[WIP] Fast tokenizer for debertaV2

docs/source/index.mdx

@@ -206,7 +206,7 @@ Flax), PyTorch, and/or TensorFlow.
 |          ConvBERT           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |            CTRL             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
 |           DeBERTa           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
-|         DeBERTa-v2          |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|         DeBERTa-v2          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |            DeiT             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            DETR             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |         DistilBERT          |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |

docs/source/main_classes/tokenizer.mdx

@@ -18,9 +18,7 @@ Rust library [🤗 Tokenizers](https://github.com/huggingface/tokenizers). The "
 
 1. a significant speed-up in particular when doing batched tokenization and
 2. additional methods to map between the original string (character and words) and the token space (e.g. getting the
-   index of the token comprising a given character or the span of characters corresponding to a given token). Currently
-   no "Fast" implementation is available for the SentencePiece-based tokenizers (for T5, ALBERT, CamemBERT, XLM-RoBERTa
-   and XLNet models).
+   index of the token comprising a given character or the span of characters corresponding to a given token).
 
 The base classes [`PreTrainedTokenizer`] and [`PreTrainedTokenizerFast`]
 implement the common methods for encoding string inputs in model inputs (see below) and instantiating/saving python and

docs/source/model_doc/deberta_v2.mdx

@@ -71,6 +71,12 @@ contributed by [kamalkraj](https://huggingface.co/kamalkraj). The original code
     - create_token_type_ids_from_sequences
     - save_vocabulary
 
+## DebertaV2TokenizerFast
+
+[[autodoc]] DebertaV2TokenizerFast
+    - build_inputs_with_special_tokens
+    - create_token_type_ids_from_sequences
+
 ## DebertaV2Model
 
 [[autodoc]] DebertaV2Model

src/transformers/__init__.py

@@ -423,6 +423,7 @@
     _import_structure["models.blenderbot"].append("BlenderbotTokenizerFast")
     _import_structure["models.camembert"].append("CamembertTokenizerFast")
     _import_structure["models.deberta"].append("DebertaTokenizerFast")
+    _import_structure["models.deberta_v2"].append("DebertaV2TokenizerFast")
     _import_structure["models.distilbert"].append("DistilBertTokenizerFast")
     _import_structure["models.dpr"].extend(
         ["DPRContextEncoderTokenizerFast", "DPRQuestionEncoderTokenizerFast", "DPRReaderTokenizerFast"]
@@ -2454,6 +2455,7 @@
         from .models.clip import CLIPTokenizerFast
         from .models.convbert import ConvBertTokenizerFast
         from .models.deberta import DebertaTokenizerFast
+        from .models.deberta_v2 import DebertaV2TokenizerFast
         from .models.distilbert import DistilBertTokenizerFast
         from .models.dpr import DPRContextEncoderTokenizerFast, DPRQuestionEncoderTokenizerFast, DPRReaderTokenizerFast
         from .models.electra import ElectraTokenizerFast

src/transformers/convert_slow_tokenizer.py

@@ -819,6 +819,30 @@ def post_processor(self):
         )
 
 
+class DebertaV2Converter(SpmConverter):
+    def normalizer(self, proto):
+        list_normalizers = []
+        if self.original_tokenizer.do_lower_case:
+            list_normalizers.append(normalizers.Lowercase())
+
+        precompiled_charsmap = proto.normalizer_spec.precompiled_charsmap
+        if precompiled_charsmap:
+            list_normalizers.append(normalizers.Precompiled(precompiled_charsmap))
+        list_normalizers.append(normalizers.Replace(Regex(" {2,}"), " "))
+
+        return normalizers.Sequence(list_normalizers)
+
+    def post_processor(self):
+        return processors.TemplateProcessing(
+            single="[CLS]:0 $A:0 [SEP]:0",
+            pair="[CLS]:0 $A:0 [SEP]:0 $B:1 [SEP]:1",
+            special_tokens=[
+                ("[CLS]", self.original_tokenizer.convert_tokens_to_ids("[CLS]")),
+                ("[SEP]", self.original_tokenizer.convert_tokens_to_ids("[SEP]")),
+            ],
+        )
+
+
 class CLIPConverter(Converter):
     def converted(self) -> Tokenizer:
         vocab = self.original_tokenizer.encoder
@@ -921,6 +945,7 @@ def converted(self) -> Tokenizer:
     "CLIPTokenizer": CLIPConverter,
     "ConvBertTokenizer": BertConverter,
     "DebertaTokenizer": DebertaConverter,
+    "DebertaV2Tokenizer": DebertaV2Converter,
     "DistilBertTokenizer": BertConverter,
     "DPRReaderTokenizer": BertConverter,
     "DPRQuestionEncoderTokenizer": BertConverter,

