"tests/models/gpt2/test_tokenization_gpt2_tf.py" did not exist on "e8d448edcfe6348177f0e99774b51eeef8e50b62"
check_repo.py 40.5 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# coding=utf-8
# Copyright 2020 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.

16
17
18
import inspect
import os
import re
19
import warnings
20
from collections import OrderedDict
21
from difflib import get_close_matches
22
from pathlib import Path
23

24
from transformers import is_flax_available, is_tf_available, is_torch_available
25
from transformers.models.auto import get_values
Yih-Dar's avatar
Yih-Dar committed
26
from transformers.models.auto.configuration_auto import CONFIG_MAPPING_NAMES
27
28
29
30
from transformers.models.auto.feature_extraction_auto import FEATURE_EXTRACTOR_MAPPING_NAMES
from transformers.models.auto.image_processing_auto import IMAGE_PROCESSOR_MAPPING_NAMES
from transformers.models.auto.processing_auto import PROCESSOR_MAPPING_NAMES
from transformers.models.auto.tokenization_auto import TOKENIZER_MAPPING_NAMES
31
from transformers.utils import ENV_VARS_TRUE_VALUES, direct_transformers_import
32

33
34
35
36
37

# All paths are set with the intent you should run this script from the root of the repo with the command
# python utils/check_repo.py
PATH_TO_TRANSFORMERS = "src/transformers"
PATH_TO_TESTS = "tests"
38
PATH_TO_DOC = "docs/source/en"
39

40
41
# Update this list with models that are supposed to be private.
PRIVATE_MODELS = [
Jongjyh's avatar
Jongjyh committed
42
    "AltRobertaModel",
43
    "DPRSpanPredictor",
Daniel Stancl's avatar
Daniel Stancl committed
44
    "LongT5Stack",
Li-Huai (Allan) Lin's avatar
Li-Huai (Allan) Lin committed
45
    "RealmBertModel",
46
    "T5Stack",
47
    "MT5Stack",
48
    "SwitchTransformersStack",
49
    "TFDPRSpanPredictor",
50
51
    "MaskFormerSwinModel",
    "MaskFormerSwinPreTrainedModel",
52
53
    "BridgeTowerTextModel",
    "BridgeTowerVisionModel",
54
55
]

