Reorganize repo (#8580)

* Put models in subfolders

* Styling

* Fix imports in tests

* More fixes in test imports

* Sneaky hidden imports

* Fix imports in doc files

* More sneaky imports

* Finish fixing tests

* Fix examples

* Fix path for copies

* More fixes for examples

* Fix dummy files

* More fixes for example

* More model import fixes

* Is this why you're unhappy GitHub?

* Fix imports in conver command

.github/workflows/github-torch-hub.yml

@@ -8,6 +8,9 @@ on:
 jobs:
   torch_hub_integration:
     runs-on: ubuntu-latest
+    env:
+      # TODO quickfix but may need more investigation
+      ACTIONS_ALLOW_UNSECURE_COMMANDS: True
     steps:
     # no checkout necessary here.
     - name: Extract branch name

.gitignore

@@ -133,7 +133,6 @@ dmypy.json
 tensorflow_code
 
 # Models
-models
 proc_data
 
 # examples

docs/source/model_doc/albert.rst

@@ -51,10 +51,10 @@ AlbertTokenizer
 Albert specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_albert.AlbertForPreTrainingOutput
+.. autoclass:: transformers.models.albert.modeling_albert.AlbertForPreTrainingOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_albert.TFAlbertForPreTrainingOutput
+.. autoclass:: transformers.models.albert.modeling_tf_albert.TFAlbertForPreTrainingOutput
     :members:
 
 

docs/source/model_doc/bart.rst

@@ -44,8 +44,8 @@ Implementation Notes
 - Bart doesn't use :obj:`token_type_ids` for sequence classification. Use :class:`~transformers.BartTokenizer` or
   :meth:`~transformers.BartTokenizer.encode` to get the proper splitting.
 - The forward pass of :class:`~transformers.BartModel` will create decoder inputs (using the helper function
-  :func:`transformers.modeling_bart._prepare_bart_decoder_inputs`) if they are not passed. This is different than some
+  :func:`transformers.models.bart.modeling_bart._prepare_bart_decoder_inputs`) if they are not passed. This is
-  other modeling APIs.
+  different than some other modeling APIs.
 - Model predictions are intended to be identical to the original implementation when
   :obj:`force_bos_token_to_be_generated=True`. This only works, however, if the string you pass to
   :func:`fairseq.encode` starts with a space.
@@ -93,7 +93,7 @@ BartModel
 .. autoclass:: transformers.BartModel
     :members: forward
 
-.. autofunction:: transformers.modeling_bart._prepare_bart_decoder_inputs
+.. autofunction:: transformers.models.bart.modeling_bart._prepare_bart_decoder_inputs
 
 
 BartForConditionalGeneration

docs/source/model_doc/bert.rst

@@ -57,10 +57,10 @@ BertTokenizerFast
 Bert specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_bert.BertForPreTrainingOutput
+.. autoclass:: transformers.models.bert.modeling_bert.BertForPreTrainingOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_bert.TFBertForPreTrainingOutput
+.. autoclass:: transformers.models.bert.modeling_tf_bert.TFBertForPreTrainingOutput
     :members:
 
 

docs/source/model_doc/dpr.rst

@@ -71,13 +71,13 @@ DPRReaderTokenizerFast
 DPR specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_dpr.DPRContextEncoderOutput
+.. autoclass:: transformers.models.dpr.modeling_dpr.DPRContextEncoderOutput
     :members:
 
-.. autoclass:: transformers.modeling_dpr.DPRQuestionEncoderOutput
+.. autoclass:: transformers.models.dpr.modeling_dpr.DPRQuestionEncoderOutput
     :members:
 
-.. autoclass:: transformers.modeling_dpr.DPRReaderOutput
+.. autoclass:: transformers.models.dpr.modeling_dpr.DPRReaderOutput
     :members:
 
