modeling_auto.py 70.7 KB
Newer Older
thomwolf's avatar
thomwolf committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
# coding=utf-8
# Copyright 2018 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.
""" Auto Model class. """


import logging
Julien Chaumond's avatar
Julien Chaumond committed
19
from collections import OrderedDict
thomwolf's avatar
thomwolf committed
20

21
22
from .configuration_auto import (
    AlbertConfig,
23
    AutoConfig,
Sam Shleifer's avatar
Sam Shleifer committed
24
    BartConfig,
25
26
27
28
    BertConfig,
    CamembertConfig,
    CTRLConfig,
    DistilBertConfig,
Lysandre's avatar
Lysandre committed
29
    FlaubertConfig,
30
31
32
    GPT2Config,
    OpenAIGPTConfig,
    RobertaConfig,
33
    T5Config,
34
35
36
    TransfoXLConfig,
    XLMConfig,
    XLMRobertaConfig,
Aymeric Augustin's avatar
Aymeric Augustin committed
37
38
    XLNetConfig,
)
39
from .configuration_utils import PretrainedConfig
Aymeric Augustin's avatar
Aymeric Augustin committed
40
41
42
43
44
45
from .modeling_albert import (
    ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
    AlbertForMaskedLM,
    AlbertForQuestionAnswering,
    AlbertForSequenceClassification,
    AlbertModel,
46
)
Sam Shleifer's avatar
Sam Shleifer committed
47
from .modeling_bart import BART_PRETRAINED_MODEL_ARCHIVE_MAP, BartForMaskedLM, BartForSequenceClassification, BartModel
48
from .modeling_bert import (
Aymeric Augustin's avatar
Aymeric Augustin committed
49
    BERT_PRETRAINED_MODEL_ARCHIVE_MAP,
50
    BertForMaskedLM,
thomwolf's avatar
thomwolf committed
51
    BertForPreTraining,
52
    BertForQuestionAnswering,
Aymeric Augustin's avatar
Aymeric Augustin committed
53
    BertForSequenceClassification,
54
    BertForTokenClassification,
Aymeric Augustin's avatar
Aymeric Augustin committed
55
    BertModel,
56
)
Aymeric Augustin's avatar
Aymeric Augustin committed
57
58
59
60
61
62
from .modeling_camembert import (
    CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
    CamembertForMaskedLM,
    CamembertForSequenceClassification,
    CamembertForTokenClassification,
    CamembertModel,
63
)
Aymeric Augustin's avatar
Aymeric Augustin committed
64
from .modeling_ctrl import CTRL_PRETRAINED_MODEL_ARCHIVE_MAP, CTRLLMHeadModel, CTRLModel
65
from .modeling_distilbert import (
Aymeric Augustin's avatar
Aymeric Augustin committed
66
    DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
67
    DistilBertForMaskedLM,
Aymeric Augustin's avatar
Aymeric Augustin committed
68
    DistilBertForQuestionAnswering,
69
70
    DistilBertForSequenceClassification,
    DistilBertForTokenClassification,
Aymeric Augustin's avatar
Aymeric Augustin committed
71
    DistilBertModel,
72
)
Lysandre's avatar
Lysandre committed
73
74
75
76
77
78
79
from .modeling_flaubert import (
    FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
    FlaubertForQuestionAnswering,
    FlaubertForSequenceClassification,
    FlaubertModel,
    FlaubertWithLMHeadModel,
)
Aymeric Augustin's avatar
Aymeric Augustin committed
80
81
82
83
84
from .modeling_gpt2 import GPT2_PRETRAINED_MODEL_ARCHIVE_MAP, GPT2LMHeadModel, GPT2Model
from .modeling_openai import OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP, OpenAIGPTLMHeadModel, OpenAIGPTModel
from .modeling_roberta import (
    ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
    RobertaForMaskedLM,
Julien Chaumond's avatar
Julien Chaumond committed
85
    RobertaForQuestionAnswering,
Aymeric Augustin's avatar
Aymeric Augustin committed
86
87
88
    RobertaForSequenceClassification,
    RobertaForTokenClassification,
    RobertaModel,
89
)
Aymeric Augustin's avatar
Aymeric Augustin committed
90
91
92
93
94
95
96
97
from .modeling_t5 import T5_PRETRAINED_MODEL_ARCHIVE_MAP, T5Model, T5WithLMHeadModel
from .modeling_transfo_xl import TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_MAP, TransfoXLLMHeadModel, TransfoXLModel
from .modeling_xlm import (
    XLM_PRETRAINED_MODEL_ARCHIVE_MAP,
    XLMForQuestionAnswering,
    XLMForSequenceClassification,
    XLMModel,
    XLMWithLMHeadModel,
98
99
)
from .modeling_xlm_roberta import (
Aymeric Augustin's avatar
Aymeric Augustin committed
100
    XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
101
    XLMRobertaForMaskedLM,
Aymeric Augustin's avatar
Aymeric Augustin committed
102
    XLMRobertaForSequenceClassification,
103
    XLMRobertaForTokenClassification,
Aymeric Augustin's avatar
Aymeric Augustin committed
104
105
106
107
108
109
110
111
112
    XLMRobertaModel,
)
from .modeling_xlnet import (
    XLNET_PRETRAINED_MODEL_ARCHIVE_MAP,
    XLNetForQuestionAnswering,
    XLNetForSequenceClassification,
    XLNetForTokenClassification,
    XLNetLMHeadModel,
    XLNetModel,
113
)
thomwolf's avatar
thomwolf committed
114

thomwolf's avatar
thomwolf committed
115

116
logger = logging.getLogger(__name__)
thomwolf's avatar
thomwolf committed
117
118


119
120
ALL_PRETRAINED_MODEL_ARCHIVE_MAP = dict(
    (key, value)
121
122
    for pretrained_map in [
        BERT_PRETRAINED_MODEL_ARCHIVE_MAP,
Sam Shleifer's avatar
Sam Shleifer committed
123
        BART_PRETRAINED_MODEL_ARCHIVE_MAP,
124
125
126
127
128
129
130
131
132
133
134
        OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP,
        TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_MAP,
        GPT2_PRETRAINED_MODEL_ARCHIVE_MAP,
        CTRL_PRETRAINED_MODEL_ARCHIVE_MAP,
        XLNET_PRETRAINED_MODEL_ARCHIVE_MAP,
        XLM_PRETRAINED_MODEL_ARCHIVE_MAP,
        ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
        DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
        ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
        CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
        T5_PRETRAINED_MODEL_ARCHIVE_MAP,
Lysandre's avatar
Lysandre committed
135
        FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
Lysandre's avatar
Lysandre committed
136
        XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
137
138
139
    ]
    for key, value, in pretrained_map.items()
)
140

Julien Chaumond's avatar
Julien Chaumond committed
141
MODEL_MAPPING = OrderedDict(
Julien Chaumond's avatar
Julien Chaumond committed
142
143
144
145
146
    [
        (T5Config, T5Model),
        (DistilBertConfig, DistilBertModel),
        (AlbertConfig, AlbertModel),
        (CamembertConfig, CamembertModel),
147
        (XLMRobertaConfig, XLMRobertaModel),
Sam Shleifer's avatar
Sam Shleifer committed
148
        (BartConfig, BartModel),
149
        (RobertaConfig, RobertaModel),
Julien Chaumond's avatar
Julien Chaumond committed
150
151
152
153
154
        (BertConfig, BertModel),
        (OpenAIGPTConfig, OpenAIGPTModel),
        (GPT2Config, GPT2Model),
        (TransfoXLConfig, TransfoXLModel),
        (XLNetConfig, XLNetModel),
Lysandre's avatar
Lysandre committed
155
        (FlaubertConfig, FlaubertModel),
Julien Chaumond's avatar
Julien Chaumond committed
156
157
158
159
160
        (XLMConfig, XLMModel),
        (CTRLConfig, CTRLModel),
    ]
)

thomwolf's avatar
thomwolf committed
161
162
163
164
165
166
167
MODEL_FOR_PRETRAINING_MAPPING = OrderedDict(
    [
        (T5Config, T5WithLMHeadModel),
        (DistilBertConfig, DistilBertForMaskedLM),
        (AlbertConfig, AlbertForMaskedLM),
        (CamembertConfig, CamembertForMaskedLM),
        (XLMRobertaConfig, XLMRobertaForMaskedLM),
Sam Shleifer's avatar
Sam Shleifer committed
168
        (BartConfig, BartForMaskedLM),
thomwolf's avatar
thomwolf committed
169
170
171
172
173
174
        (RobertaConfig, RobertaForMaskedLM),
        (BertConfig, BertForPreTraining),
        (OpenAIGPTConfig, OpenAIGPTLMHeadModel),
        (GPT2Config, GPT2LMHeadModel),
        (TransfoXLConfig, TransfoXLLMHeadModel),
        (XLNetConfig, XLNetLMHeadModel),
Lysandre's avatar
Lysandre committed
175
        (FlaubertConfig, FlaubertWithLMHeadModel),
thomwolf's avatar
thomwolf committed
176
177
178
179
180
        (XLMConfig, XLMWithLMHeadModel),
        (CTRLConfig, CTRLLMHeadModel),
    ]
)