56
57
# Update this list for models that are not tested with a comment explaining the reason it should not be.
# Being in this list is an exception and should **not** be the rule.
58
IGNORE_NON_TESTED = PRIVATE_MODELS.copy() + [
59
    # models to ignore for not tested
NielsRogge's avatar
NielsRogge committed
60
    "Blip2QFormerModel",  # Building part of bigger (tested) model.
NielsRogge's avatar
NielsRogge committed
61
62
    "DetaEncoder",  # Building part of bigger (tested) model.
    "DetaDecoder",  # Building part of bigger (tested) model.
63
    "ErnieMForInformationExtraction",
64
65
    "GraphormerEncoder",  # Building part of bigger (tested) model.
    "GraphormerDecoderHead",  # Building part of bigger (tested) model.
NielsRogge's avatar
NielsRogge committed
66
    "CLIPSegDecoder",  # Building part of bigger (tested) model.
67
68
    "TableTransformerEncoder",  # Building part of bigger (tested) model.
    "TableTransformerDecoder",  # Building part of bigger (tested) model.
69
70
    "TimeSeriesTransformerEncoder",  # Building part of bigger (tested) model.
    "TimeSeriesTransformerDecoder",  # Building part of bigger (tested) model.
71
72
    "InformerEncoder",  # Building part of bigger (tested) model.
    "InformerDecoder",  # Building part of bigger (tested) model.
73
74
    "JukeboxVQVAE",  # Building part of bigger (tested) model.
    "JukeboxPrior",  # Building part of bigger (tested) model.
NielsRogge's avatar
NielsRogge committed
75
76
    "DeformableDetrEncoder",  # Building part of bigger (tested) model.
    "DeformableDetrDecoder",  # Building part of bigger (tested) model.
Younes Belkada's avatar
Younes Belkada committed
77
    "OPTDecoder",  # Building part of bigger (tested) model.
78
79
    "FlaxWhisperDecoder",  # Building part of bigger (tested) model.
    "FlaxWhisperEncoder",  # Building part of bigger (tested) model.
80
81
    "WhisperDecoder",  # Building part of bigger (tested) model.
    "WhisperEncoder",  # Building part of bigger (tested) model.
82
    "DecisionTransformerGPT2Model",  # Building part of bigger (tested) model.
NielsRogge's avatar
NielsRogge committed
83
    "SegformerDecodeHead",  # Building part of bigger (tested) model.
Gunjan Chhablani's avatar
Gunjan Chhablani committed
84
85
86
    "PLBartEncoder",  # Building part of bigger (tested) model.
    "PLBartDecoder",  # Building part of bigger (tested) model.
    "PLBartDecoderWrapper",  # Building part of bigger (tested) model.
Vasudev Gupta's avatar
Vasudev Gupta committed
87
88
89
    "BigBirdPegasusEncoder",  # Building part of bigger (tested) model.
    "BigBirdPegasusDecoder",  # Building part of bigger (tested) model.
    "BigBirdPegasusDecoderWrapper",  # Building part of bigger (tested) model.
NielsRogge's avatar
NielsRogge committed
90
91
92
    "DetrEncoder",  # Building part of bigger (tested) model.
    "DetrDecoder",  # Building part of bigger (tested) model.
    "DetrDecoderWrapper",  # Building part of bigger (tested) model.
93
94
    "ConditionalDetrEncoder",  # Building part of bigger (tested) model.
    "ConditionalDetrDecoder",  # Building part of bigger (tested) model.
Suraj Patil's avatar
Suraj Patil committed
95
96
    "M2M100Encoder",  # Building part of bigger (tested) model.
    "M2M100Decoder",  # Building part of bigger (tested) model.
Chan Woo Kim's avatar
Chan Woo Kim committed
97
    "MCTCTEncoder",  # Building part of bigger (tested) model.
wangpeng's avatar
wangpeng committed
98
    "MgpstrModel",  # Building part of bigger (tested) model.
Suraj Patil's avatar
Suraj Patil committed
99
100
    "Speech2TextEncoder",  # Building part of bigger (tested) model.
    "Speech2TextDecoder",  # Building part of bigger (tested) model.
Patrick von Platen's avatar
Patrick von Platen committed
101
102
    "LEDEncoder",  # Building part of bigger (tested) model.
    "LEDDecoder",  # Building part of bigger (tested) model.
103
    "BartDecoderWrapper",  # Building part of bigger (tested) model.
104
    "BartEncoder",  # Building part of bigger (tested) model.
105
    "BertLMHeadModel",  # Needs to be setup as decoder.
106
    "BlenderbotSmallEncoder",  # Building part of bigger (tested) model.
107
    "BlenderbotSmallDecoderWrapper",  # Building part of bigger (tested) model.
108
    "BlenderbotEncoder",  # Building part of bigger (tested) model.
109
    "BlenderbotDecoderWrapper",  # Building part of bigger (tested) model.
110
    "MBartEncoder",  # Building part of bigger (tested) model.
111
    "MBartDecoderWrapper",  # Building part of bigger (tested) model.
112
113
114
115
    "MegatronBertLMHeadModel",  # Building part of bigger (tested) model.
    "MegatronBertEncoder",  # Building part of bigger (tested) model.
    "MegatronBertDecoder",  # Building part of bigger (tested) model.
    "MegatronBertDecoderWrapper",  # Building part of bigger (tested) model.
StevenTang1998's avatar
StevenTang1998 committed
116
117
    "MvpDecoderWrapper",  # Building part of bigger (tested) model.
    "MvpEncoder",  # Building part of bigger (tested) model.
118
    "PegasusEncoder",  # Building part of bigger (tested) model.
119
    "PegasusDecoderWrapper",  # Building part of bigger (tested) model.
Jason Phang's avatar
Jason Phang committed
120
121
122
    "PegasusXEncoder",  # Building part of bigger (tested) model.
    "PegasusXDecoder",  # Building part of bigger (tested) model.
    "PegasusXDecoderWrapper",  # Building part of bigger (tested) model.
123
    "DPREncoder",  # Building part of bigger (tested) model.
124
    "ProphetNetDecoderWrapper",  # Building part of bigger (tested) model.
Li-Huai (Allan) Lin's avatar
Li-Huai (Allan) Lin committed
125
126
127
128
    "RealmBertModel",  # Building part of bigger (tested) model.
    "RealmReader",  # Not regular model.
    "RealmScorer",  # Not regular model.
    "RealmForOpenQA",  # Not regular model.
129
    "ReformerForMaskedLM",  # Needs to be setup as decoder.
130
    "Speech2Text2DecoderWrapper",  # Building part of bigger (tested) model.
Ratthachat (Jung)'s avatar
Ratthachat (Jung) committed
131
    "TFDPREncoder",  # Building part of bigger (tested) model.
132
133
    "TFElectraMainLayer",  # Building part of bigger (tested) model (should it be a TFPreTrainedModel ?)
    "TFRobertaForMultipleChoice",  # TODO: fix
134
    "TFRobertaPreLayerNormForMultipleChoice",  # TODO: fix
135
    "TrOCRDecoderWrapper",  # Building part of bigger (tested) model.
amyeroberts's avatar
amyeroberts committed
136
137
    "TFWhisperEncoder",  # Building part of bigger (tested) model.
    "TFWhisperDecoder",  # Building part of bigger (tested) model.
abhishek thakur's avatar
abhishek thakur committed
138
    "SeparableConv1D",  # Building part of bigger (tested) model.
139
    "FlaxBartForCausalLM",  # Building part of bigger (tested) model.
140
    "FlaxBertForCausalLM",  # Building part of bigger (tested) model. Tested implicitly through FlaxRobertaForCausalLM.
Younes Belkada's avatar
Younes Belkada committed
141
    "OPTDecoderWrapper",
142
    "TFSegformerDecodeHead",  # Not a regular model.
Jongjyh's avatar
Jongjyh committed
143
    "AltRobertaModel",  # Building part of bigger (tested) model.
Younes Belkada's avatar
Younes Belkada committed
144
    "BlipTextLMHeadModel",  # No need to test it as it is tested by BlipTextVision models
145
146
    "BridgeTowerTextModel",  # No need to test it as it is tested by BridgeTowerModel model.
    "BridgeTowerVisionModel",  # No need to test it as it is tested by BridgeTowerModel model.
147
148
149
150
151
152
153
154
155
156
157
158
    "SpeechT5Decoder",  # Building part of bigger (tested) model.
    "SpeechT5DecoderWithoutPrenet",  # Building part of bigger (tested) model.
    "SpeechT5DecoderWithSpeechPrenet",  # Building part of bigger (tested) model.
    "SpeechT5DecoderWithTextPrenet",  # Building part of bigger (tested) model.
    "SpeechT5Encoder",  # Building part of bigger (tested) model.
    "SpeechT5EncoderWithoutPrenet",  # Building part of bigger (tested) model.
    "SpeechT5EncoderWithSpeechPrenet",  # Building part of bigger (tested) model.
    "SpeechT5EncoderWithTextPrenet",  # Building part of bigger (tested) model.
    "SpeechT5SpeechDecoder",  # Building part of bigger (tested) model.
    "SpeechT5SpeechEncoder",  # Building part of bigger (tested) model.
    "SpeechT5TextDecoder",  # Building part of bigger (tested) model.
    "SpeechT5TextEncoder",  # Building part of bigger (tested) model.
159
160
161
162
163
]

# Update this list with test files that don't have a tester with a `all_model_classes` variable and which don't
# trigger the common tests.
TEST_FILES_WITH_NO_COMMON_TESTS = [
Yih-Dar's avatar
Yih-Dar committed
164
165
166
167
168
169
170
171
172
173
174
175
176
    "models/decision_transformer/test_modeling_decision_transformer.py",
    "models/camembert/test_modeling_camembert.py",
    "models/mt5/test_modeling_flax_mt5.py",
    "models/mbart/test_modeling_mbart.py",
    "models/mt5/test_modeling_mt5.py",
    "models/pegasus/test_modeling_pegasus.py",
    "models/camembert/test_modeling_tf_camembert.py",
    "models/mt5/test_modeling_tf_mt5.py",
    "models/xlm_roberta/test_modeling_tf_xlm_roberta.py",
    "models/xlm_roberta/test_modeling_flax_xlm_roberta.py",
    "models/xlm_prophetnet/test_modeling_xlm_prophetnet.py",
    "models/xlm_roberta/test_modeling_xlm_roberta.py",
    "models/vision_text_dual_encoder/test_modeling_vision_text_dual_encoder.py",
Matt's avatar
Matt committed
177
    "models/vision_text_dual_encoder/test_modeling_tf_vision_text_dual_encoder.py",
Yih-Dar's avatar
Yih-Dar committed
178
179
    "models/vision_text_dual_encoder/test_modeling_flax_vision_text_dual_encoder.py",
    "models/decision_transformer/test_modeling_decision_transformer.py",
180
181
]