 

docs/source/model_doc/electra.rst

@@ -69,10 +69,10 @@ ElectraTokenizerFast
 Electra specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_electra.ElectraForPreTrainingOutput
+.. autoclass:: transformers.models.electra.modeling_electra.ElectraForPreTrainingOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_electra.TFElectraForPreTrainingOutput
+.. autoclass:: transformers.models.electra.modeling_tf_electra.TFElectraForPreTrainingOutput
     :members:
 
 

docs/source/model_doc/funnel.rst

@@ -65,10 +65,10 @@ FunnelTokenizerFast
 Funnel specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_funnel.FunnelForPreTrainingOutput
+.. autoclass:: transformers.models.funnel.modeling_funnel.FunnelForPreTrainingOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_funnel.TFFunnelForPreTrainingOutput
+.. autoclass:: transformers.models.funnel.modeling_tf_funnel.TFFunnelForPreTrainingOutput
     :members:
 
 

docs/source/model_doc/gpt.rst

@@ -72,10 +72,10 @@ OpenAIGPTTokenizerFast
 OpenAI specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_openai.OpenAIGPTDoubleHeadsModelOutput
+.. autoclass:: transformers.models.openai.modeling_openai.OpenAIGPTDoubleHeadsModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_openai.TFOpenAIGPTDoubleHeadsModelOutput
+.. autoclass:: transformers.models.openai.modeling_tf_openai.TFOpenAIGPTDoubleHeadsModelOutput
     :members:
 
 

docs/source/model_doc/gpt2.rst

@@ -60,10 +60,10 @@ GPT2TokenizerFast
 GPT2 specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_gpt2.GPT2DoubleHeadsModelOutput
+.. autoclass:: transformers.models.gpt2.modeling_gpt2.GPT2DoubleHeadsModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_gpt2.TFGPT2DoubleHeadsModelOutput
+.. autoclass:: transformers.models.gpt2.modeling_tf_gpt2.TFGPT2DoubleHeadsModelOutput
     :members:
 
 

docs/source/model_doc/longformer.rst

@@ -93,29 +93,27 @@ LongformerTokenizerFast
 Longformer specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_longformer.LongformerBaseModelOutput
+.. autoclass:: transformers.models.longformer.modeling_longformer.LongformerBaseModelOutput
     :members: 
 
-.. autoclass:: transformers.modeling_longformer.LongformerBaseModelOutputWithPooling
+.. autoclass:: transformers.models.longformer.modeling_longformer.LongformerBaseModelOutputWithPooling
     :members: 
 
-.. autoclass:: transformers.modeling_longformer.LongformerMultipleChoiceModelOutput
+.. autoclass:: transformers.models.longformer.modeling_longformer.LongformerMultipleChoiceModelOutput
     :members: 
 
-.. autoclass:: transformers.modeling_longformer.LongformerQuestionAnsweringModelOutput
+.. autoclass:: transformers.models.longformer.modeling_longformer.LongformerQuestionAnsweringModelOutput
     :members: 
 
-.. autoclass:: transformers.modeling_tf_longformer.TFLongformerBaseModelOutput
+.. autoclass:: transformers.models.longformer.modeling_tf_longformer.TFLongformerBaseModelOutput
     :members: 
 
-.. autoclass:: transformers.modeling_tf_longformer.TFLongformerBaseModelOutputWithPooling
+.. autoclass:: transformers.models.longformer.modeling_tf_longformer.TFLongformerBaseModelOutputWithPooling
     :members: 
 
-.. autoclass:: transformers.modeling_tf_longformer.TFLongformerQuestionAnsweringModelOutput
+.. autoclass:: transformers.models.longformer.modeling_tf_longformer.TFLongformerQuestionAnsweringModelOutput
     :members: 
 
-LongformerModel
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 LongformerModel
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

docs/source/model_doc/lxmert.rst

@@ -67,19 +67,19 @@ LxmertTokenizerFast
 Lxmert specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_lxmert.LxmertModelOutput
+.. autoclass:: transformers.models.lxmert.modeling_lxmert.LxmertModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_lxmert.LxmertForPreTrainingOutput
+.. autoclass:: transformers.models.lxmert.modeling_lxmert.LxmertForPreTrainingOutput
     :members:
 