Julien Chaumond's avatar
Julien Chaumond committed
181
MODEL_WITH_LM_HEAD_MAPPING = OrderedDict(
182
183
184
185
186
187
    [
        (T5Config, T5WithLMHeadModel),
        (DistilBertConfig, DistilBertForMaskedLM),
        (AlbertConfig, AlbertForMaskedLM),
        (CamembertConfig, CamembertForMaskedLM),
        (XLMRobertaConfig, XLMRobertaForMaskedLM),
Sam Shleifer's avatar
Sam Shleifer committed
188
        (BartConfig, BartForMaskedLM),
189
        (RobertaConfig, RobertaForMaskedLM),
190
191
192
193
194
        (BertConfig, BertForMaskedLM),
        (OpenAIGPTConfig, OpenAIGPTLMHeadModel),
        (GPT2Config, GPT2LMHeadModel),
        (TransfoXLConfig, TransfoXLLMHeadModel),
        (XLNetConfig, XLNetLMHeadModel),
Lysandre's avatar
Lysandre committed
195
        (FlaubertConfig, FlaubertWithLMHeadModel),
196
197
198
199
200
        (XLMConfig, XLMWithLMHeadModel),
        (CTRLConfig, CTRLLMHeadModel),
    ]
)

Julien Chaumond's avatar
Julien Chaumond committed
201
MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
202
203
204
205
206
    [
        (DistilBertConfig, DistilBertForSequenceClassification),
        (AlbertConfig, AlbertForSequenceClassification),
        (CamembertConfig, CamembertForSequenceClassification),
        (XLMRobertaConfig, XLMRobertaForSequenceClassification),
Sam Shleifer's avatar
Sam Shleifer committed
207
        (BartConfig, BartForSequenceClassification),
208
        (RobertaConfig, RobertaForSequenceClassification),
209
210
        (BertConfig, BertForSequenceClassification),
        (XLNetConfig, XLNetForSequenceClassification),
Lysandre's avatar
Lysandre committed
211
        (FlaubertConfig, FlaubertForSequenceClassification),
Lysandre's avatar
Lysandre committed
212
        (XLMConfig, XLMForSequenceClassification),
213
214
215
    ]
)

Julien Chaumond's avatar
Julien Chaumond committed
216
MODEL_FOR_QUESTION_ANSWERING_MAPPING = OrderedDict(
217
218
219
    [
        (DistilBertConfig, DistilBertForQuestionAnswering),
        (AlbertConfig, AlbertForQuestionAnswering),
Malte Pietsch's avatar
Malte Pietsch committed
220
        (RobertaConfig, RobertaForQuestionAnswering),
221
222
        (BertConfig, BertForQuestionAnswering),
        (XLNetConfig, XLNetForQuestionAnswering),
Lysandre's avatar
Lysandre committed
223
        (FlaubertConfig, FlaubertForQuestionAnswering),
Lysandre's avatar
Lysandre committed
224
        (XLMConfig, XLMForQuestionAnswering),
225
226
227
    ]
)

Julien Chaumond's avatar
Julien Chaumond committed
228
MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
Julien Chaumond's avatar
Julien Chaumond committed
229
230
231
    [
        (DistilBertConfig, DistilBertForTokenClassification),
        (CamembertConfig, CamembertForTokenClassification),
232
        (XLMRobertaConfig, XLMRobertaForTokenClassification),
233
        (RobertaConfig, RobertaForTokenClassification),
Julien Chaumond's avatar
Julien Chaumond committed
234
235
236
237
238
        (BertConfig, BertForTokenClassification),
        (XLNetConfig, XLNetForTokenClassification),
    ]
)

239

thomwolf's avatar
thomwolf committed
240
241
class AutoModel(object):
    r"""
242
        :class:`~transformers.AutoModel` is a generic model class
thomwolf's avatar
thomwolf committed
243
244
        that will be instantiated as one of the base model classes of the library
        when created with the `AutoModel.from_pretrained(pretrained_model_name_or_path)`
245
        or the `AutoModel.from_config(config)` class methods.
thomwolf's avatar
thomwolf committed
246

247
        This class cannot be instantiated using `__init__()` (throws an error).
thomwolf's avatar
thomwolf committed
248
    """
249

thomwolf's avatar
thomwolf committed
250
    def __init__(self):
251
252
        raise EnvironmentError(
            "AutoModel is designed to be instantiated "
253
            "using the `AutoModel.from_pretrained(pretrained_model_name_or_path)` or "
254
255
            "`AutoModel.from_config(config)` methods."
        )
256
257
258
259
260
261

    @classmethod
    def from_config(cls, config):
        r""" Instantiates one of the base model classes of the library
        from a configuration.

Lysandre's avatar
Lysandre committed
262
263
        Args:
            config (:class:`~transformers.PretrainedConfig`):
264
                The model class to instantiate is selected based on the configuration class:
Lysandre's avatar
Lysandre committed
265
266
267
268
269
270
271
272
273
274

                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModel` (DistilBERT model)
                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModel` (RoBERTa model)
                - isInstance of `bert` configuration class: :class:`~transformers.BertModel` (Bert model)
                - isInstance of `openai-gpt` configuration class: :class:`~transformers.OpenAIGPTModel` (OpenAI GPT model)
                - isInstance of `gpt2` configuration class: :class:`~transformers.GPT2Model` (OpenAI GPT-2 model)
                - isInstance of `ctrl` configuration class: :class:`~transformers.CTRLModel` (Salesforce CTRL  model)
                - isInstance of `transfo-xl` configuration class: :class:`~transformers.TransfoXLModel` (Transformer-XL model)
                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetModel` (XLNet model)
                - isInstance of `xlm` configuration class: :class:`~transformers.XLMModel` (XLM model)
Lysandre's avatar
Lysandre committed
275
                - isInstance of `flaubert` configuration class: :class:`~transformers.FlaubertModel` (XLM model)
276
277
278
279
280
281

        Examples::

            config = BertConfig.from_pretrained('bert-base-uncased')    # Download configuration from S3 and cache.
            model = AutoModel.from_config(config)  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
        """
Julien Chaumond's avatar
Julien Chaumond committed
282
283
284
        for config_class, model_class in MODEL_MAPPING.items():
            if isinstance(config, config_class):
                return model_class(config)
285
286
287
288
289
290
        raise ValueError(
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__, cls.__name__, ", ".join(c.__name__ for c in MODEL_MAPPING.keys())
            )
        )
thomwolf's avatar
thomwolf committed
291
292
293

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
294
        r""" Instantiates one of the base model classes of the library
thomwolf's avatar
thomwolf committed
295
296
        from a pre-trained model configuration.

Lysandre's avatar
Lysandre committed
297
298
299
300
301
        The `from_pretrained()` method takes care of returning the correct model class instance
        based on the `model_type` property of the config object, or when it's missing,
        falling back to using pattern matching on the `pretrained_model_name_or_path` string.

