]) > max_length:
+            encoded_inputs["input_ids"] = encoded_inputs["input_ids"][:max_length]
+            encoded_inputs["token_type_ids"] = encoded_inputs["token_type_ids"][:max_length]
+            encoded_inputs["special_tokens_mask"] = encoded_inputs["special_tokens_mask"][:max_length]
+
         return encoded_inputs
 
-    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1):
+    def truncate_sequences(self, ids, pair_ids=None, num_tokens_to_remove=0, truncation_strategy='longest_first', stride=0):
+        """Truncates a sequence pair in place to the maximum length.
+            truncation_strategy: string selected in the following options:
+                - 'longest_first' (default) Iteratively reduce the inputs sequence until the input is under max_length
+                    starting from the longest one at each token (when there is a pair of input sequences).
+                    Overflowing tokens only contains overflow from the first sequence.
+                - 'only_first': Only truncate the first sequence. raise an error if the first sequence is shorter or equal to than num_tokens_to_remove.
+                - 'only_second': Only truncate the second sequence
+                - 'do_not_truncate': Does not truncate (raise an error if the input sequence is longer than max_length)
+        """
+        if num_tokens_to_remove <= 0:
+            return ids, pair_ids, []
+
+        if truncation_strategy == 'longest_first':
+            overflowing_tokens = []
+            for _ in range(num_tokens_to_remove):
+                if pair_ids is None or len(ids) > len(pair_ids):
+                    overflowing_tokens = [ids[-1]] + overflowing_tokens
+                    ids = ids[:-1]
+                else:
+                    pair_ids = pair_ids[:-1]
+            window_len = min(len(ids), stride)
+            if window_len > 0:
+                overflowing_tokens = ids[-window_len:] + overflowing_tokens
+        elif truncation_strategy == 'only_first':
+            assert len(ids) > num_tokens_to_remove
+            window_len = min(len(ids), stride + num_tokens_to_remove)
+            overflowing_tokens = ids[-window_len:]
+            ids = ids[:-num_tokens_to_remove]
+        elif truncation_strategy == 'only_second':
+            assert pair_ids is not None and len(pair_ids) > num_tokens_to_remove
+            window_len = min(len(pair_ids), stride + num_tokens_to_remove)
+            overflowing_tokens = pair_ids[-window_len:]
+            pair_ids = pair_ids[:-num_tokens_to_remove]
+        elif truncation_strategy == 'do_not_truncate':
+            raise ValueError("Input sequence are too long for max_length. Please select a truncation strategy.")
+        else:
+            raise ValueError("Truncation_strategy should be selected in ['longest_first', 'only_first', 'only_second', 'do_not_truncate']")
+        return (ids, pair_ids, overflowing_tokens)
+
+    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
         logger.warning("This tokenizer does not make use of special tokens.")
+        if token_ids_1 is None:
+            return len(token_ids_0) * [0]
         return [0] * len(token_ids_0) + [1] * len(token_ids_1)
 
-    def add_special_tokens_single_sequence(self, token_ids):
-        logger.warning("This tokenizer does not make use of special tokens. The sequence has been returned with no modification.")
-        return token_ids
-
-    def add_special_tokens_sequence_pair(self, token_ids_0, token_ids_1):
-        logger.warning("This tokenizer does not make use of special tokens. The two sequences have been concatenated.")
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
+        """
+        Build model inputs from a sequence or a pair of sequence for sequence classification tasks
+        by concatenating and adding special tokens.
+        A RoBERTa sequence has the following format:
+            single sequence: <s> X </s>
+            pair of sequences: <s> A </s></s> B </s>
+        """
+        logger.warning("This tokenizer does not make use of special tokens. Input is returned with no modification.")
+        if token_ids_1 is None:
+            return token_ids_0
         return token_ids_0 + token_ids_1
 
+    def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
+        """
+        Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer ``prepare_for_model`` or ``encode_plus`` methods.
+
+        Args:
+            token_ids_0: list of ids (must not contain special tokens)
+            token_ids_1: Optional list of ids (must not contain special tokens), necessary when fetching sequence ids
+                for sequence pairs
+            already_has_special_tokens: (default False) Set to True if the token list is already formated with
+                special tokens for the model
+
+        Returns:
+            A list of integers in the range [0, 1]: 0 for a special token, 1 for a sequence token.
+        """
+        return [0] * ((len(token_ids_1) if token_ids_1 else 0) + len(token_ids_0))
+
     def convert_ids_to_tokens(self, ids, skip_special_tokens=False):
         """ Converts a single index or a sequence of indices (integers) in a token "
             (resp.) a sequence of tokens (str/unicode), using the vocabulary and added tokens.

transformers/tokenization_xlm.py

@@ -754,32 +754,59 @@ class XLMTokenizer(PreTrainedTokenizer):
         out_string = ''.join(tokens).replace('</w>', ' ').strip()
         return out_string
 