182
183
# Update this list for models that are not in any of the auto MODEL_XXX_MAPPING. Being in this list is an exception and
# should **not** be the rule.
184
IGNORE_NON_AUTO_CONFIGURED = PRIVATE_MODELS.copy() + [
185
    # models to ignore for model xxx mapping
186
187
    "AlignTextModel",
    "AlignVisionModel",
188
189
190
191
    "ClapTextModel",
    "ClapTextModelWithProjection",
    "ClapAudioModel",
    "ClapAudioModelWithProjection",
NielsRogge's avatar
NielsRogge committed
192
193
194
    "Blip2ForConditionalGeneration",
    "Blip2QFormerModel",
    "Blip2VisionModel",
195
    "ErnieMForInformationExtraction",
196
    "GitVisionModel",
197
198
    "GraphormerModel",
    "GraphormerForGraphClassification",
Younes Belkada's avatar
Younes Belkada committed
199
200
201
202
203
204
    "BlipForConditionalGeneration",
    "BlipForImageTextRetrieval",
    "BlipForQuestionAnswering",
    "BlipVisionModel",
    "BlipTextLMHeadModel",
    "BlipTextModel",
NielsRogge's avatar
NielsRogge committed
205
    "Swin2SRForImageSuperResolution",
206
207
    "BridgeTowerForImageAndTextRetrieval",
    "BridgeTowerForMaskedLM",
208
    "BridgeTowerForContrastiveLearning",
NielsRogge's avatar
NielsRogge committed
209
210
211
    "CLIPSegForImageSegmentation",
    "CLIPSegVisionModel",
    "CLIPSegTextModel",
Matt's avatar
Matt committed
212
    "EsmForProteinFolding",
213
    "GPTSanJapaneseModel",
214
    "TimeSeriesTransformerForPrediction",
215
    "InformerForPrediction",
216
217
    "JukeboxVQVAE",
    "JukeboxPrior",
Jason Phang's avatar
Jason Phang committed
218
219
220
221
222
223
    "PegasusXEncoder",
    "PegasusXDecoder",
    "PegasusXDecoderWrapper",
    "PegasusXEncoder",
    "PegasusXDecoder",
    "PegasusXDecoderWrapper",
NielsRogge's avatar
NielsRogge committed
224
    "DPTForDepthEstimation",
225
    "DecisionTransformerGPT2Model",
NielsRogge's avatar
NielsRogge committed
226
    "GLPNForDepthEstimation",
NielsRogge's avatar
NielsRogge committed
227
228
    "ViltForImagesAndTextClassification",
    "ViltForImageAndTextRetrieval",
229
    "ViltForTokenClassification",
NielsRogge's avatar
NielsRogge committed
230
    "ViltForMaskedLM",
Suraj Patil's avatar
Suraj Patil committed
231
232
233
    "XGLMEncoder",
    "XGLMDecoder",
    "XGLMDecoderWrapper",
NielsRogge's avatar
NielsRogge committed
234
235
    "PerceiverForMultimodalAutoencoding",
    "PerceiverForOpticalFlow",
NielsRogge's avatar
NielsRogge committed
236
    "SegformerDecodeHead",
237
    "TFSegformerDecodeHead",
Kamal Raj's avatar
Kamal Raj committed
238
    "FlaxBeitForMaskedImageModeling",
Gunjan Chhablani's avatar
Gunjan Chhablani committed
239
240
241
    "PLBartEncoder",
    "PLBartDecoder",
    "PLBartDecoderWrapper",
NielsRogge's avatar
NielsRogge committed
242
    "BeitForMaskedImageModeling",
243
244
    "ChineseCLIPTextModel",
    "ChineseCLIPVisionModel",
Suraj Patil's avatar
Suraj Patil committed
245
    "CLIPTextModel",
246
    "CLIPTextModelWithProjection",
Suraj Patil's avatar
Suraj Patil committed
247
    "CLIPVisionModel",
248
    "CLIPVisionModelWithProjection",
249
250
    "GroupViTTextModel",
    "GroupViTVisionModel",
Yih-Dar's avatar
Yih-Dar committed
251
252
    "TFCLIPTextModel",
    "TFCLIPVisionModel",
253
254
    "TFGroupViTTextModel",
    "TFGroupViTVisionModel",
Suraj Patil's avatar
Suraj Patil committed
255
256
    "FlaxCLIPTextModel",
    "FlaxCLIPVisionModel",
257
    "FlaxWav2Vec2ForCTC",
NielsRogge's avatar
NielsRogge committed
258
    "DetrForSegmentation",
259
    "ConditionalDetrForSegmentation",
260
261
    "DPRReader",
    "FlaubertForQuestionAnswering",
262
263
264
265
    "FlavaImageCodebook",
    "FlavaTextModel",
    "FlavaImageModel",
    "FlavaMultimodalModel",
266
    "GPT2DoubleHeadsModel",
267
    "GPTSw3DoubleHeadsModel",
268
    "LayoutLMForQuestionAnswering",
Ryokan RI's avatar
Ryokan RI committed
269
    "LukeForMaskedLM",
NielsRogge's avatar
NielsRogge committed
270
271
272
    "LukeForEntityClassification",
    "LukeForEntityPairClassification",
    "LukeForEntitySpanClassification",
wangpeng's avatar
wangpeng committed
273
    "MgpstrModel",
274
    "OpenAIGPTDoubleHeadsModel",
275
276
277
    "OwlViTTextModel",
    "OwlViTVisionModel",
    "OwlViTForObjectDetection",
278
279
280
    "RagModel",
    "RagSequenceForGeneration",
    "RagTokenForGeneration",
Li-Huai (Allan) Lin's avatar
Li-Huai (Allan) Lin committed
281
282
283
284
    "RealmEmbedder",
    "RealmForOpenQA",
    "RealmScorer",
    "RealmReader",
Ratthachat (Jung)'s avatar
Ratthachat (Jung) committed
285
    "TFDPRReader",
286
    "TFGPT2DoubleHeadsModel",
287
    "TFLayoutLMForQuestionAnswering",
288
    "TFOpenAIGPTDoubleHeadsModel",
Ratthachat (Jung)'s avatar
Ratthachat (Jung) committed
289
290
291
    "TFRagModel",
    "TFRagSequenceForGeneration",
    "TFRagTokenForGeneration",
292
    "Wav2Vec2ForCTC",
Patrick von Platen's avatar
Patrick von Platen committed
293
    "HubertForCTC",
294
295
    "SEWForCTC",
    "SEWDForCTC",
296
297
    "XLMForQuestionAnswering",
    "XLNetForQuestionAnswering",
abhishek thakur's avatar
abhishek thakur committed
298
    "SeparableConv1D",
Gunjan Chhablani's avatar
Gunjan Chhablani committed
299
300
301
302
    "VisualBertForRegionToPhraseAlignment",
    "VisualBertForVisualReasoning",
    "VisualBertForQuestionAnswering",
    "VisualBertForMultipleChoice",
Will Rice's avatar
Will Rice committed
303
    "TFWav2Vec2ForCTC",
Will Rice's avatar
Will Rice committed
304
    "TFHubertForCTC",
NielsRogge's avatar
NielsRogge committed
305
306
    "XCLIPVisionModel",
    "XCLIPTextModel",
Jongjyh's avatar
Jongjyh committed
307
308
309
    "AltCLIPTextModel",
    "AltCLIPVisionModel",
    "AltRobertaModel",
Zineng Tang's avatar
Zineng Tang committed
310
    "TvltForAudioVisualClassification",
311
312
313
    "SpeechT5ForSpeechToSpeech",
    "SpeechT5ForTextToSpeech",
    "SpeechT5HifiGan",
314
315
]