        The base model class to instantiate is selected as the first pattern matching
thomwolf's avatar
thomwolf committed
302
        in the `pretrained_model_name_or_path` string (in the following order):
Lysandre's avatar
Lysandre committed
303
304
305
306
307
308
309
310
311
312
313
314
315
            - contains `t5`: :class:`~transformers.T5Model` (T5 model)
            - contains `distilbert`: :class:`~transformers.DistilBertModel` (DistilBERT model)
            - contains `albert`: :class:`~transformers.AlbertModel` (ALBERT model)
            - contains `camembert`: :class:`~transformers.CamembertModel` (CamemBERT model)
            - contains `xlm-roberta`: :class:`~transformers.XLMRobertaModel` (XLM-RoBERTa model)
            - contains `roberta`: :class:`~transformers.RobertaModel` (RoBERTa model)
            - contains `bert`: :class:`~transformers.BertModel` (Bert model)
            - contains `openai-gpt`: :class:`~transformers.OpenAIGPTModel` (OpenAI GPT model)
            - contains `gpt2`: :class:`~transformers.GPT2Model` (OpenAI GPT-2 model)
            - contains `transfo-xl`: :class:`~transformers.TransfoXLModel` (Transformer-XL model)
            - contains `xlnet`: :class:`~transformers.XLNetModel` (XLNet model)
            - contains `xlm`: :class:`~transformers.XLMModel` (XLM model)
            - contains `ctrl`: :class:`~transformers.CTRLModel` (Salesforce CTRL  model)
Lysandre's avatar
Lysandre committed
316
            - contains `flaubert`: :class:`~transformers.Flaubert` (Flaubert  model)
thomwolf's avatar
thomwolf committed
317

thomwolf's avatar
typos  
thomwolf committed
318
            The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
thomwolf's avatar
thomwolf committed
319
320
            To train the model, you should first set it back in training mode with `model.train()`

Lysandre's avatar
Lysandre committed
321
        Args:
thomwolf's avatar
thomwolf committed
322
323
324
            pretrained_model_name_or_path: either:

                - a string with the `shortcut name` of a pre-trained model to load from cache or download, e.g.: ``bert-base-uncased``.
325
                - a string with the `identifier name` of a pre-trained model that was user-uploaded to our S3, e.g.: ``dbmdz/bert-base-german-cased``.
326
                - a path to a `directory` containing model weights saved using :func:`~transformers.PreTrainedModel.save_pretrained`, e.g.: ``./my_model_directory/``.
thomwolf's avatar
thomwolf committed
327
328
329
330
331
                - a path or url to a `tensorflow index checkpoint file` (e.g. `./tf_model/model.ckpt.index`). In this case, ``from_tf`` should be set to True and a configuration object should be provided as ``config`` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.

            model_args: (`optional`) Sequence of positional arguments:
                All remaning positional arguments will be passed to the underlying model's ``__init__`` method

332
            config: (`optional`) instance of a class derived from :class:`~transformers.PretrainedConfig`:
thomwolf's avatar
thomwolf committed
333
334
335
                Configuration for the model to use instead of an automatically loaded configuation. Configuration can be automatically loaded when:

                - the model is a model provided by the library (loaded with the ``shortcut-name`` string of a pretrained model), or
336
                - the model was saved using :func:`~transformers.PreTrainedModel.save_pretrained` and is reloaded by suppling the save directory.
thomwolf's avatar
thomwolf committed
337
338
339
340
                - the model is loaded by suppling a local directory as ``pretrained_model_name_or_path`` and a configuration JSON file named `config.json` is found in the directory.

            state_dict: (`optional`) dict:
                an optional state dictionnary for the model to use instead of a state dictionary loaded from saved weights file.
thomwolf's avatar
typos  
thomwolf committed
341
                This option can be used if you want to create a model from a pretrained configuration but load your own weights.
342
                In this case though, you should check if using :func:`~transformers.PreTrainedModel.save_pretrained` and :func:`~transformers.PreTrainedModel.from_pretrained` is not a simpler option.
thomwolf's avatar
thomwolf committed
343
344

            cache_dir: (`optional`) string:
thomwolf's avatar
thomwolf committed
345
346
                Path to a directory in which a downloaded pre-trained model
                configuration should be cached if the standard cache should not be used.
thomwolf's avatar
thomwolf committed
347
348
349
350

            force_download: (`optional`) boolean, default False:
                Force to (re-)download the model weights and configuration files and override the cached versions if they exists.

351
352
353
            resume_download: (`optional`) boolean, default False:
                Do not delete incompletely recieved file. Attempt to resume the download if such a file exists.

thomwolf's avatar
thomwolf committed
354
355
356
357
358
359
360
361
362
363
364
            proxies: (`optional`) dict, default None:
                A dictionary of proxy servers to use by protocol or endpoint, e.g.: {'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}.
                The proxies are used on each request.

            output_loading_info: (`optional`) boolean:
                Set to ``True`` to also return a dictionnary containing missing keys, unexpected keys and error messages.

            kwargs: (`optional`) Remaining dictionary of keyword arguments:
                Can be used to update the configuration object (after it being loaded) and initiate the model. (e.g. ``output_attention=True``). Behave differently depending on whether a `config` is provided or automatically loaded:

                - If a configuration is provided with ``config``, ``**kwargs`` will be directly passed to the underlying model's ``__init__`` method (we assume all relevant updates to the configuration have already been done)
365
                - If a configuration is not provided, ``kwargs`` will be first passed to the configuration class initialization function (:func:`~transformers.PretrainedConfig.from_pretrained`). Each key of ``kwargs`` that corresponds to a configuration attribute will be used to override said attribute with the supplied ``kwargs`` value. Remaining keys that do not correspond to any configuration attribute will be passed to the underlying model's ``__init__`` function.
thomwolf's avatar
thomwolf committed
366
367
368

        Examples::

thomwolf's avatar
thomwolf committed
369
370
371
372
373
374
375
            model = AutoModel.from_pretrained('bert-base-uncased')    # Download model and configuration from S3 and cache.
            model = AutoModel.from_pretrained('./test/bert_model/')  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
            model = AutoModel.from_pretrained('bert-base-uncased', output_attention=True)  # Update configuration during loading
            assert model.config.output_attention == True
            # Loading from a TF checkpoint file instead of a PyTorch model (slower)
            config = AutoConfig.from_json_file('./tf_model/bert_tf_model_config.json')
            model = AutoModel.from_pretrained('./tf_model/bert_tf_checkpoint.ckpt.index', from_tf=True, config=config)
thomwolf's avatar
thomwolf committed
376
377

        """
378
379
380
381
        config = kwargs.pop("config", None)
        if not isinstance(config, PretrainedConfig):
            config = AutoConfig.from_pretrained(pretrained_model_name_or_path, **kwargs)

Julien Chaumond's avatar
Julien Chaumond committed
382
383
384
        for config_class, model_class in MODEL_MAPPING.items():
            if isinstance(config, config_class):
                return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
385
        raise ValueError(
386
387
388
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__, cls.__name__, ", ".join(c.__name__ for c in MODEL_MAPPING.keys())
389
390
            )
        )
391
392


thomwolf's avatar
thomwolf committed
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
class AutoModelForPreTraining(object):
    r"""
        :class:`~transformers.AutoModelForPreTraining` is a generic model class
        that will be instantiated as one of the model classes of the library -with the architecture used for pretraining this model– when created with the `AutoModelForPreTraining.from_pretrained(pretrained_model_name_or_path)`
        class method.

        This class cannot be instantiated using `__init__()` (throws an error).
    """

    def __init__(self):
        raise EnvironmentError(
            "AutoModelForPreTraining is designed to be instantiated "
            "using the `AutoModelForPreTraining.from_pretrained(pretrained_model_name_or_path)` or "
            "`AutoModelForPreTraining.from_config(config)` methods."
        )

    @classmethod
    def from_config(cls, config):
        r""" Instantiates one of the base model classes of the library
        from a configuration.

        Args:
            config (:class:`~transformers.PretrainedConfig`):
                The model class to instantiate is selected based on the configuration class:

                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForMaskedLM` (DistilBERT model)
                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForMaskedLM` (RoBERTa model)
                - isInstance of `bert` configuration class: :class:`~transformers.BertForPreTraining` (Bert model)
                - isInstance of `openai-gpt` configuration class: :class:`~transformers.OpenAIGPTLMHeadModel` (OpenAI GPT model)
                - isInstance of `gpt2` configuration class: :class:`~transformers.GPT2ModelLMHeadModel` (OpenAI GPT-2 model)
                - isInstance of `ctrl` configuration class: :class:`~transformers.CTRLModelLMHeadModel` (Salesforce CTRL  model)
                - isInstance of `transfo-xl` configuration class: :class:`~transformers.TransfoXLLMHeadModel` (Transformer-XL model)
                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetLMHeadModel` (XLNet model)
                - isInstance of `xlm` configuration class: :class:`~transformers.XLMWithLMHeadModel` (XLM model)
Lysandre's avatar
Lysandre committed
427
                - isInstance of `flaubert` configuration class: :class:`~transformers.FlaubertWithLMHeadModel` (Flaubert model)
thomwolf's avatar
thomwolf committed
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466

        Examples::

            config = BertConfig.from_pretrained('bert-base-uncased')    # Download configuration from S3 and cache.
            model = AutoModelForPreTraining.from_config(config)  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
        """
        for config_class, model_class in MODEL_FOR_PRETRAINING_MAPPING.items():
            if isinstance(config, config_class):
                return model_class(config)
        raise ValueError(
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__, cls.__name__, ", ".join(c.__name__ for c in MODEL_FOR_PRETRAINING_MAPPING.keys())
            )
        )

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
        r""" Instantiates one of the model classes of the library -with the architecture used for pretraining this model– from a pre-trained model configuration.