-.. autoclass:: transformers.modeling_lxmert.LxmertForQuestionAnsweringOutput
+.. autoclass:: transformers.models.lxmert.modeling_lxmert.LxmertForQuestionAnsweringOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_lxmert.TFLxmertModelOutput
+.. autoclass:: transformers.models.lxmert.modeling_tf_lxmert.TFLxmertModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_lxmert.TFLxmertForPreTrainingOutput
+.. autoclass:: transformers.models.lxmert.modeling_tf_lxmert.TFLxmertForPreTrainingOutput
     :members:
 
 

docs/source/model_doc/mobilebert.rst

@@ -58,10 +58,10 @@ MobileBertTokenizerFast
 MobileBert specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_mobilebert.MobileBertForPreTrainingOutput
+.. autoclass:: transformers.models.mobilebert.modeling_mobilebert.MobileBertForPreTrainingOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_mobilebert.TFMobileBertForPreTrainingOutput
+.. autoclass:: transformers.models.mobilebert.modeling_tf_mobilebert.TFMobileBertForPreTrainingOutput
     :members:
 
 

docs/source/model_doc/prophetnet.rst

@@ -47,16 +47,16 @@ ProphetNetTokenizer
 ProphetNet specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_prophetnet.ProphetNetSeq2SeqLMOutput
+.. autoclass:: transformers.models.prophetnet.modeling_prophetnet.ProphetNetSeq2SeqLMOutput
     :members:
 
-.. autoclass:: transformers.modeling_prophetnet.ProphetNetSeq2SeqModelOutput
+.. autoclass:: transformers.models.prophetnet.modeling_prophetnet.ProphetNetSeq2SeqModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_prophetnet.ProphetNetDecoderModelOutput
+.. autoclass:: transformers.models.prophetnet.modeling_prophetnet.ProphetNetDecoderModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_prophetnet.ProphetNetDecoderLMOutput
+.. autoclass:: transformers.models.prophetnet.modeling_prophetnet.ProphetNetDecoderLMOutput
     :members:
 
 ProphetNetModel

docs/source/model_doc/rag.rst

@@ -50,10 +50,10 @@ RagTokenizer
 Rag specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_rag.RetrievAugLMMarginOutput
+.. autoclass:: transformers.models.rag.modeling_rag.RetrievAugLMMarginOutput
     :members:
 
-.. autoclass:: transformers.modeling_rag.RetrievAugLMOutput
+.. autoclass:: transformers.models.rag.modeling_rag.RetrievAugLMOutput
     :members:
 
 RagRetriever

docs/source/model_doc/transformerxl.rst

@@ -49,16 +49,16 @@ TransfoXLTokenizer
 TransfoXL specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_transfo_xl.TransfoXLModelOutput
+.. autoclass:: transformers.models.transfo_xl.modeling_transfo_xl.TransfoXLModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_transfo_xl.TransfoXLLMHeadModelOutput
+.. autoclass:: transformers.models.transfo_xl.modeling_transfo_xl.TransfoXLLMHeadModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_transfo_xl.TFTransfoXLModelOutput
+.. autoclass:: transformers.models.transfo_xl.modeling_tf_transfo_xl.TFTransfoXLModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_transfo_xl.TFTransfoXLLMHeadModelOutput
+.. autoclass:: transformers.models.transfo_xl.modeling_tf_transfo_xl.TFTransfoXLLMHeadModelOutput
     :members:
 
 

docs/source/model_doc/xlm.rst

@@ -50,7 +50,7 @@ XLMTokenizer
 XLM specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_xlm.XLMForQuestionAnsweringOutput
+.. autoclass:: transformers.models.xlm.modeling_xlm.XLMForQuestionAnsweringOutput
     :members:
 