316
317
318
319
320
321
# Update this list for models that have multiple model types for the same
# model doc
MODEL_TYPE_TO_DOC_MAPPING = OrderedDict(
    [
        ("data2vec-text", "data2vec"),
        ("data2vec-audio", "data2vec"),
322
        ("data2vec-vision", "data2vec"),
NielsRogge's avatar
NielsRogge committed
323
        ("donut-swin", "donut"),
324
325
326
327
    ]
)


328
# This is to make sure the transformers module imported is the one in the repo.
329
transformers = direct_transformers_import(PATH_TO_TRANSFORMERS)
330
331


332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
def check_missing_backends():
    missing_backends = []
    if not is_torch_available():
        missing_backends.append("PyTorch")
    if not is_tf_available():
        missing_backends.append("TensorFlow")
    if not is_flax_available():
        missing_backends.append("Flax")
    if len(missing_backends) > 0:
        missing = ", ".join(missing_backends)
        if os.getenv("TRANSFORMERS_IS_CI", "").upper() in ENV_VARS_TRUE_VALUES:
            raise Exception(
                "Full repo consistency checks require all backends to be installed (with `pip install -e .[dev]` in the "
                f"Transformers repo, the following are missing: {missing}."
            )
        else:
            warnings.warn(
                "Full repo consistency checks require all backends to be installed (with `pip install -e .[dev]` in the "
                f"Transformers repo, the following are missing: {missing}. While it's probably fine as long as you "
                "didn't make any change in one of those backends modeling files, you should probably execute the "
                "command above to be on the safe side."
            )


356
357
358
359
360
361
362
363
364
365
366
367
368
def check_model_list():
    """Check the model list inside the transformers library."""
    # Get the models from the directory structure of `src/transformers/models/`
    models_dir = os.path.join(PATH_TO_TRANSFORMERS, "models")
    _models = []
    for model in os.listdir(models_dir):
        model_dir = os.path.join(models_dir, model)
        if os.path.isdir(model_dir) and "__init__.py" in os.listdir(model_dir):
            _models.append(model)

    # Get the models from the directory structure of `src/transformers/models/`
    models = [model for model in dir(transformers.models) if not model.startswith("__")]

369
    missing_models = sorted(set(_models).difference(models))
370
371
372
373
374
375
    if missing_models:
        raise Exception(
            f"The following models should be included in {models_dir}/__init__.py: {','.join(missing_models)}."
        )


376
377
378
# If some modeling modules should be ignored for all checks, they should be added in the nested list
# _ignore_modules of this function.
def get_model_modules():
Patrick von Platen's avatar
Patrick von Platen committed
379
    """Get the model modules inside the transformers library."""
380
381
382
383
384
385
386
387
    _ignore_modules = [
        "modeling_auto",
        "modeling_encoder_decoder",
        "modeling_marian",
        "modeling_mmbt",
        "modeling_outputs",
        "modeling_retribert",
        "modeling_utils",
Sylvain Gugger's avatar
Sylvain Gugger committed
388
        "modeling_flax_auto",
389
        "modeling_flax_encoder_decoder",
Stas Bekman's avatar
Stas Bekman committed
390
        "modeling_flax_utils",
391
        "modeling_speech_encoder_decoder",
392
        "modeling_flax_speech_encoder_decoder",
393
        "modeling_flax_vision_encoder_decoder",
394
395
        "modeling_transfo_xl_utilities",
        "modeling_tf_auto",
396
        "modeling_tf_encoder_decoder",
397
398
399
400
        "modeling_tf_outputs",
        "modeling_tf_pytorch_utils",
        "modeling_tf_utils",
        "modeling_tf_transfo_xl_utilities",
401
        "modeling_tf_vision_encoder_decoder",
402
        "modeling_vision_encoder_decoder",
403
404
    ]
    modules = []
Sylvain Gugger's avatar
Sylvain Gugger committed
405
406
407
408
409
410
411
412
413
    for model in dir(transformers.models):
        # There are some magic dunder attributes in the dir, we ignore them
        if not model.startswith("__"):
            model_module = getattr(transformers.models, model)
            for submodule in dir(model_module):
                if submodule.startswith("modeling") and submodule not in _ignore_modules:
                    modeling_module = getattr(model_module, submodule)
                    if inspect.ismodule(modeling_module):
                        modules.append(modeling_module)
414
415
416
    return modules


417
def get_models(module, include_pretrained=False):
Patrick von Platen's avatar
Patrick von Platen committed
418
    """Get the objects in module that are models."""
419
    models = []
420
    model_classes = (transformers.PreTrainedModel, transformers.TFPreTrainedModel, transformers.FlaxPreTrainedModel)
421
    for attr_name in dir(module):
422
        if not include_pretrained and ("Pretrained" in attr_name or "PreTrained" in attr_name):
423
424
425
426
427
428
429
            continue
        attr = getattr(module, attr_name)
        if isinstance(attr, type) and issubclass(attr, model_classes) and attr.__module__ == module.__name__:
            models.append((attr_name, attr))
    return models


430
431
432
433
434
435
436
437
438
439
440
441
def is_a_private_model(model):
    """Returns True if the model should not be in the main init."""
    if model in PRIVATE_MODELS:
        return True

    # Wrapper, Encoder and Decoder are all privates
    if model.endswith("Wrapper"):
        return True
    if model.endswith("Encoder"):
        return True
    if model.endswith("Decoder"):
        return True
442
443
    if model.endswith("Prenet"):
        return True
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
    return False


def check_models_are_in_init():
    """Checks all models defined in the library are in the main init."""
    models_not_in_init = []
    dir_transformers = dir(transformers)
    for module in get_model_modules():
        models_not_in_init += [
            model[0] for model in get_models(module, include_pretrained=True) if model[0] not in dir_transformers
        ]

    # Remove private models
    models_not_in_init = [model for model in models_not_in_init if not is_a_private_model(model)]
    if len(models_not_in_init) > 0:
        raise Exception(f"The following models should be in the main init: {','.join(models_not_in_init)}.")