        The `from_pretrained()` method takes care of returning the correct model class instance
        based on the `model_type` property of the config object, or when it's missing,
        falling back to using pattern matching on the `pretrained_model_name_or_path` string.

        The model class to instantiate is selected as the first pattern matching
        in the `pretrained_model_name_or_path` string (in the following order):
            - contains `t5`: :class:`~transformers.T5ModelWithLMHead` (T5 model)
            - contains `distilbert`: :class:`~transformers.DistilBertForMaskedLM` (DistilBERT model)
            - contains `albert`: :class:`~transformers.AlbertForMaskedLM` (ALBERT model)
            - contains `camembert`: :class:`~transformers.CamembertForMaskedLM` (CamemBERT model)
            - contains `xlm-roberta`: :class:`~transformers.XLMRobertaForMaskedLM` (XLM-RoBERTa model)
            - contains `roberta`: :class:`~transformers.RobertaForMaskedLM` (RoBERTa model)
            - contains `bert`: :class:`~transformers.BertForPreTraining` (Bert model)
            - contains `openai-gpt`: :class:`~transformers.OpenAIGPTLMHeadModel` (OpenAI GPT model)
            - contains `gpt2`: :class:`~transformers.GPT2LMHeadModel` (OpenAI GPT-2 model)
            - contains `transfo-xl`: :class:`~transformers.TransfoXLLMHeadModel` (Transformer-XL model)
            - contains `xlnet`: :class:`~transformers.XLNetLMHeadModel` (XLNet model)
            - contains `xlm`: :class:`~transformers.XLMWithLMHeadModel` (XLM model)
            - contains `ctrl`: :class:`~transformers.CTRLLMHeadModel` (Salesforce CTRL model)
Lysandre's avatar
Lysandre committed
467
            - contains `flaubert`: :class:`~transformers.FlaubertWithLMHeadModel` (Flaubert model)
thomwolf's avatar
thomwolf committed
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544

        The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
        To train the model, you should first set it back in training mode with `model.train()`

        Args:
            pretrained_model_name_or_path:
                Either:

                - a string with the `shortcut name` of a pre-trained model to load from cache or download, e.g.: ``bert-base-uncased``.
                - a string with the `identifier name` of a pre-trained model that was user-uploaded to our S3, e.g.: ``dbmdz/bert-base-german-cased``.
                - a path to a `directory` containing model weights saved using :func:`~transformers.PreTrainedModel.save_pretrained`, e.g.: ``./my_model_directory/``.
                - a path or url to a `tensorflow index checkpoint file` (e.g. `./tf_model/model.ckpt.index`). In this case, ``from_tf`` should be set to True and a configuration object should be provided as ``config`` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.
            model_args: (`optional`) Sequence of positional arguments:
                All remaning positional arguments will be passed to the underlying model's ``__init__`` method
            config: (`optional`) instance of a class derived from :class:`~transformers.PretrainedConfig`:
                Configuration for the model to use instead of an automatically loaded configuation. Configuration can be automatically loaded when:

                - the model is a model provided by the library (loaded with the ``shortcut-name`` string of a pretrained model), or
                - the model was saved using :func:`~transformers.PreTrainedModel.save_pretrained` and is reloaded by suppling the save directory.
                - the model is loaded by suppling a local directory as ``pretrained_model_name_or_path`` and a configuration JSON file named `config.json` is found in the directory.

            state_dict: (`optional`) dict:
                an optional state dictionnary for the model to use instead of a state dictionary loaded from saved weights file.
                This option can be used if you want to create a model from a pretrained configuration but load your own weights.
                In this case though, you should check if using :func:`~transformers.PreTrainedModel.save_pretrained` and :func:`~transformers.PreTrainedModel.from_pretrained` is not a simpler option.
            cache_dir: (`optional`) string:
                Path to a directory in which a downloaded pre-trained model
                configuration should be cached if the standard cache should not be used.
            force_download: (`optional`) boolean, default False:
                Force to (re-)download the model weights and configuration files and override the cached versions if they exists.
            resume_download: (`optional`) boolean, default False:
                Do not delete incompletely received file. Attempt to resume the download if such a file exists.
            proxies: (`optional`) dict, default None:
                A dictionary of proxy servers to use by protocol or endpoint, e.g.: {'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}.
                The proxies are used on each request.
            output_loading_info: (`optional`) boolean:
                Set to ``True`` to also return a dictionnary containing missing keys, unexpected keys and error messages.
            kwargs: (`optional`) Remaining dictionary of keyword arguments:
                Can be used to update the configuration object (after it being loaded) and initiate the model.
                (e.g. ``output_attention=True``). Behave differently depending on whether a `config` is provided or
                automatically loaded:

                - If a configuration is provided with ``config``, ``**kwargs`` will be directly passed to the
                  underlying model's ``__init__`` method (we assume all relevant updates to the configuration have
                  already been done)
                - If a configuration is not provided, ``kwargs`` will be first passed to the configuration class
                  initialization function (:func:`~transformers.PretrainedConfig.from_pretrained`). Each key of
                  ``kwargs`` that corresponds to a configuration attribute will be used to override said attribute
                  with the supplied ``kwargs`` value. Remaining keys that do not correspond to any configuration
                  attribute will be passed to the underlying model's ``__init__`` function.

        Examples::

            model = AutoModelForPreTraining.from_pretrained('bert-base-uncased')    # Download model and configuration from S3 and cache.
            model = AutoModelForPreTraining.from_pretrained('./test/bert_model/')  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
            model = AutoModelForPreTraining.from_pretrained('bert-base-uncased', output_attention=True)  # Update configuration during loading
            assert model.config.output_attention == True
            # Loading from a TF checkpoint file instead of a PyTorch model (slower)
            config = AutoConfig.from_json_file('./tf_model/bert_tf_model_config.json')
            model = AutoModelForPreTraining.from_pretrained('./tf_model/bert_tf_checkpoint.ckpt.index', from_tf=True, config=config)

        """
        config = kwargs.pop("config", None)
        if not isinstance(config, PretrainedConfig):
            config = AutoConfig.from_pretrained(pretrained_model_name_or_path, **kwargs)

        for config_class, model_class in MODEL_FOR_PRETRAINING_MAPPING.items():
            if isinstance(config, config_class):
                return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
        raise ValueError(
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__, cls.__name__, ", ".join(c.__name__ for c in MODEL_FOR_PRETRAINING_MAPPING.keys())
            )
        )


545
546
class AutoModelWithLMHead(object):
    r"""
547
        :class:`~transformers.AutoModelWithLMHead` is a generic model class
548
549
550
551
552
553
        that will be instantiated as one of the language modeling model classes of the library
        when created with the `AutoModelWithLMHead.from_pretrained(pretrained_model_name_or_path)`
        class method.

        This class cannot be instantiated using `__init__()` (throws an error).
    """
554

555
    def __init__(self):
556
557
        raise EnvironmentError(
            "AutoModelWithLMHead is designed to be instantiated "
558
            "using the `AutoModelWithLMHead.from_pretrained(pretrained_model_name_or_path)` or "
559
560
            "`AutoModelWithLMHead.from_config(config)` methods."
        )
561
562
563
564
565
566

    @classmethod
    def from_config(cls, config):
        r""" Instantiates one of the base model classes of the library
        from a configuration.

Lysandre's avatar
Lysandre committed
567
568
        Args:
            config (:class:`~transformers.PretrainedConfig`):
569
                The model class to instantiate is selected based on the configuration class:
Lysandre's avatar
Lysandre committed
570

Lysandre's avatar
Lysandre committed
571
572
573
574
575
576
577
578
579
                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForMaskedLM` (DistilBERT model)
                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForMaskedLM` (RoBERTa model)
                - isInstance of `bert` configuration class: :class:`~transformers.BertModelForMaskedLM` (Bert model)
                - isInstance of `openai-gpt` configuration class: :class:`~transformers.OpenAIGPTLMHeadModel` (OpenAI GPT model)
                - isInstance of `gpt2` configuration class: :class:`~transformers.GPT2ModelLMHeadModel` (OpenAI GPT-2 model)
                - isInstance of `ctrl` configuration class: :class:`~transformers.CTRLModelLMHeadModel` (Salesforce CTRL  model)
                - isInstance of `transfo-xl` configuration class: :class:`~transformers.TransfoXLLMHeadModel` (Transformer-XL model)
                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetLMHeadModel` (XLNet model)
                - isInstance of `xlm` configuration class: :class:`~transformers.XLMWithLMHeadModel` (XLM model)
Lysandre's avatar
Lysandre committed
580
                - isInstance of `flaubert` configuration class: :class:`~transformers.FlaubertWithLMHeadModel` (Flaubert model)
581
582
583
584
585
586

        Examples::

            config = BertConfig.from_pretrained('bert-base-uncased')    # Download configuration from S3 and cache.
            model = AutoModelWithLMHead.from_config(config)  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
        """
587
588
589
590
591
592
593
594
595
        for config_class, model_class in MODEL_WITH_LM_HEAD_MAPPING.items():
            if isinstance(config, config_class):
                return model_class(config)
        raise ValueError(
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__, cls.__name__, ", ".join(c.__name__ for c in MODEL_WITH_LM_HEAD_MAPPING.keys())
            )
        )
596
597
598
599
600
601
602

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
        r""" Instantiates one of the language modeling model classes of the library
        from a pre-trained model configuration.