 

docs/source/model_doc/xlnet.rst

@@ -53,43 +53,43 @@ XLNetTokenizer
 XLNet specific outputs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autoclass:: transformers.modeling_xlnet.XLNetModelOutput
+.. autoclass:: transformers.models.xlnet.modeling_xlnet.XLNetModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_xlnet.XLNetLMHeadModelOutput
+.. autoclass:: transformers.models.xlnet.modeling_xlnet.XLNetLMHeadModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_xlnet.XLNetForSequenceClassificationOutput
+.. autoclass:: transformers.models.xlnet.modeling_xlnet.XLNetForSequenceClassificationOutput
     :members:
 
-.. autoclass:: transformers.modeling_xlnet.XLNetForMultipleChoiceOutput
+.. autoclass:: transformers.models.xlnet.modeling_xlnet.XLNetForMultipleChoiceOutput
     :members:
 
-.. autoclass:: transformers.modeling_xlnet.XLNetForTokenClassificationOutput
+.. autoclass:: transformers.models.xlnet.modeling_xlnet.XLNetForTokenClassificationOutput
     :members:
 
-.. autoclass:: transformers.modeling_xlnet.XLNetForQuestionAnsweringSimpleOutput
+.. autoclass:: transformers.models.xlnet.modeling_xlnet.XLNetForQuestionAnsweringSimpleOutput
     :members:
 
-.. autoclass:: transformers.modeling_xlnet.XLNetForQuestionAnsweringOutput
+.. autoclass:: transformers.models.xlnet.modeling_xlnet.XLNetForQuestionAnsweringOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_xlnet.TFXLNetModelOutput
+.. autoclass:: transformers.models.xlnet.modeling_tf_xlnet.TFXLNetModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_xlnet.TFXLNetLMHeadModelOutput
+.. autoclass:: transformers.models.xlnet.modeling_tf_xlnet.TFXLNetLMHeadModelOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_xlnet.TFXLNetForSequenceClassificationOutput
+.. autoclass:: transformers.models.xlnet.modeling_tf_xlnet.TFXLNetForSequenceClassificationOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_xlnet.TFXLNetForMultipleChoiceOutput
+.. autoclass:: transformers.models.xlnet.modeling_tf_xlnet.TFXLNetForMultipleChoiceOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_xlnet.TFXLNetForTokenClassificationOutput
+.. autoclass:: transformers.models.xlnet.modeling_tf_xlnet.TFXLNetForTokenClassificationOutput
     :members:
 
-.. autoclass:: transformers.modeling_tf_xlnet.TFXLNetForQuestionAnsweringSimpleOutput
+.. autoclass:: transformers.models.xlnet.modeling_tf_xlnet.TFXLNetForQuestionAnsweringSimpleOutput
     :members:
 
 

examples/bert-loses-patience/pabee/modeling_pabee_albert.py

@@ -21,7 +21,7 @@ import torch.nn as nn
 from torch.nn import CrossEntropyLoss, MSELoss
 
 from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
-from transformers.modeling_albert import (
+from transformers.models.albert.modeling_albert import (
     ALBERT_INPUTS_DOCSTRING,
     ALBERT_START_DOCSTRING,
     AlbertModel,

examples/bert-loses-patience/pabee/modeling_pabee_bert.py

@@ -23,7 +23,7 @@ from torch import nn
 from torch.nn import CrossEntropyLoss, MSELoss
 
 from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
-from transformers.modeling_bert import (
+from transformers.models.bert.modeling_bert import (
     BERT_INPUTS_DOCSTRING,
     BERT_START_DOCSTRING,
     BertEncoder,

examples/contrib/run_camembert.py

@@ -1,7 +1,6 @@
 import torch
 
-from transformers.modeling_camembert import CamembertForMaskedLM
+from transformers import CamembertForMaskedLM, CamembertTokenizer
-from transformers.tokenization_camembert import CamembertTokenizer
 
 
 def fill_mask(masked_input, model, tokenizer, topk=5):