462
463
464
# If some test_modeling files should be ignored when checking models are all tested, they should be added in the
# nested list _ignore_files of this function.
def get_model_test_files():
Yih-Dar's avatar
Yih-Dar committed
465
466
467
468
469
470
    """Get the model test files.

    The returned files should NOT contain the `tests` (i.e. `PATH_TO_TESTS` defined in this script). They will be
    considered as paths relative to `tests`. A caller has to use `os.path.join(PATH_TO_TESTS, ...)` to access the files.
    """

471
472
473
    _ignore_files = [
        "test_modeling_common",
        "test_modeling_encoder_decoder",
474
        "test_modeling_flax_encoder_decoder",
475
        "test_modeling_flax_speech_encoder_decoder",
476
477
        "test_modeling_marian",
        "test_modeling_tf_common",
478
        "test_modeling_tf_encoder_decoder",
479
480
    ]
    test_files = []
Yih-Dar's avatar
Yih-Dar committed
481
482
483
484
485
486
487
488
489
490
491
492
493
    # Check both `PATH_TO_TESTS` and `PATH_TO_TESTS/models`
    model_test_root = os.path.join(PATH_TO_TESTS, "models")
    model_test_dirs = []
    for x in os.listdir(model_test_root):
        x = os.path.join(model_test_root, x)
        if os.path.isdir(x):
            model_test_dirs.append(x)

    for target_dir in [PATH_TO_TESTS] + model_test_dirs:
        for file_or_dir in os.listdir(target_dir):
            path = os.path.join(target_dir, file_or_dir)
            if os.path.isfile(path):
                filename = os.path.split(path)[-1]
494
                if "test_modeling" in filename and os.path.splitext(filename)[0] not in _ignore_files:
Yih-Dar's avatar
Yih-Dar committed
495
496
497
                    file = os.path.join(*path.split(os.sep)[1:])
                    test_files.append(file)

498
499
500
501
502
503
    return test_files


# This is a bit hacky but I didn't find a way to import the test_file as a module and read inside the tester class
# for the all_model_classes variable.
def find_tested_models(test_file):
Patrick von Platen's avatar
Patrick von Platen committed
504
    """Parse the content of test_file to detect what's in all_model_classes"""
Sylvain Gugger's avatar
Sylvain Gugger committed
505
    # This is a bit hacky but I didn't find a way to import the test_file as a module and read inside the class
506
    with open(os.path.join(PATH_TO_TESTS, test_file), "r", encoding="utf-8", newline="\n") as f:
507
        content = f.read()
Sylvain Gugger's avatar
Sylvain Gugger committed
508
    all_models = re.findall(r"all_model_classes\s+=\s+\(\s*\(([^\)]*)\)", content)
509
510
    # Check with one less parenthesis as well
    all_models += re.findall(r"all_model_classes\s+=\s+\(([^\)]*)\)", content)
Sylvain Gugger's avatar
Sylvain Gugger committed
511
    if len(all_models) > 0:
512
        model_tested = []
Sylvain Gugger's avatar
Sylvain Gugger committed
513
514
515
516
517
        for entry in all_models:
            for line in entry.split(","):
                name = line.strip()
                if len(name) > 0:
                    model_tested.append(name)
518
519
520
521
        return model_tested


def check_models_are_tested(module, test_file):
Patrick von Platen's avatar
Patrick von Platen committed
522
    """Check models defined in module are tested in test_file."""
523
    # XxxPreTrainedModel are not tested
524
525
526
    defined_models = get_models(module)
    tested_models = find_tested_models(test_file)
    if tested_models is None:
527
        if test_file.replace(os.path.sep, "/") in TEST_FILES_WITH_NO_COMMON_TESTS:
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
            return
        return [
            f"{test_file} should define `all_model_classes` to apply common tests to the models it tests. "
            + "If this intentional, add the test filename to `TEST_FILES_WITH_NO_COMMON_TESTS` in the file "
            + "`utils/check_repo.py`."
        ]
    failures = []
    for model_name, _ in defined_models:
        if model_name not in tested_models and model_name not in IGNORE_NON_TESTED:
            failures.append(
                f"{model_name} is defined in {module.__name__} but is not tested in "
                + f"{os.path.join(PATH_TO_TESTS, test_file)}. Add it to the all_model_classes in that file."
                + "If common tests should not applied to that model, add its name to `IGNORE_NON_TESTED`"
                + "in the file `utils/check_repo.py`."
            )
    return failures


def check_all_models_are_tested():
Patrick von Platen's avatar
Patrick von Platen committed
547
    """Check all models are properly tested."""
548
549
550
551
    modules = get_model_modules()
    test_files = get_model_test_files()
    failures = []
    for module in modules:
552
553
        test_file = [file for file in test_files if f"test_{module.__name__.split('.')[-1]}.py" in file]
        if len(test_file) == 0:
554
            failures.append(f"{module.__name__} does not have its corresponding test file {test_file}.")
555
556
557
558
        elif len(test_file) > 1:
            failures.append(f"{module.__name__} has several test files: {test_file}.")
        else:
            test_file = test_file[0]
559
560
561
            new_failures = check_models_are_tested(module, test_file)
            if new_failures is not None:
                failures += new_failures
562
563
564
565
    if len(failures) > 0:
        raise Exception(f"There were {len(failures)} failures:\n" + "\n".join(failures))