-    def add_special_tokens_single_sequence(self, token_ids):
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
         """
-        Adds special tokens to a sequence for sequence classification tasks.
-        An XLM sequence has the following format: [CLS] X [SEP]
-        """
-        return [self.cls_token_id] + token_ids + [self.sep_token_id]
-
-    def add_special_tokens_sequence_pair(self, token_ids_0, token_ids_1):
-        """
-        Adds special tokens to a sequence pair for sequence classification tasks.
-        An XLM sequence pair has the following format: [CLS] A [SEP] B [SEP]
+        Build model inputs from a sequence or a pair of sequence for sequence classification tasks
+        by concatenating and adding special tokens.
+        A RoBERTa sequence has the following format:
+            single sequence: <s> X </s>
+            pair of sequences: <s> A </s></s> B </s>
         """
+        if token_ids_1 is None:
+            return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
         sep = [self.sep_token_id]
         cls = [self.cls_token_id]
         return cls + token_ids_0 + sep + token_ids_1 + sep
 
-    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1):
+    def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
+        """
+        Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer ``prepare_for_model`` or ``encode_plus`` methods.
+
+        Args:
+            token_ids_0: list of ids (must not contain special tokens)
+            token_ids_1: Optional list of ids (must not contain special tokens), necessary when fetching sequence ids
+                for sequence pairs
+            already_has_special_tokens: (default False) Set to True if the token list is already formated with
+                special tokens for the model
+
+        Returns:
+            A list of integers in the range [0, 1]: 0 for a special token, 1 for a sequence token.
+        """
+
+        if already_has_special_tokens:
+            if token_ids_1 is not None:
+                raise ValueError("You should not supply a second sequence if the provided sequence of "
+                                 "ids is already formated with special tokens for the model.")
+            return list(map(lambda x: 1 if x in [self.sep_token_id, self.cls_token_id] else 0, token_ids_0))
+
+        if token_ids_1 is not None:
+            return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
+        return [1] + ([0] * len(token_ids_0)) + [1]
+
+    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
         """
         Creates a mask from the two sequences passed to be used in a sequence-pair classification task.
         An XLM sequence pair mask has the following format:
         0 0 0 0 0 0 0 0 0 0 1 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]
         cls = [self.cls_token_id]
-
+        if token_ids_1 is None:
+            return len(cls + token_ids_0 + sep) * [0]
         return len(cls + token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1]
 
     def save_vocabulary(self, save_directory):

transformers/tokenization_xlnet.py

@@ -181,36 +181,61 @@ class XLNetTokenizer(PreTrainedTokenizer):
         out_string = ''.join(tokens).replace(SPIECE_UNDERLINE, ' ').strip()
         return out_string
 
-    def add_special_tokens_single_sequence(self, token_ids):
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
         """
-        Adds special tokens to a sequence for sequence classification tasks.
-        An XLNet sequence has the following format: X [SEP][CLS]
+        Build model inputs from a sequence or a pair of sequence for sequence classification tasks
+        by concatenating and adding special tokens.
+        A RoBERTa sequence has the following format:
+            single sequence: <s> X </s>
+            pair of sequences: <s> A </s></s> B </s>
         """
         sep = [self.sep_token_id]
         cls = [self.cls_token_id]
-        return token_ids + sep + cls
+        if token_ids_1 is None:
+            return token_ids_0 + sep + cls
+        return token_ids_0 + sep + token_ids_1 + sep + cls
 
-    def add_special_tokens_sequence_pair(self, token_ids_0, token_ids_1):
+    def get_special_tokens_mask(self, token_ids_0, token_ids_1=None, already_has_special_tokens=False):
         """
-        Adds special tokens to a sequence pair for sequence classification tasks.
-        An XLNet sequence pair has the following format: A [SEP] B [SEP][CLS]
+        Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer ``prepare_for_model`` or ``encode_plus`` methods.
+
+        Args:
+            token_ids_0: list of ids (must not contain special tokens)
+            token_ids_1: Optional list of ids (must not contain special tokens), necessary when fetching sequence ids
+                for sequence pairs
+            already_has_special_tokens: (default False) Set to True if the token list is already formated with
+                special tokens for the model
+
+        Returns:
+            A list of integers in the range [0, 1]: 0 for a special token, 1 for a sequence token.
         """
 
-        sep = [self.sep_token_id]
-        cls = [self.cls_token_id]
-        return token_ids_0 + sep + token_ids_1 + sep + cls
+        if already_has_special_tokens:
+            if token_ids_1 is not None:
+                raise ValueError("You should not supply a second sequence if the provided sequence of "
+                                 "ids is already formated with special tokens for the model.")
+            return list(map(lambda x: 1 if x in [self.sep_token_id, self.cls_token_id] else 0, token_ids_0))
+
+        if token_ids_1 is not None:
+            return ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1, 1]
+        return ([0] * len(token_ids_0)) + [1, 1]
 
-    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1):
+    def create_token_type_ids_from_sequences(self, token_ids_0, token_ids_1=None):
         """
         Creates a mask from the two sequences passed to be used in a sequence-pair classification task.
         A BERT sequence pair mask has the following format:
         0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 2
         | first sequence    | second sequence     | CLS segment ID
+        
+        if token_ids_1 is None, only returns the first portion of the mask (0's).
         """
         sep = [self.sep_token_id]
         cls = [self.cls_token_id]
         cls_segment_id = [2]
 
+        if token_ids_1 is None:
+            return len(token_ids_0 + sep + cls) * [0]
         return len(token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1] + cls_segment_id
 
     def save_vocabulary(self, save_directory):