examples/contrib/run_swag.py

@@ -32,8 +32,14 @@ from torch.utils.data.distributed import DistributedSampler
 from tqdm import tqdm, trange
 
 import transformers
-from transformers import WEIGHTS_NAME, AdamW, AutoConfig, AutoTokenizer, get_linear_schedule_with_warmup
+from transformers import (
-from transformers.modeling_auto import AutoModelForMultipleChoice
+    WEIGHTS_NAME,
+    AdamW,
+    AutoConfig,
+    AutoModelForMultipleChoice,
+    AutoTokenizer,
+    get_linear_schedule_with_warmup,
+)
 from transformers.trainer_utils import is_main_process
 
 

examples/deebert/src/modeling_highway_bert.py

@@ -3,7 +3,7 @@ from torch import nn
 from torch.nn import CrossEntropyLoss, MSELoss
 
 from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
-from transformers.modeling_bert import (
+from transformers.models.bert.modeling_bert import (
     BERT_INPUTS_DOCSTRING,
     BERT_START_DOCSTRING,
     BertEmbeddings,

examples/deebert/src/modeling_highway_roberta.py

@@ -3,9 +3,13 @@ from __future__ import absolute_import, division, print_function, unicode_litera
 import torch.nn as nn
 from torch.nn import CrossEntropyLoss, MSELoss
 
-from transformers.configuration_roberta import RobertaConfig
+from transformers import RobertaConfig
 from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
-from transformers.modeling_roberta import ROBERTA_INPUTS_DOCSTRING, ROBERTA_START_DOCSTRING, RobertaEmbeddings
+from transformers.models.roberta.modeling_roberta import (
+    ROBERTA_INPUTS_DOCSTRING,
+    ROBERTA_START_DOCSTRING,
+    RobertaEmbeddings,
+)
 
 from .modeling_highway_bert import BertPreTrainedModel, DeeBertModel, HighwayException, entropy
 

examples/movement-pruning/emmental/modeling_bert_masked.py

@@ -16,7 +16,7 @@
 """Masked Version of BERT. It replaces the `torch.nn.Linear` layers with
 :class:`~emmental.MaskedLinear` and add an additional parameters in the forward pass to
 compute the adaptive mask.
-Built on top of `transformers.modeling_bert`"""
+Built on top of `transformers.models.bert.modeling_bert`"""
 
 
 import logging
@@ -29,8 +29,8 @@ from torch.nn import CrossEntropyLoss, MSELoss
 from emmental import MaskedBertConfig
 from emmental.modules import MaskedLinear
 from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
-from transformers.modeling_bert import ACT2FN, BertLayerNorm, load_tf_weights_in_bert
 from transformers.modeling_utils import PreTrainedModel, prune_linear_layer
+from transformers.models.bert.modeling_bert import ACT2FN, BertLayerNorm, load_tf_weights_in_bert
 
 
 logger = logging.getLogger(__name__)

examples/rag/distributed_retriever.py

@@ -27,7 +27,7 @@ class RagPyTorchDistributedRetriever(RagRetriever):
             It is used to decode the question and then use the generator_tokenizer.
         generator_tokenizer (:class:`~transformers.PretrainedTokenizer`):
             The tokenizer used for the generator part of the RagModel.
-        index (:class:`~transformers.retrieval_rag.Index`, optional, defaults to the one defined by the configuration):
+        index (:class:`~transformers.models.rag.retrieval_rag.Index`, optional, defaults to the one defined by the configuration):
             If specified, use this index instead of the one built using the configuration
     """
 