566
def get_all_auto_configured_models():
Patrick von Platen's avatar
Patrick von Platen committed
567
    """Return the list of all models in at least one auto class."""
568
    result = set()  # To avoid duplicates we concatenate all model classes in a set.
569
570
    if is_torch_available():
        for attr_name in dir(transformers.models.auto.modeling_auto):
571
            if attr_name.startswith("MODEL_") and attr_name.endswith("MAPPING_NAMES"):
572
573
574
                result = result | set(get_values(getattr(transformers.models.auto.modeling_auto, attr_name)))
    if is_tf_available():
        for attr_name in dir(transformers.models.auto.modeling_tf_auto):
575
            if attr_name.startswith("TF_MODEL_") and attr_name.endswith("MAPPING_NAMES"):
576
577
578
                result = result | set(get_values(getattr(transformers.models.auto.modeling_tf_auto, attr_name)))
    if is_flax_available():
        for attr_name in dir(transformers.models.auto.modeling_flax_auto):
579
            if attr_name.startswith("FLAX_MODEL_") and attr_name.endswith("MAPPING_NAMES"):
580
                result = result | set(get_values(getattr(transformers.models.auto.modeling_flax_auto, attr_name)))
581
    return list(result)
582
583


584
585
586
587
588
589
590
591
592
593
594
def ignore_unautoclassed(model_name):
    """Rules to determine if `name` should be in an auto class."""
    # Special white list
    if model_name in IGNORE_NON_AUTO_CONFIGURED:
        return True
    # Encoder and Decoder should be ignored
    if "Encoder" in model_name or "Decoder" in model_name:
        return True
    return False


595
def check_models_are_auto_configured(module, all_auto_models):
Patrick von Platen's avatar
Patrick von Platen committed
596
    """Check models defined in module are each in an auto class."""
597
598
599
    defined_models = get_models(module)
    failures = []
    for model_name, _ in defined_models:
600
        if model_name not in all_auto_models and not ignore_unautoclassed(model_name):
601
602
603
604
605
606
607
608
609
            failures.append(
                f"{model_name} is defined in {module.__name__} but is not present in any of the auto mapping. "
                "If that is intended behavior, add its name to `IGNORE_NON_AUTO_CONFIGURED` in the file "
                "`utils/check_repo.py`."
            )
    return failures


def check_all_models_are_auto_configured():
Patrick von Platen's avatar
Patrick von Platen committed
610
    """Check all models are each in an auto class."""
611
    check_missing_backends()
612
613
614
615
616
617
618
619
620
621
622
    modules = get_model_modules()
    all_auto_models = get_all_auto_configured_models()
    failures = []
    for module in modules:
        new_failures = check_models_are_auto_configured(module, all_auto_models)
        if new_failures is not None:
            failures += new_failures
    if len(failures) > 0:
        raise Exception(f"There were {len(failures)} failures:\n" + "\n".join(failures))


623
624
def check_all_auto_object_names_being_defined():
    """Check all names defined in auto (name) mappings exist in the library."""
625
    check_missing_backends()
626

627
    failures = []
628
    mappings_to_check = {
629
630
631
632
633
634
        "TOKENIZER_MAPPING_NAMES": TOKENIZER_MAPPING_NAMES,
        "IMAGE_PROCESSOR_MAPPING_NAMES": IMAGE_PROCESSOR_MAPPING_NAMES,
        "FEATURE_EXTRACTOR_MAPPING_NAMES": FEATURE_EXTRACTOR_MAPPING_NAMES,
        "PROCESSOR_MAPPING_NAMES": PROCESSOR_MAPPING_NAMES,
    }

635
636
    # Each auto modeling files contains multiple mappings. Let's get them in a dynamic way.
    for module_name in ["modeling_auto", "modeling_tf_auto", "modeling_flax_auto"]:
637
638
639
        module = getattr(transformers.models.auto, module_name, None)
        if module is None:
            continue
640
641
642
643
644
        # all mappings in a single auto modeling file
        mapping_names = [x for x in dir(module) if x.endswith("_MAPPING_NAMES")]
        mappings_to_check.update({name: getattr(module, name) for name in mapping_names})

    for name, mapping in mappings_to_check.items():
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
        for model_type, class_names in mapping.items():
            if not isinstance(class_names, tuple):
                class_names = (class_names,)
                for class_name in class_names:
                    if class_name is None:
                        continue
                    # dummy object is accepted
                    if not hasattr(transformers, class_name):
                        # If the class name is in a model name mapping, let's not check if there is a definition in any modeling
                        # module, if it's a private model defined in this file.
                        if name.endswith("MODEL_MAPPING_NAMES") and is_a_private_model(class_name):
                            continue
                        failures.append(
                            f"`{class_name}` appears in the mapping `{name}` but it is not defined in the library."
                        )
    if len(failures) > 0:
        raise Exception(f"There were {len(failures)} failures:\n" + "\n".join(failures))


Yih-Dar's avatar
Yih-Dar committed
664
665
def check_all_auto_mapping_names_in_config_mapping_names():
    """Check all keys defined in auto mappings (mappings of names) appear in `CONFIG_MAPPING_NAMES`."""
666
    check_missing_backends()
Yih-Dar's avatar
Yih-Dar committed
667

668
    failures = []
Yih-Dar's avatar
Yih-Dar committed
669
    # `TOKENIZER_PROCESSOR_MAPPING_NAMES` and `AutoTokenizer` is special, and don't need to follow the rule.
670
    mappings_to_check = {
Yih-Dar's avatar
Yih-Dar committed
671
672
673
674
675
        "IMAGE_PROCESSOR_MAPPING_NAMES": IMAGE_PROCESSOR_MAPPING_NAMES,
        "FEATURE_EXTRACTOR_MAPPING_NAMES": FEATURE_EXTRACTOR_MAPPING_NAMES,
        "PROCESSOR_MAPPING_NAMES": PROCESSOR_MAPPING_NAMES,
    }

676
677
    # Each auto modeling files contains multiple mappings. Let's get them in a dynamic way.
    for module_name in ["modeling_auto", "modeling_tf_auto", "modeling_flax_auto"]:
678
679
680
        module = getattr(transformers.models.auto, module_name, None)
        if module is None:
            continue
681
682
683
684
685
        # all mappings in a single auto modeling file
        mapping_names = [x for x in dir(module) if x.endswith("_MAPPING_NAMES")]
        mappings_to_check.update({name: getattr(module, name) for name in mapping_names})

    for name, mapping in mappings_to_check.items():
Yih-Dar's avatar
Yih-Dar committed
686
687
688
689
690
691
692
693
694
695
        for model_type, class_names in mapping.items():
            if model_type not in CONFIG_MAPPING_NAMES:
                failures.append(
                    f"`{model_type}` appears in the mapping `{name}` but it is not defined in the keys of "
                    "`CONFIG_MAPPING_NAMES`."
                )
    if len(failures) > 0:
        raise Exception(f"There were {len(failures)} failures:\n" + "\n".join(failures))