        The `from_pretrained()` method takes care of returning the correct model class instance
603
604
        based on the `model_type` property of the config object, or when it's missing,
        falling back to using pattern matching on the `pretrained_model_name_or_path` string.
605
606
607

        The model class to instantiate is selected as the first pattern matching
        in the `pretrained_model_name_or_path` string (in the following order):
Lysandre's avatar
Lysandre committed
608
609
610
611
612
613
614
615
616
617
618
619
620
            - contains `t5`: :class:`~transformers.T5ModelWithLMHead` (T5 model)
            - contains `distilbert`: :class:`~transformers.DistilBertForMaskedLM` (DistilBERT model)
            - contains `albert`: :class:`~transformers.AlbertForMaskedLM` (ALBERT model)
            - contains `camembert`: :class:`~transformers.CamembertForMaskedLM` (CamemBERT model)
            - contains `xlm-roberta`: :class:`~transformers.XLMRobertaForMaskedLM` (XLM-RoBERTa model)
            - contains `roberta`: :class:`~transformers.RobertaForMaskedLM` (RoBERTa model)
            - contains `bert`: :class:`~transformers.BertForMaskedLM` (Bert model)
            - contains `openai-gpt`: :class:`~transformers.OpenAIGPTLMHeadModel` (OpenAI GPT model)
            - contains `gpt2`: :class:`~transformers.GPT2LMHeadModel` (OpenAI GPT-2 model)
            - contains `transfo-xl`: :class:`~transformers.TransfoXLLMHeadModel` (Transformer-XL model)
            - contains `xlnet`: :class:`~transformers.XLNetLMHeadModel` (XLNet model)
            - contains `xlm`: :class:`~transformers.XLMWithLMHeadModel` (XLM model)
            - contains `ctrl`: :class:`~transformers.CTRLLMHeadModel` (Salesforce CTRL model)
Lysandre's avatar
Lysandre committed
621
            - contains `flaubert`: :class:`~transformers.FlaubertWithLMHeadModel` (Flaubert model)
622
623
624
625

        The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
        To train the model, you should first set it back in training mode with `model.train()`

Lysandre's avatar
Lysandre committed
626
627
628
        Args:
            pretrained_model_name_or_path:
                Either:
thomwolf's avatar
thomwolf committed
629
630

                - a string with the `shortcut name` of a pre-trained model to load from cache or download, e.g.: ``bert-base-uncased``.
631
                - a string with the `identifier name` of a pre-trained model that was user-uploaded to our S3, e.g.: ``dbmdz/bert-base-german-cased``.
632
                - a path to a `directory` containing model weights saved using :func:`~transformers.PreTrainedModel.save_pretrained`, e.g.: ``./my_model_directory/``.
thomwolf's avatar
thomwolf committed
633
634
635
                - a path or url to a `tensorflow index checkpoint file` (e.g. `./tf_model/model.ckpt.index`). In this case, ``from_tf`` should be set to True and a configuration object should be provided as ``config`` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.
            model_args: (`optional`) Sequence of positional arguments:
                All remaning positional arguments will be passed to the underlying model's ``__init__`` method
636
            config: (`optional`) instance of a class derived from :class:`~transformers.PretrainedConfig`:
thomwolf's avatar
thomwolf committed
637
638
639
                Configuration for the model to use instead of an automatically loaded configuation. Configuration can be automatically loaded when:

                - the model is a model provided by the library (loaded with the ``shortcut-name`` string of a pretrained model), or
640
                - the model was saved using :func:`~transformers.PreTrainedModel.save_pretrained` and is reloaded by suppling the save directory.
thomwolf's avatar
thomwolf committed
641
642
643
644
                - the model is loaded by suppling a local directory as ``pretrained_model_name_or_path`` and a configuration JSON file named `config.json` is found in the directory.

            state_dict: (`optional`) dict:
                an optional state dictionnary for the model to use instead of a state dictionary loaded from saved weights file.
645
                This option can be used if you want to create a model from a pretrained configuration but load your own weights.
646
                In this case though, you should check if using :func:`~transformers.PreTrainedModel.save_pretrained` and :func:`~transformers.PreTrainedModel.from_pretrained` is not a simpler option.
thomwolf's avatar
thomwolf committed
647
            cache_dir: (`optional`) string:
648
649
                Path to a directory in which a downloaded pre-trained model
                configuration should be cached if the standard cache should not be used.
thomwolf's avatar
thomwolf committed
650
651
            force_download: (`optional`) boolean, default False:
                Force to (re-)download the model weights and configuration files and override the cached versions if they exists.
652
            resume_download: (`optional`) boolean, default False:
Lysandre's avatar
Lysandre committed
653
                Do not delete incompletely received file. Attempt to resume the download if such a file exists.
thomwolf's avatar
thomwolf committed
654
655
656
657
658
659
            proxies: (`optional`) dict, default None:
                A dictionary of proxy servers to use by protocol or endpoint, e.g.: {'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}.
                The proxies are used on each request.
            output_loading_info: (`optional`) boolean:
                Set to ``True`` to also return a dictionnary containing missing keys, unexpected keys and error messages.
            kwargs: (`optional`) Remaining dictionary of keyword arguments:
Lysandre's avatar
Lysandre committed
660
661
662
663
664
665
666
667
668
669
670
671
                Can be used to update the configuration object (after it being loaded) and initiate the model.
                (e.g. ``output_attention=True``). Behave differently depending on whether a `config` is provided or
                automatically loaded:

                - If a configuration is provided with ``config``, ``**kwargs`` will be directly passed to the
                  underlying model's ``__init__`` method (we assume all relevant updates to the configuration have
                  already been done)
                - If a configuration is not provided, ``kwargs`` will be first passed to the configuration class
                  initialization function (:func:`~transformers.PretrainedConfig.from_pretrained`). Each key of
                  ``kwargs`` that corresponds to a configuration attribute will be used to override said attribute
                  with the supplied ``kwargs`` value. Remaining keys that do not correspond to any configuration
                  attribute will be passed to the underlying model's ``__init__`` function.
672
673
674
675
676
677
678
679
680
681
682
683

        Examples::

            model = AutoModelWithLMHead.from_pretrained('bert-base-uncased')    # Download model and configuration from S3 and cache.
            model = AutoModelWithLMHead.from_pretrained('./test/bert_model/')  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
            model = AutoModelWithLMHead.from_pretrained('bert-base-uncased', output_attention=True)  # Update configuration during loading
            assert model.config.output_attention == True
            # Loading from a TF checkpoint file instead of a PyTorch model (slower)
            config = AutoConfig.from_json_file('./tf_model/bert_tf_model_config.json')
            model = AutoModelWithLMHead.from_pretrained('./tf_model/bert_tf_checkpoint.ckpt.index', from_tf=True, config=config)

        """
684
685
686
687
        config = kwargs.pop("config", None)
        if not isinstance(config, PretrainedConfig):
            config = AutoConfig.from_pretrained(pretrained_model_name_or_path, **kwargs)

688
689
690
        for config_class, model_class in MODEL_WITH_LM_HEAD_MAPPING.items():
            if isinstance(config, config_class):
                return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
691
        raise ValueError(
692
693
694
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__, cls.__name__, ", ".join(c.__name__ for c in MODEL_WITH_LM_HEAD_MAPPING.keys())
695
696
            )
        )
697
698
699
700


class AutoModelForSequenceClassification(object):
    r"""
701
        :class:`~transformers.AutoModelForSequenceClassification` is a generic model class
702
703
704
705
706
707
        that will be instantiated as one of the sequence classification model classes of the library
        when created with the `AutoModelForSequenceClassification.from_pretrained(pretrained_model_name_or_path)`
        class method.