examples/rag/test_distributed_retriever.py

@@ -11,16 +11,12 @@ import numpy as np
 from datasets import Dataset
 
 import faiss
-from transformers.configuration_bart import BartConfig
+from transformers import BartConfig, BartTokenizer, DPRConfig, DPRQuestionEncoderTokenizer, RagConfig
-from transformers.configuration_dpr import DPRConfig
-from transformers.configuration_rag import RagConfig
 from transformers.file_utils import is_datasets_available, is_faiss_available, is_psutil_available, is_torch_available
-from transformers.retrieval_rag import CustomHFIndex
+from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES
+from transformers.models.rag.retrieval_rag import CustomHFIndex
+from transformers.models.roberta.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
 from transformers.testing_utils import require_torch_non_multi_gpu_but_fix_me
-from transformers.tokenization_bart import BartTokenizer
-from transformers.tokenization_bert import VOCAB_FILES_NAMES as DPR_VOCAB_FILES_NAMES
-from transformers.tokenization_dpr import DPRQuestionEncoderTokenizer
-from transformers.tokenization_roberta import VOCAB_FILES_NAMES as BART_VOCAB_FILES_NAMES
 
 
 sys.path.append(os.path.join(os.getcwd()))  # noqa: E402 # noqa: E402 # isort:skip
@@ -137,7 +133,7 @@ class RagRetrieverTest(TestCase):
             question_encoder=DPRConfig().to_dict(),
             generator=BartConfig().to_dict(),
         )
-        with patch("transformers.retrieval_rag.load_dataset") as mock_load_dataset:
+        with patch("transformers.models.rag.retrieval_rag.load_dataset") as mock_load_dataset:
             mock_load_dataset.return_value = dataset
             retriever = RagPyTorchDistributedRetriever(
                 config,

examples/seq2seq/distillation.py

@@ -16,7 +16,7 @@ from finetune import SummarizationModule, TranslationModule
 from finetune import main as ft_main
 from make_student import create_student_by_copying_alternating_layers, get_layers_to_supervise
 from transformers import AutoModelForSeq2SeqLM, MBartTokenizer, T5ForConditionalGeneration
-from transformers.modeling_bart import shift_tokens_right
+from transformers.models.bart.modeling_bart import shift_tokens_right
 from utils import calculate_bleu, check_output_dir, freeze_params, label_smoothed_nll_loss, use_task_specific_params
 
 

examples/seq2seq/finetune.py

@@ -17,7 +17,7 @@ from torch.utils.data import DataLoader
 
 from callbacks import Seq2SeqLoggingCallback, get_checkpoint_callback, get_early_stopping_callback
 from transformers import MBartTokenizer, T5ForConditionalGeneration
-from transformers.modeling_bart import shift_tokens_right
+from transformers.models.bart.modeling_bart import shift_tokens_right
 from utils import (
     ROUGE_KEYS,
     LegacySeq2SeqDataset,

examples/seq2seq/seq2seq_trainer.py

@@ -5,8 +5,8 @@ from torch import nn
 from torch.utils.data import DistributedSampler, RandomSampler
 
 from transformers import PreTrainedModel, Trainer, logging
-from transformers.configuration_fsmt import FSMTConfig
 from transformers.file_utils import is_torch_tpu_available
+from transformers.models.fsmt.configuration_fsmt import FSMTConfig
 from transformers.optimization import (
     Adafactor,
     AdamW,

examples/seq2seq/test_datasets.py

@@ -10,7 +10,7 @@ from parameterized import parameterized
 from save_len_file import save_len_file
 from test_seq2seq_examples import ARTICLES, BART_TINY, MARIAN_TINY, MBART_TINY, SUMMARIES, T5_TINY, make_test_data_dir
 from transformers import AutoTokenizer
-from transformers.modeling_bart import shift_tokens_right
+from transformers.models.bart.modeling_bart import shift_tokens_right
 from transformers.testing_utils import TestCasePlus, require_torch_non_multi_gpu_but_fix_me, slow
 from utils import FAIRSEQ_AVAILABLE, DistributedSortishSampler, LegacySeq2SeqDataset, Seq2SeqDataset
 

examples/seq2seq/test_tatoeba_conversion.py

@@ -2,8 +2,8 @@ import os
 import tempfile
 import unittest
 
-from transformers.convert_marian_tatoeba_to_pytorch import DEFAULT_REPO, TatoebaConverter
 from transformers.file_utils import cached_property
+from transformers.models.marian.convert_marian_tatoeba_to_pytorch import DEFAULT_REPO, TatoebaConverter
 from transformers.testing_utils import require_torch_non_multi_gpu_but_fix_me, slow
 