src/transformers/models/auto/tokenization_auto.py

@@ -156,7 +156,13 @@
             ("fsmt", ("FSMTTokenizer", None)),
             ("bert-generation", ("BertGenerationTokenizer" if is_sentencepiece_available() else None, None)),
             ("deberta", ("DebertaTokenizer", "DebertaTokenizerFast" if is_tokenizers_available() else None)),
-            ("deberta-v2", ("DebertaV2Tokenizer" if is_sentencepiece_available() else None, None)),
+            (
+                "deberta-v2",
+                (
+                    "DebertaV2Tokenizer" if is_sentencepiece_available() else None,
+                    "DebertaV2TokenizerFast" if is_tokenizers_available() else None,
+                ),
+            ),
             ("rag", ("RagTokenizer", None)),
             ("xlm-prophetnet", ("XLMProphetNetTokenizer" if is_sentencepiece_available() else None, None)),
             ("speech_to_text", ("Speech2TextTokenizer" if is_sentencepiece_available() else None, None)),

src/transformers/models/deberta_v2/__init__.py

@@ -18,14 +18,17 @@
 
 from typing import TYPE_CHECKING
 
-from ...file_utils import _LazyModule, is_tf_available, is_torch_available
+from ...file_utils import _LazyModule, is_tf_available, is_tokenizers_available, is_torch_available
 
 
 _import_structure = {
     "configuration_deberta_v2": ["DEBERTA_V2_PRETRAINED_CONFIG_ARCHIVE_MAP", "DebertaV2Config"],
     "tokenization_deberta_v2": ["DebertaV2Tokenizer"],
 }
 
+if is_tokenizers_available():
+    _import_structure["tokenization_deberta_v2_fast"] = ["DebertaV2TokenizerFast"]
+
 if is_tf_available():
     _import_structure["modeling_tf_deberta_v2"] = [
         "TF_DEBERTA_V2_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -53,6 +56,9 @@
     from .configuration_deberta_v2 import DEBERTA_V2_PRETRAINED_CONFIG_ARCHIVE_MAP, DebertaV2Config
     from .tokenization_deberta_v2 import DebertaV2Tokenizer
 
+    if is_tokenizers_available():
+        from .tokenization_deberta_v2_fast import DebertaV2TokenizerFast
+
     if is_tf_available():
         from .modeling_tf_deberta_v2 import (
             TF_DEBERTA_V2_PRETRAINED_MODEL_ARCHIVE_LIST,

src/transformers/models/deberta_v2/tokenization_deberta_v2.py

@@ -140,6 +140,7 @@ def __init__(
             )
         self.do_lower_case = do_lower_case
         self.split_by_punct = split_by_punct
+        self.vocab_file = vocab_file
         self._tokenizer = SPMTokenizer(vocab_file, split_by_punct=split_by_punct, sp_model_kwargs=self.sp_model_kwargs)
 