        This class cannot be instantiated using `__init__()` (throws an error).
    """
708

709
    def __init__(self):
710
711
        raise EnvironmentError(
            "AutoModelForSequenceClassification is designed to be instantiated "
712
            "using the `AutoModelForSequenceClassification.from_pretrained(pretrained_model_name_or_path)` or "
713
714
            "`AutoModelForSequenceClassification.from_config(config)` methods."
        )
715
716
717
718
719
720

    @classmethod
    def from_config(cls, config):
        r""" Instantiates one of the base model classes of the library
        from a configuration.

Lysandre's avatar
Lysandre committed
721
722
        Args:
            config (:class:`~transformers.PretrainedConfig`):
723
                The model class to instantiate is selected based on the configuration class:
Lysandre's avatar
Lysandre committed
724

Lysandre's avatar
Lysandre committed
725
726
727
728
729
730
731
732
                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForSequenceClassification` (DistilBERT model)
                - isInstance of `albert` configuration class: :class:`~transformers.AlbertModelForSequenceClassification` (ALBERT model)
                - isInstance of `camembert` configuration class: :class:`~transformers.CamembertModelForSequenceClassification` (CamemBERT model)
                - isInstance of `xlm roberta` configuration class: :class:`~transformers.XLMRobertaModelForSequenceClassification` (XLM-RoBERTa model)
                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForSequenceClassification` (RoBERTa model)
                - isInstance of `bert` configuration class: :class:`~transformers.BertModelForSequenceClassification` (Bert model)
                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetModelForSequenceClassification` (XLNet model)
                - isInstance of `xlm` configuration class: :class:`~transformers.XLMModelForSequenceClassification` (XLM model)
Lysandre's avatar
Lysandre committed
733
                - isInstance of `flaubert` configuration class: :class:`~transformers.FlaubertForSequenceClassification` (Flaubert model)
Lysandre's avatar
Lysandre committed
734

735
736
737
738
739
740

        Examples::

            config = BertConfig.from_pretrained('bert-base-uncased')    # Download configuration from S3 and cache.
            model = AutoModelForSequenceClassification.from_config(config)  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
        """
741
742
743
744
745
746
747
748
749
750
751
        for config_class, model_class in MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING.items():
            if isinstance(config, config_class):
                return model_class(config)
        raise ValueError(
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__,
                cls.__name__,
                ", ".join(c.__name__ for c in MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING.keys()),
            )
        )
752
753
754
755
756
757
758

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
        r""" Instantiates one of the sequence classification model classes of the library
        from a pre-trained model configuration.

        The `from_pretrained()` method takes care of returning the correct model class instance
759
760
        based on the `model_type` property of the config object, or when it's missing,
        falling back to using pattern matching on the `pretrained_model_name_or_path` string.
761
762
763

        The model class to instantiate is selected as the first pattern matching
        in the `pretrained_model_name_or_path` string (in the following order):
Lysandre's avatar
Lysandre committed
764
765
766
767
768
769
770
            - contains `distilbert`: :class:`~transformers.DistilBertForSequenceClassification` (DistilBERT model)
            - contains `albert`: :class:`~transformers.AlbertForSequenceClassification` (ALBERT model)
            - contains `camembert`: :class:`~transformers.CamembertForSequenceClassification` (CamemBERT model)
            - contains `xlm-roberta`: :class:`~transformers.XLMRobertaForSequenceClassification` (XLM-RoBERTa model)
            - contains `roberta`: :class:`~transformers.RobertaForSequenceClassification` (RoBERTa model)
            - contains `bert`: :class:`~transformers.BertForSequenceClassification` (Bert model)
            - contains `xlnet`: :class:`~transformers.XLNetForSequenceClassification` (XLNet model)
Lysandre's avatar
Lysandre committed
771
            - contains `flaubert`: :class:`~transformers.FlaubertForSequenceClassification` (Flaubert model)
772
773
774
775

        The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
        To train the model, you should first set it back in training mode with `model.train()`

Lysandre's avatar
Lysandre committed
776
        Args:
thomwolf's avatar
thomwolf committed
777
778
779
            pretrained_model_name_or_path: either:

                - a string with the `shortcut name` of a pre-trained model to load from cache or download, e.g.: ``bert-base-uncased``.
780
                - a string with the `identifier name` of a pre-trained model that was user-uploaded to our S3, e.g.: ``dbmdz/bert-base-german-cased``.
781
                - a path to a `directory` containing model weights saved using :func:`~transformers.PreTrainedModel.save_pretrained`, e.g.: ``./my_model_directory/``.
thomwolf's avatar
thomwolf committed
782
783
784
                - a path or url to a `tensorflow index checkpoint file` (e.g. `./tf_model/model.ckpt.index`). In this case, ``from_tf`` should be set to True and a configuration object should be provided as ``config`` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.

            model_args: (`optional`) Sequence of positional arguments:
Lysandre's avatar
Lysandre committed
785
                All remaining positional arguments will be passed to the underlying model's ``__init__`` method
thomwolf's avatar
thomwolf committed
786

787
            config: (`optional`) instance of a class derived from :class:`~transformers.PretrainedConfig`:
thomwolf's avatar
thomwolf committed
788
789
790
                Configuration for the model to use instead of an automatically loaded configuation. Configuration can be automatically loaded when:

                - the model is a model provided by the library (loaded with the ``shortcut-name`` string of a pretrained model), or
791
                - the model was saved using :func:`~transformers.PreTrainedModel.save_pretrained` and is reloaded by suppling the save directory.
thomwolf's avatar
thomwolf committed
792
793
794
795
                - the model is loaded by suppling a local directory as ``pretrained_model_name_or_path`` and a configuration JSON file named `config.json` is found in the directory.

            state_dict: (`optional`) dict:
                an optional state dictionnary for the model to use instead of a state dictionary loaded from saved weights file.
796
                This option can be used if you want to create a model from a pretrained configuration but load your own weights.
797
                In this case though, you should check if using :func:`~transformers.PreTrainedModel.save_pretrained` and :func:`~transformers.PreTrainedModel.from_pretrained` is not a simpler option.
thomwolf's avatar
thomwolf committed
798
799

            cache_dir: (`optional`) string:
800
801
                Path to a directory in which a downloaded pre-trained model
                configuration should be cached if the standard cache should not be used.
thomwolf's avatar
thomwolf committed
802
803
804
805

            force_download: (`optional`) boolean, default False:
                Force to (re-)download the model weights and configuration files and override the cached versions if they exists.

806
807
808
            resume_download: (`optional`) boolean, default False:
                Do not delete incompletely recieved file. Attempt to resume the download if such a file exists.

thomwolf's avatar
thomwolf committed
809
810
811
812
813
814
815
816
817
818
819
            proxies: (`optional`) dict, default None:
                A dictionary of proxy servers to use by protocol or endpoint, e.g.: {'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}.
                The proxies are used on each request.

            output_loading_info: (`optional`) boolean:
                Set to ``True`` to also return a dictionnary containing missing keys, unexpected keys and error messages.

            kwargs: (`optional`) Remaining dictionary of keyword arguments:
                Can be used to update the configuration object (after it being loaded) and initiate the model. (e.g. ``output_attention=True``). Behave differently depending on whether a `config` is provided or automatically loaded:

                - If a configuration is provided with ``config``, ``**kwargs`` will be directly passed to the underlying model's ``__init__`` method (we assume all relevant updates to the configuration have already been done)
820
                - If a configuration is not provided, ``kwargs`` will be first passed to the configuration class initialization function (:func:`~transformers.PretrainedConfig.from_pretrained`). Each key of ``kwargs`` that corresponds to a configuration attribute will be used to override said attribute with the supplied ``kwargs`` value. Remaining keys that do not correspond to any configuration attribute will be passed to the underlying model's ``__init__`` function.
821
822
823
824
825
826
827
828
829
830
831
832

        Examples::

            model = AutoModelForSequenceClassification.from_pretrained('bert-base-uncased')    # Download model and configuration from S3 and cache.
            model = AutoModelForSequenceClassification.from_pretrained('./test/bert_model/')  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
            model = AutoModelForSequenceClassification.from_pretrained('bert-base-uncased', output_attention=True)  # Update configuration during loading
            assert model.config.output_attention == True
            # Loading from a TF checkpoint file instead of a PyTorch model (slower)
            config = AutoConfig.from_json_file('./tf_model/bert_tf_model_config.json')
            model = AutoModelForSequenceClassification.from_pretrained('./tf_model/bert_tf_checkpoint.ckpt.index', from_tf=True, config=config)

        """
833
834
835
836
        config = kwargs.pop("config", None)
        if not isinstance(config, PretrainedConfig):
            config = AutoConfig.from_pretrained(pretrained_model_name_or_path, **kwargs)

837
838
839
        for config_class, model_class in MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING.items():
            if isinstance(config, config_class):
                return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
840
        raise ValueError(
841
842
843
844
845
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__,
                cls.__name__,
                ", ".join(c.__name__ for c in MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING.keys()),
846
847
            )
        )
848
849
850
851


class AutoModelForQuestionAnswering(object):
    r"""
852
        :class:`~transformers.AutoModelForQuestionAnswering` is a generic model class
853
854
855
856
857
858
        that will be instantiated as one of the question answering model classes of the library
        when created with the `AutoModelForQuestionAnswering.from_pretrained(pretrained_model_name_or_path)`
        class method.