 

examples/seq2seq/utils.py

@@ -21,7 +21,7 @@ from torch.utils.data import Dataset, Sampler
 from sentence_splitter import add_newline_to_end_of_each_sentence
 from transformers import BartTokenizer, EvalPrediction, PreTrainedTokenizer, T5Tokenizer
 from transformers.file_utils import cached_property
-from transformers.modeling_bart import shift_tokens_right
+from transformers.models.bart.modeling_bart import shift_tokens_right
 
 
 try:

examples/text-generation/pplm/run_pplm.py

@@ -34,9 +34,8 @@ import torch.nn.functional as F
 from tqdm import trange
 
 from pplm_classification_head import ClassificationHead
-from transformers import GPT2Tokenizer
+from transformers import GPT2LMHeadModel, GPT2Tokenizer
 from transformers.file_utils import cached_path
-from transformers.modeling_gpt2 import GPT2LMHeadModel
 
 
 PPLM_BOW = 1

model_cards/allenai/wmt16-en-de-12-1/README.md

@@ -35,8 +35,7 @@ All 3 models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "allenai/wmt16-en-de-12-1"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/allenai/wmt16-en-de-dist-12-1/README.md

@@ -35,8 +35,7 @@ All 3 models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "allenai/wmt16-en-de-dist-12-1"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/allenai/wmt16-en-de-dist-6-1/README.md

@@ -35,8 +35,7 @@ All 3 models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "allenai/wmt16-en-de-dist-6-1"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/allenai/wmt19-de-en-6-6-base/README.md

@@ -35,8 +35,7 @@ For more details, please, see [Deep Encoder, Shallow Decoder: Reevaluating the S
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "allenai/wmt19-de-en-6-6-base"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/allenai/wmt19-de-en-6-6-big/README.md

@@ -35,8 +35,7 @@ For more details, please, see [Deep Encoder, Shallow Decoder: Reevaluating the S
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "allenai/wmt19-de-en-6-6-big"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/deepset/electra-base-squad2/README.md

@@ -47,9 +47,7 @@ Evaluated on the SQuAD 2.0 dev set with the [official eval script](https://works
 
 ### In Transformers
 ```python
-from transformers.pipelines import pipeline
+from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
-from transformers.modeling_auto import AutoModelForQuestionAnswering
-from transformers.tokenization_auto import AutoTokenizer
 
 model_name = "deepset/electra-base-squad2"
 

model_cards/deepset/minilm-uncased-squad2/README.md

@@ -48,9 +48,7 @@ Evaluated on the SQuAD 2.0 dev set with the [official eval script](https://works
 
 ### In Transformers
 ```python
-from transformers.pipelines import pipeline
+from transformers import AutoModelForQuestionAnswering,  AutoTokenizer, pipeline
-from transformers.modeling_auto import AutoModelForQuestionAnswering
-from transformers.tokenization_auto import AutoTokenizer
 
 model_name = "deepset/minilm-uncased-squad2"
 

model_cards/deepset/roberta-base-squad2-covid/README.md

@@ -39,9 +39,8 @@ This model is the model obtained from the **third** fold of the cross-validation
 
 ### In Transformers
 ```python
-from transformers.pipelines import pipeline
+from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
-from transformers.modeling_auto import AutoModelForQuestionAnswering
+
-from transformers.tokenization_auto import AutoTokenizer
 
 model_name = "deepset/roberta-base-squad2-covid"
 

model_cards/deepset/roberta-base-squad2-v2/README.md

@@ -48,9 +48,7 @@ Evaluated on the SQuAD 2.0 dev set with the [official eval script](https://works
 
 ### In Transformers
 ```python
-from transformers.pipelines import pipeline
+from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
-from transformers.modeling_auto import AutoModelForQuestionAnswering
-from transformers.tokenization_auto import AutoTokenizer
 
 model_name = "deepset/roberta-base-squad2-v2"
 

model_cards/deepset/roberta-base-squad2/README.md

@@ -54,9 +54,7 @@ Evaluated on the SQuAD 2.0 dev set with the [official eval script](https://works
 