Sylvain Gugger's avatar
Sylvain Gugger committed
696
697
698
699
_re_decorator = re.compile(r"^\s*@(\S+)\s+$")


def check_decorator_order(filename):
Patrick von Platen's avatar
Patrick von Platen committed
700
    """Check that in the test file `filename` the slow decorator is always last."""
701
    with open(filename, "r", encoding="utf-8", newline="\n") as f:
Sylvain Gugger's avatar
Sylvain Gugger committed
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
        lines = f.readlines()
    decorator_before = None
    errors = []
    for i, line in enumerate(lines):
        search = _re_decorator.search(line)
        if search is not None:
            decorator_name = search.groups()[0]
            if decorator_before is not None and decorator_name.startswith("parameterized"):
                errors.append(i)
            decorator_before = decorator_name
        elif decorator_before is not None:
            decorator_before = None
    return errors


def check_all_decorator_order():
Patrick von Platen's avatar
Patrick von Platen committed
718
    """Check that in all test files, the slow decorator is always last."""
Sylvain Gugger's avatar
Sylvain Gugger committed
719
720
721
722
723
724
725
726
727
    errors = []
    for fname in os.listdir(PATH_TO_TESTS):
        if fname.endswith(".py"):
            filename = os.path.join(PATH_TO_TESTS, fname)
            new_errors = check_decorator_order(filename)
            errors += [f"- {filename}, line {i}" for i in new_errors]
    if len(errors) > 0:
        msg = "\n".join(errors)
        raise ValueError(
Sylvain Gugger's avatar
Sylvain Gugger committed
728
729
            "The parameterized decorator (and its variants) should always be first, but this is not the case in the"
            f" following files:\n{msg}"
Sylvain Gugger's avatar
Sylvain Gugger committed
730
731
732
        )


733
def find_all_documented_objects():
Patrick von Platen's avatar
Patrick von Platen committed
734
    """Parse the content of all doc files to detect which classes and functions it documents"""
735
736
    documented_obj = []
    for doc_file in Path(PATH_TO_DOC).glob("**/*.rst"):
Julien Plu's avatar
Julien Plu committed
737
        with open(doc_file, "r", encoding="utf-8", newline="\n") as f:
738
739
740
            content = f.read()
        raw_doc_objs = re.findall(r"(?:autoclass|autofunction):: transformers.(\S+)\s+", content)
        documented_obj += [obj.split(".")[-1] for obj in raw_doc_objs]
Sylvain Gugger's avatar
Sylvain Gugger committed
741
742
743
744
745
    for doc_file in Path(PATH_TO_DOC).glob("**/*.mdx"):
        with open(doc_file, "r", encoding="utf-8", newline="\n") as f:
            content = f.read()
        raw_doc_objs = re.findall("\[\[autodoc\]\]\s+(\S+)\s+", content)
        documented_obj += [obj.split(".")[-1] for obj in raw_doc_objs]
746
747
748
749
750
751
    return documented_obj


# One good reason for not being documented is to be deprecated. Put in this list deprecated objects.
DEPRECATED_OBJECTS = [
    "AutoModelWithLMHead",
752
    "BartPretrainedModel",
753
754
    "DataCollator",
    "DataCollatorForSOP",
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
    "GlueDataset",
    "GlueDataTrainingArguments",
    "LineByLineTextDataset",
    "LineByLineWithRefDataset",
    "LineByLineWithSOPTextDataset",
    "PretrainedBartModel",
    "PretrainedFSMTModel",
    "SingleSentenceClassificationProcessor",
    "SquadDataTrainingArguments",
    "SquadDataset",
    "SquadExample",
    "SquadFeatures",
    "SquadV1Processor",
    "SquadV2Processor",
    "TFAutoModelWithLMHead",
770
    "TFBartPretrainedModel",
771
772
    "TextDataset",
    "TextDatasetForNextSentencePrediction",
773
    "Wav2Vec2ForMaskedLM",
774
    "Wav2Vec2Tokenizer",
775
776
777
778
779
780
781
782
783
784
    "glue_compute_metrics",
    "glue_convert_examples_to_features",
    "glue_output_modes",
    "glue_processors",
    "glue_tasks_num_labels",
    "squad_convert_examples_to_features",
    "xnli_compute_metrics",
    "xnli_output_modes",
    "xnli_processors",
    "xnli_tasks_num_labels",
785
786
    "TFTrainer",
    "TFTrainingArguments",
787
788
789
790
791
792
793
]

# Exceptionally, some objects should not be documented after all rules passed.
# ONLY PUT SOMETHING IN THIS LIST AS A LAST RESORT!
UNDOCUMENTED_OBJECTS = [
    "AddedToken",  # This is a tokenizers class.
    "BasicTokenizer",  # Internal, should never have been in the main init.
794
    "CharacterTokenizer",  # Internal, should never have been in the main init.
795
    "DPRPretrainedReader",  # Like an Encoder.
Sylvain Gugger's avatar
Sylvain Gugger committed
796
    "DummyObject",  # Just picked by mistake sometimes.
797
    "MecabTokenizer",  # Internal, should never have been in the main init.
798
799
800
801
802
803
804
805
806
807
808
    "ModelCard",  # Internal type.
    "SqueezeBertModule",  # Internal building block (should have been called SqueezeBertLayer)
    "TFDPRPretrainedReader",  # Like an Encoder.
    "TransfoXLCorpus",  # Internal type.
    "WordpieceTokenizer",  # Internal, should never have been in the main init.
    "absl",  # External module
    "add_end_docstrings",  # Internal, should never have been in the main init.
    "add_start_docstrings",  # Internal, should never have been in the main init.
    "convert_tf_weight_name_to_pt_weight_name",  # Internal used to convert model weights
    "logger",  # Internal logger
    "logging",  # External module
809
    "requires_backends",  # Internal function
Jongjyh's avatar
Jongjyh committed
810
    "AltRobertaModel",  # Internal module
811
812
813
814
815
816
817
818
819
]