        This class cannot be instantiated using `__init__()` (throws an error).
    """
859

860
    def __init__(self):
861
862
        raise EnvironmentError(
            "AutoModelForQuestionAnswering is designed to be instantiated "
863
            "using the `AutoModelForQuestionAnswering.from_pretrained(pretrained_model_name_or_path)` or "
864
865
            "`AutoModelForQuestionAnswering.from_config(config)` methods."
        )
866
867
868
869
870
871

    @classmethod
    def from_config(cls, config):
        r""" Instantiates one of the base model classes of the library
        from a configuration.

Lysandre's avatar
Lysandre committed
872
873
        Args:
            config (:class:`~transformers.PretrainedConfig`):
874
                The model class to instantiate is selected based on the configuration class:
Lysandre's avatar
Lysandre committed
875

Lysandre's avatar
Lysandre committed
876
877
878
879
880
                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForQuestionAnswering` (DistilBERT model)
                - isInstance of `albert` configuration class: :class:`~transformers.AlbertModelForQuestionAnswering` (ALBERT model)
                - isInstance of `bert` configuration class: :class:`~transformers.BertModelForQuestionAnswering` (Bert model)
                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetModelForQuestionAnswering` (XLNet model)
                - isInstance of `xlm` configuration class: :class:`~transformers.XLMModelForQuestionAnswering` (XLM model)
Lysandre's avatar
Lysandre committed
881
                - isInstance of `flaubert` configuration class: :class:`~transformers.FlaubertForQuestionAnswering` (XLM model)
882
883
884
885
886
887

        Examples::

            config = BertConfig.from_pretrained('bert-base-uncased')    # Download configuration from S3 and cache.
            model = AutoModelForSequenceClassification.from_config(config)  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
        """
888
889
890
891
892
893
894
895
896
897
898
899
        for config_class, model_class in MODEL_FOR_QUESTION_ANSWERING_MAPPING.items():
            if isinstance(config, config_class):
                return model_class(config)

        raise ValueError(
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__,
                cls.__name__,
                ", ".join(c.__name__ for c in MODEL_FOR_QUESTION_ANSWERING_MAPPING.keys()),
            )
        )
900
901
902
903
904
905
906

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
        r""" Instantiates one of the question answering model classes of the library
        from a pre-trained model configuration.

        The `from_pretrained()` method takes care of returning the correct model class instance
907
908
        based on the `model_type` property of the config object, or when it's missing,
        falling back to using pattern matching on the `pretrained_model_name_or_path` string.
909
910
911

        The model class to instantiate is selected as the first pattern matching
        in the `pretrained_model_name_or_path` string (in the following order):
Lysandre's avatar
Lysandre committed
912
913
914
915
916
            - contains `distilbert`: :class:`~transformers.DistilBertForQuestionAnswering` (DistilBERT model)
            - contains `albert`: :class:`~transformers.AlbertForQuestionAnswering` (ALBERT model)
            - contains `bert`: :class:`~transformers.BertForQuestionAnswering` (Bert model)
            - contains `xlnet`: :class:`~transformers.XLNetForQuestionAnswering` (XLNet model)
            - contains `xlm`: :class:`~transformers.XLMForQuestionAnswering` (XLM model)
Lysandre's avatar
Lysandre committed
917
            - contains `flaubert`: :class:`~transformers.FlaubertForQuestionAnswering` (XLM model)
918
919
920
921

        The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
        To train the model, you should first set it back in training mode with `model.train()`

Lysandre's avatar
Lysandre committed
922
        Args:
thomwolf's avatar
thomwolf committed
923
924
925
            pretrained_model_name_or_path: either:

                - a string with the `shortcut name` of a pre-trained model to load from cache or download, e.g.: ``bert-base-uncased``.
926
                - a string with the `identifier name` of a pre-trained model that was user-uploaded to our S3, e.g.: ``dbmdz/bert-base-german-cased``.
927
                - a path to a `directory` containing model weights saved using :func:`~transformers.PreTrainedModel.save_pretrained`, e.g.: ``./my_model_directory/``.
thomwolf's avatar
thomwolf committed
928
929
930
931
932
                - a path or url to a `tensorflow index checkpoint file` (e.g. `./tf_model/model.ckpt.index`). In this case, ``from_tf`` should be set to True and a configuration object should be provided as ``config`` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.

            model_args: (`optional`) Sequence of positional arguments:
                All remaning positional arguments will be passed to the underlying model's ``__init__`` method

933
            config: (`optional`) instance of a class derived from :class:`~transformers.PretrainedConfig`:
thomwolf's avatar
thomwolf committed
934
935
936
                Configuration for the model to use instead of an automatically loaded configuation. Configuration can be automatically loaded when:

                - the model is a model provided by the library (loaded with the ``shortcut-name`` string of a pretrained model), or
937
                - the model was saved using :func:`~transformers.PreTrainedModel.save_pretrained` and is reloaded by suppling the save directory.
thomwolf's avatar
thomwolf committed
938
939
940
941
                - the model is loaded by suppling a local directory as ``pretrained_model_name_or_path`` and a configuration JSON file named `config.json` is found in the directory.

            state_dict: (`optional`) dict:
                an optional state dictionnary for the model to use instead of a state dictionary loaded from saved weights file.
942
                This option can be used if you want to create a model from a pretrained configuration but load your own weights.
943
                In this case though, you should check if using :func:`~transformers.PreTrainedModel.save_pretrained` and :func:`~transformers.PreTrainedModel.from_pretrained` is not a simpler option.
thomwolf's avatar
thomwolf committed
944
945

            cache_dir: (`optional`) string:
946
947
                Path to a directory in which a downloaded pre-trained model
                configuration should be cached if the standard cache should not be used.
thomwolf's avatar
thomwolf committed
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962

            force_download: (`optional`) boolean, default False:
                Force to (re-)download the model weights and configuration files and override the cached versions if they exists.

            proxies: (`optional`) dict, default None:
                A dictionary of proxy servers to use by protocol or endpoint, e.g.: {'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}.
                The proxies are used on each request.

            output_loading_info: (`optional`) boolean:
                Set to ``True`` to also return a dictionnary containing missing keys, unexpected keys and error messages.

            kwargs: (`optional`) Remaining dictionary of keyword arguments:
                Can be used to update the configuration object (after it being loaded) and initiate the model. (e.g. ``output_attention=True``). Behave differently depending on whether a `config` is provided or automatically loaded:

                - If a configuration is provided with ``config``, ``**kwargs`` will be directly passed to the underlying model's ``__init__`` method (we assume all relevant updates to the configuration have already been done)
963
                - If a configuration is not provided, ``kwargs`` will be first passed to the configuration class initialization function (:func:`~transformers.PretrainedConfig.from_pretrained`). Each key of ``kwargs`` that corresponds to a configuration attribute will be used to override said attribute with the supplied ``kwargs`` value. Remaining keys that do not correspond to any configuration attribute will be passed to the underlying model's ``__init__`` function.
964
965
966
967
968
969
970
971
972
973
974
975

        Examples::

            model = AutoModelForQuestionAnswering.from_pretrained('bert-base-uncased')    # Download model and configuration from S3 and cache.
            model = AutoModelForQuestionAnswering.from_pretrained('./test/bert_model/')  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
            model = AutoModelForQuestionAnswering.from_pretrained('bert-base-uncased', output_attention=True)  # Update configuration during loading
            assert model.config.output_attention == True
            # Loading from a TF checkpoint file instead of a PyTorch model (slower)
            config = AutoConfig.from_json_file('./tf_model/bert_tf_model_config.json')
            model = AutoModelForQuestionAnswering.from_pretrained('./tf_model/bert_tf_checkpoint.ckpt.index', from_tf=True, config=config)

        """
976
977
978
979
        config = kwargs.pop("config", None)
        if not isinstance(config, PretrainedConfig):
            config = AutoConfig.from_pretrained(pretrained_model_name_or_path, **kwargs)

980
981
982
        for config_class, model_class in MODEL_FOR_QUESTION_ANSWERING_MAPPING.items():
            if isinstance(config, config_class):
                return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
983

984
        raise ValueError(
985
986
987
988
989
990
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__,
                cls.__name__,
                ", ".join(c.__name__ for c in MODEL_FOR_QUESTION_ANSWERING_MAPPING.keys()),
            )
991
        )
992
993
994


class AutoModelForTokenClassification:
Lysandre's avatar
Lysandre committed
995
996
997
998
999
1000
1001
1002
1003
    r"""
        :class:`~transformers.AutoModelForTokenClassification` is a generic model class
        that will be instantiated as one of the token classification model classes of the library
        when created with the `AutoModelForTokenClassification.from_pretrained(pretrained_model_name_or_path)`
        class method.