     @property

src/transformers/models/deberta_v2/tokenization_deberta_v2_fast.py

@@ -0,0 +1,247 @@
+# coding=utf-8
+# Copyright 2020 Microsoft 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 class for model DeBERTa."""
+
+import os
+from shutil import copyfile
+from typing import Optional, Tuple
+
+from ...file_utils import is_sentencepiece_available
+from ...tokenization_utils_fast import PreTrainedTokenizerFast
+from ...utils import logging
+
+
+if is_sentencepiece_available():
+    from .tokenization_deberta_v2 import DebertaV2Tokenizer
+else:
+    DebertaV2Tokenizer = None
+
+logger = logging.get_logger(__name__)
+
+PRETRAINED_VOCAB_FILES_MAP = {
+    "vocab_file": {
+        "microsoft/deberta-v2-xlarge": "https://huggingface.co/microsoft/deberta-v2-xlarge/resolve/main/spm.model",
+        "microsoft/deberta-v2-xxlarge": "https://huggingface.co/microsoft/deberta-v2-xxlarge/resolve/main/spm.model",
+        "microsoft/deberta-v2-xlarge-mnli": "https://huggingface.co/microsoft/deberta-v2-xlarge-mnli/resolve/main/spm.model",
+        "microsoft/deberta-v2-xxlarge-mnli": "https://huggingface.co/microsoft/deberta-v2-xxlarge-mnli/resolve/main/spm.model",
+    }
+}
+
+PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
+    "microsoft/deberta-v2-xlarge": 512,
+    "microsoft/deberta-v2-xxlarge": 512,
+    "microsoft/deberta-v2-xlarge-mnli": 512,
+    "microsoft/deberta-v2-xxlarge-mnli": 512,
+}
+
+PRETRAINED_INIT_CONFIGURATION = {
+    "microsoft/deberta-v2-xlarge": {"do_lower_case": False},
+    "microsoft/deberta-v2-xxlarge": {"do_lower_case": False},
+    "microsoft/deberta-v2-xlarge-mnli": {"do_lower_case": False},
+    "microsoft/deberta-v2-xxlarge-mnli": {"do_lower_case": False},
+}
+
+VOCAB_FILES_NAMES = {"vocab_file": "spm.model"}
+
+
+class DebertaV2TokenizerFast(PreTrainedTokenizerFast):
+    r"""
+    Constructs a DeBERTa-v2 fast tokenizer. Based on [SentencePiece](https://github.com/google/sentencepiece).
+
+    Args:
+        vocab_file (`str`):
+            [SentencePiece](https://github.com/google/sentencepiece) file (generally has a *.spm* extension) that
+            contains the vocabulary necessary to instantiate a tokenizer.
+        do_lower_case (`bool`, *optional*, defaults to `False`):
+            Whether or not to lowercase the input when tokenizing.
+        bos_token (`string`, *optional*, defaults to "[CLS]"):
+            The beginning of sequence token that was used during pre-training. Can be used a sequence classifier token.
+            When building a sequence using special tokens, this is not the token that is used for the beginning of
+            sequence. The token used is the `cls_token`.
+        eos_token (`string`, *optional*, defaults to "[SEP]"):
+            The end of sequence token. When building a sequence using special tokens, this is not the token that is
+            used for the end of sequence. The token used is the `sep_token`.
+        unk_token (`str`, *optional*, defaults to `"[UNK]"`):
+            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
+            token instead.
+        sep_token (`str`, *optional*, defaults to `"[SEP]"`):
+            The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
+            sequence classification or for a text and a question for question answering. It is also used as the last
+            token of a sequence built with special tokens.
+        pad_token (`str`, *optional*, defaults to `"[PAD]"`):
+            The token used for padding, for example when batching sequences of different lengths.
+        cls_token (`str`, *optional*, defaults to `"[CLS]"`):
+            The classifier token which is used when doing sequence classification (classification of the whole sequence
+            instead of per-token classification). It is the first token of the sequence when built with special tokens.
+        mask_token (`str`, *optional*, defaults to `"[MASK]"`):
+            The token used for masking values. This is the token used when training this model with masked language
+            modeling. This is the token which the model will try to predict.
+        sp_model_kwargs (`dict`, *optional*):
+            Will be passed to the `SentencePieceProcessor.__init__()` method. The [Python wrapper for SentencePiece](https://github.com/google/sentencepiece/tree/master/python) can be used, among other things, to set:
+
+            - `enable_sampling`: Enable subword regularization.
+            - `nbest_size`: Sampling parameters for unigram. Invalid for BPE-Dropout.
+
+              - `nbest_size = {0,1}`: No sampling is performed.
+              - `nbest_size > 1`: samples from the nbest_size results.
+              - `nbest_size < 0`: assuming that nbest_size is infinite and samples from the all hypothesis (lattice)
+                using forward-filtering-and-backward-sampling algorithm.
+
+            - `alpha`: Smoothing parameter for unigram sampling, and dropout probability of merge operations for
+              BPE-dropout.
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
+    pretrained_init_configuration = PRETRAINED_INIT_CONFIGURATION
+    max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
+    slow_tokenizer_class = DebertaV2Tokenizer
+
+    def __init__(
+        self,
+        vocab_file=None,
+        tokenizer_file=None,
+        do_lower_case=False,
+        split_by_punct=False,
+        bos_token="[CLS]",
+        eos_token="[SEP]",
+        unk_token="[UNK]",
+        sep_token="[SEP]",
+        pad_token="[PAD]",
+        cls_token="[CLS]",
+        mask_token="[MASK]",
+        **kwargs
+    ) -> None:
+        super().__init__(
+            vocab_file,
+            tokenizer_file=tokenizer_file,
+            do_lower_case=do_lower_case,
+            bos_token=bos_token,
+            eos_token=eos_token,
+            unk_token=unk_token,
+            sep_token=sep_token,
+            pad_token=pad_token,
+            cls_token=cls_token,
+            mask_token=mask_token,
+            split_by_punct=split_by_punct,
+            **kwargs,
+        )
+
+        if not os.path.isfile(vocab_file):
+            raise ValueError(
+                f"Can't find a vocabulary file at path '{vocab_file}'. To load the vocabulary from a Google pretrained "
+                "model use `tokenizer = AutoTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`"
+            )
+
+        self.do_lower_case = do_lower_case
+        self.vocab_file = vocab_file
+        self.can_save_slow_tokenizer = False if not self.vocab_file else True
+
+    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 DeBERTa sequence has the following format:
+
+        - single sequence: [CLS] X [SEP]
+        - pair of sequences: [CLS] A [SEP] B [SEP]
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs to which the special tokens will be added.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
+        """
+
+        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 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[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
+                Whether or not the token list is already formatted with special tokens for the model.
+
+        Returns:
+            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
+        """
+
+        if already_has_special_tokens:
+            return super().get_special_tokens_mask(
+                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
+            )
+
+        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):
+        """
+        Create a mask from the two sequences passed to be used in a sequence-pair classification task. A DeBERTa
+        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
+        | first sequence    | second sequence |
+        ```
+
+        If `token_ids_1` is `None`, this method only returns the first portion of the mask (0s).
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given
+            sequence(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: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        if not self.can_save_slow_tokenizer:
+            raise ValueError(
+                "Your fast tokenizer does not have the necessary information to save the vocabulary for a slow "
+                "tokenizer."
+            )
+
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+
+        return (out_vocab_file,)