# This list should be empty. Objects in it should get their own doc page.
SHOULD_HAVE_THEIR_OWN_PAGE = [
    # Benchmarks
    "PyTorchBenchmark",
    "PyTorchBenchmarkArguments",
    "TensorFlowBenchmark",
    "TensorFlowBenchmarkArguments",
820
    "AutoBackbone",
NielsRogge's avatar
NielsRogge committed
821
822
    "BitBackbone",
    "ConvNextBackbone",
823
    "DinatBackbone",
NielsRogge's avatar
NielsRogge committed
824
    "MaskFormerSwinBackbone",
825
826
    "MaskFormerSwinConfig",
    "MaskFormerSwinModel",
NielsRogge's avatar
NielsRogge committed
827
828
    "NatBackbone",
    "ResNetBackbone",
NielsRogge's avatar
NielsRogge committed
829
    "SwinBackbone",
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
]


def ignore_undocumented(name):
    """Rules to determine if `name` should be undocumented."""
    # NOT DOCUMENTED ON PURPOSE.
    # Constants uppercase are not documented.
    if name.isupper():
        return True
    # PreTrainedModels / Encoders / Decoders / Layers / Embeddings / Attention are not documented.
    if (
        name.endswith("PreTrainedModel")
        or name.endswith("Decoder")
        or name.endswith("Encoder")
        or name.endswith("Layer")
        or name.endswith("Embeddings")
        or name.endswith("Attention")
    ):
        return True
    # Submodules are not documented.
    if os.path.isdir(os.path.join(PATH_TO_TRANSFORMERS, name)) or os.path.isfile(
        os.path.join(PATH_TO_TRANSFORMERS, f"{name}.py")
    ):
        return True
    # All load functions are not documented.
    if name.startswith("load_tf") or name.startswith("load_pytorch"):
        return True
    # is_xxx_available functions are not documented.
    if name.startswith("is_") and name.endswith("_available"):
        return True
    # Deprecated objects are not documented.
    if name in DEPRECATED_OBJECTS or name in UNDOCUMENTED_OBJECTS:
        return True
    # MMBT model does not really work.
    if name.startswith("MMBT"):
        return True
    if name in SHOULD_HAVE_THEIR_OWN_PAGE:
        return True
    return False


def check_all_objects_are_documented():
Patrick von Platen's avatar
Patrick von Platen committed
872
    """Check all models are properly documented."""
873
    documented_objs = find_all_documented_objects()
874
875
876
    modules = transformers._modules
    objects = [c for c in dir(transformers) if c not in modules and not c.startswith("_")]
    undocumented_objs = [c for c in objects if c not in documented_objs and not ignore_undocumented(c)]
877
878
879
880
881
    if len(undocumented_objs) > 0:
        raise Exception(
            "The following objects are in the public init so should be documented:\n - "
            + "\n - ".join(undocumented_objs)
        )
882
    check_docstrings_are_in_md()
883
884
885
886
887
888
889
890
891
    check_model_type_doc_match()


def check_model_type_doc_match():
    """Check all doc pages have a corresponding model type."""
    model_doc_folder = Path(PATH_TO_DOC) / "model_doc"
    model_docs = [m.stem for m in model_doc_folder.glob("*.mdx")]

    model_types = list(transformers.models.auto.configuration_auto.MODEL_NAMES_MAPPING.keys())
892
    model_types = [MODEL_TYPE_TO_DOC_MAPPING[m] if m in MODEL_TYPE_TO_DOC_MAPPING else m for m in model_types]
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910

    errors = []
    for m in model_docs:
        if m not in model_types and m != "auto":
            close_matches = get_close_matches(m, model_types)
            error_message = f"{m} is not a proper model identifier."
            if len(close_matches) > 0:
                close_matches = "/".join(close_matches)
                error_message += f" Did you mean {close_matches}?"
            errors.append(error_message)

    if len(errors) > 0:
        raise ValueError(
            "Some model doc pages do not match any existing model type:\n"
            + "\n".join(errors)
            + "\nYou can add any missing model type to the `MODEL_NAMES_MAPPING` constant in "
            "models/auto/configuration_auto.py."
        )
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937


# Re pattern to catch :obj:`xx`, :class:`xx`, :func:`xx` or :meth:`xx`.
_re_rst_special_words = re.compile(r":(?:obj|func|class|meth):`([^`]+)`")
# Re pattern to catch things between double backquotes.
_re_double_backquotes = re.compile(r"(^|[^`])``([^`]+)``([^`]|$)")
# Re pattern to catch example introduction.
_re_rst_example = re.compile(r"^\s*Example.*::\s*$", flags=re.MULTILINE)


def is_rst_docstring(docstring):
    """
    Returns `True` if `docstring` is written in rst.
    """
    if _re_rst_special_words.search(docstring) is not None:
        return True
    if _re_double_backquotes.search(docstring) is not None:
        return True
    if _re_rst_example.search(docstring) is not None:
        return True
    return False


def check_docstrings_are_in_md():
    """Check all docstrings are in md"""
    files_with_rst = []
    for file in Path(PATH_TO_TRANSFORMERS).glob("**/*.py"):
938
        with open(file, encoding="utf-8") as f:
939
940
941
942
943
944
945
946
947
948
949
950
951
            code = f.read()
        docstrings = code.split('"""')

        for idx, docstring in enumerate(docstrings):
            if idx % 2 == 0 or not is_rst_docstring(docstring):
                continue
            files_with_rst.append(file)
            break

    if len(files_with_rst) > 0:
        raise ValueError(
            "The following files have docstrings written in rst:\n"
            + "\n".join([f"- {f}" for f in files_with_rst])
Kamal Raj's avatar
Kamal Raj committed
952
            + "\nTo fix this run `doc-builder convert path_to_py_file` after installing `doc-builder`\n"
953
954
            "(`pip install git+https://github.com/huggingface/doc-builder`)"
        )
955
956


957
def check_repo_quality():
Patrick von Platen's avatar
Patrick von Platen committed
958
    """Check all models are properly tested and documented."""
959
960
    print("Checking all models are included.")
    check_model_list()
961
962
    print("Checking all models are public.")
    check_models_are_in_init()
963
    print("Checking all models are properly tested.")
Sylvain Gugger's avatar
Sylvain Gugger committed
964
    check_all_decorator_order()
965
    check_all_models_are_tested()
966
    print("Checking all objects are properly documented.")
967
    check_all_objects_are_documented()
968
969
    print("Checking all models are in at least one auto class.")
    check_all_models_are_auto_configured()
970
971
    print("Checking all names in auto name mappings are defined.")
    check_all_auto_object_names_being_defined()
Yih-Dar's avatar
Yih-Dar committed
972
973
    print("Checking all keys in auto name mappings are defined in `CONFIG_MAPPING_NAMES`.")
    check_all_auto_mapping_names_in_config_mapping_names()
974
975
976
977


if __name__ == "__main__":
    check_repo_quality()