        This class cannot be instantiated using `__init__()` (throws an error).
    """

1004
    def __init__(self):
1005
1006
1007
1008
1009
        raise EnvironmentError(
            "AutoModelForTokenClassification is designed to be instantiated "
            "using the `AutoModelForTokenClassification.from_pretrained(pretrained_model_name_or_path)` or "
            "`AutoModelForTokenClassification.from_config(config)` methods."
        )
1010
1011
1012
1013
1014

    @classmethod
    def from_config(cls, config):
        r""" Instantiates one of the base model classes of the library
        from a configuration.
1015

Lysandre's avatar
Lysandre committed
1016
1017
        Args:
            config (:class:`~transformers.PretrainedConfig`):
1018
                The model class to instantiate is selected based on the configuration class:
Lysandre's avatar
Lysandre committed
1019

Lysandre's avatar
Lysandre committed
1020
1021
1022
1023
1024
1025
                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForTokenClassification` (DistilBERT model)
                - isInstance of `xlm roberta` configuration class: :class:`~transformers.XLMRobertaModelForTokenClassification` (XLMRoberta model)
                - isInstance of `bert` configuration class: :class:`~transformers.BertModelForTokenClassification` (Bert model)
                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetModelForTokenClassification` (XLNet model)
                - isInstance of `camembert` configuration class: :class:`~transformers.CamembertModelForTokenClassification` (Camembert model)
                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForTokenClassification` (Roberta model)
1026

1027
        Examples::
1028

1029
1030
1031
            config = BertConfig.from_pretrained('bert-base-uncased')    # Download configuration from S3 and cache.
            model = AutoModelForTokenClassification.from_config(config)  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
        """
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
        for config_class, model_class in MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING.items():
            if isinstance(config, config_class):
                return model_class(config)

        raise ValueError(
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__,
                cls.__name__,
                ", ".join(c.__name__ for c in MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING.keys()),
            )
        )
1044

1045
1046
1047
1048
1049
1050
    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
        r""" Instantiates one of the question answering model classes of the library
        from a pre-trained model configuration.

        The `from_pretrained()` method takes care of returning the correct model class instance
1051
1052
        based on the `model_type` property of the config object, or when it's missing,
        falling back to using pattern matching on the `pretrained_model_name_or_path` string.
1053
1054
1055

        The model class to instantiate is selected as the first pattern matching
        in the `pretrained_model_name_or_path` string (in the following order):
Lysandre's avatar
Lysandre committed
1056
1057
1058
1059
1060
1061
            - contains `distilbert`: :class:`~transformers.DistilBertForTokenClassification` (DistilBERT model)
            - contains `xlm-roberta`: :class:`~transformers.XLMRobertaForTokenClassification` (XLM-RoBERTa?Para model)
            - contains `camembert`: :class:`~transformers.CamembertForTokenClassification` (Camembert model)
            - contains `bert`: :class:`~transformers.BertForTokenClassification` (Bert model)
            - contains `xlnet`: :class:`~transformers.XLNetForTokenClassification` (XLNet model)
            - contains `roberta`: :class:`~transformers.RobertaForTokenClassification` (Roberta model)
1062
1063
1064
1065

        The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
        To train the model, you should first set it back in training mode with `model.train()`

Lysandre's avatar
Lysandre committed
1066
1067
1068
        Args:
            pretrained_model_name_or_path:
                Either:
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119

                - a string with the `shortcut name` of a pre-trained model to load from cache or download, e.g.: ``bert-base-uncased``.
                - a path to a `directory` containing model weights saved using :func:`~transformers.PreTrainedModel.save_pretrained`, e.g.: ``./my_model_directory/``.
                - a path or url to a `tensorflow index checkpoint file` (e.g. `./tf_model/model.ckpt.index`). In this case, ``from_tf`` should be set to True and a configuration object should be provided as ``config`` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.

            model_args: (`optional`) Sequence of positional arguments:
                All remaning positional arguments will be passed to the underlying model's ``__init__`` method

            config: (`optional`) instance of a class derived from :class:`~transformers.PretrainedConfig`:
                Configuration for the model to use instead of an automatically loaded configuation. Configuration can be automatically loaded when:

                - the model is a model provided by the library (loaded with the ``shortcut-name`` string of a pretrained model), or
                - the model was saved using :func:`~transformers.PreTrainedModel.save_pretrained` and is reloaded by suppling the save directory.
                - the model is loaded by suppling a local directory as ``pretrained_model_name_or_path`` and a configuration JSON file named `config.json` is found in the directory.

            state_dict: (`optional`) dict:
                an optional state dictionnary for the model to use instead of a state dictionary loaded from saved weights file.
                This option can be used if you want to create a model from a pretrained configuration but load your own weights.
                In this case though, you should check if using :func:`~transformers.PreTrainedModel.save_pretrained` and :func:`~transformers.PreTrainedModel.from_pretrained` is not a simpler option.

            cache_dir: (`optional`) string:
                Path to a directory in which a downloaded pre-trained model
                configuration should be cached if the standard cache should not be used.

            force_download: (`optional`) boolean, default False:
                Force to (re-)download the model weights and configuration files and override the cached versions if they exists.

            proxies: (`optional`) dict, default None:
                A dictionary of proxy servers to use by protocol or endpoint, e.g.: {'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}.
                The proxies are used on each request.

            output_loading_info: (`optional`) boolean:
                Set to ``True`` to also return a dictionnary containing missing keys, unexpected keys and error messages.

            kwargs: (`optional`) Remaining dictionary of keyword arguments:
                Can be used to update the configuration object (after it being loaded) and initiate the model. (e.g. ``output_attention=True``). Behave differently depending on whether a `config` is provided or automatically loaded:

                - If a configuration is provided with ``config``, ``**kwargs`` will be directly passed to the underlying model's ``__init__`` method (we assume all relevant updates to the configuration have already been done)
                - If a configuration is not provided, ``kwargs`` will be first passed to the configuration class initialization function (:func:`~transformers.PretrainedConfig.from_pretrained`). Each key of ``kwargs`` that corresponds to a configuration attribute will be used to override said attribute with the supplied ``kwargs`` value. Remaining keys that do not correspond to any configuration attribute will be passed to the underlying model's ``__init__`` function.

        Examples::

            model = AutoModelForTokenClassification.from_pretrained('bert-base-uncased')    # Download model and configuration from S3 and cache.
            model = AutoModelForTokenClassification.from_pretrained('./test/bert_model/')  # E.g. model was saved using `save_pretrained('./test/saved_model/')`
            model = AutoModelForTokenClassification.from_pretrained('bert-base-uncased', output_attention=True)  # Update configuration during loading
            assert model.config.output_attention == True
            # Loading from a TF checkpoint file instead of a PyTorch model (slower)
            config = AutoConfig.from_json_file('./tf_model/bert_tf_model_config.json')
            model = AutoModelForTokenClassification.from_pretrained('./tf_model/bert_tf_checkpoint.ckpt.index', from_tf=True, config=config)

        """
1120
1121
1122
1123
        config = kwargs.pop("config", None)
        if not isinstance(config, PretrainedConfig):
            config = AutoConfig.from_pretrained(pretrained_model_name_or_path, **kwargs)

1124
1125
1126
        for config_class, model_class in MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING.items():
            if isinstance(config, config_class):
                return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
1127

1128
        raise ValueError(
1129
1130
1131
1132
1133
            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
            "Model type should be one of {}.".format(
                config.__class__,
                cls.__name__,
                ", ".join(c.__name__ for c in MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING.keys()),
1134
1135
            )
        )