 ### In Transformers
 ```python
-from transformers.pipelines import pipeline
+from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
-from transformers.modeling_auto import AutoModelForQuestionAnswering
-from transformers.tokenization_auto import AutoTokenizer
 
 model_name = "deepset/roberta-base-squad2"
 

model_cards/deepset/xlm-roberta-large-squad2/README.md

@@ -63,9 +63,7 @@ Evaluated on German [XQuAD: xquad.de.json](https://github.com/deepmind/xquad)
 
 ### In Transformers
 ```python
-from transformers.pipelines import pipeline
+from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
-from transformers.modeling_auto import AutoModelForQuestionAnswering
-from transformers.tokenization_auto import AutoTokenizer
 
 model_name = "deepset/xlm-roberta-large-squad2"
 

model_cards/facebook/wmt19-de-en/README.md

@@ -36,8 +36,7 @@ All four models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "facebook/wmt19-de-en"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/facebook/wmt19-en-de/README.md

@@ -36,8 +36,7 @@ All four models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "facebook/wmt19-en-de"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/facebook/wmt19-en-ru/README.md

@@ -36,8 +36,7 @@ All four models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "facebook/wmt19-en-ru"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/facebook/wmt19-ru-en/README.md

@@ -36,8 +36,7 @@ All four models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "facebook/wmt19-ru-en"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

model_cards/julien-c/bert-xsmall-dummy/README.md

@@ -2,11 +2,7 @@
 
 
 ```python
-from transformers.configuration_bert import BertConfig
+from transformers BertConfig, BertForMaskedLM, BertTokenizer, TFBertForMaskedLM
-from transformers.modeling_bert import BertForMaskedLM
-from transformers.modeling_tf_bert import TFBertForMaskedLM
-from transformers.tokenization_bert import BertTokenizer
-
 
 SMALL_MODEL_IDENTIFIER = "julien-c/bert-xsmall-dummy"
 DIRNAME = "./bert-xsmall-dummy"

scripts/fsmt/gen-card-allenai-wmt16.py

@@ -60,8 +60,7 @@ All 3 models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "allenai/{model_name}"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

scripts/fsmt/gen-card-allenai-wmt19.py

@@ -59,8 +59,7 @@ For more details, please, see [Deep Encoder, Shallow Decoder: Reevaluating the S
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "allenai/{model_name}"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

scripts/fsmt/gen-card-facebook-wmt19.py

@@ -63,8 +63,7 @@ All four models are available:
 #### How to use
 
 ```python
-from transformers.tokenization_fsmt import FSMTTokenizer
+from transformers import FSMTForConditionalGeneration, FSMTTokenizer
-from transformers.modeling_fsmt import FSMTForConditionalGeneration
 mname = "facebook/wmt19-{src_lang}-{tgt_lang}"
 tokenizer = FSMTTokenizer.from_pretrained(mname)
 model = FSMTForConditionalGeneration.from_pretrained(mname)

src/transformers/__init__.py

@@ -27,46 +27,10 @@ from .integrations import (  # isort:skip
     is_wandb_available,
 )
 
-# Configurations
+# Configuration
-from .configuration_albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig
-from .configuration_auto import ALL_PRETRAINED_CONFIG_ARCHIVE_MAP, CONFIG_MAPPING, AutoConfig
-from .configuration_bart import BartConfig
-from .configuration_bert import BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, BertConfig
-from .configuration_bert_generation import BertGenerationConfig
-from .configuration_blenderbot import BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP, BlenderbotConfig
-from .configuration_camembert import CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, CamembertConfig
-from .configuration_ctrl import CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP, CTRLConfig
-from .configuration_deberta import DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, DebertaConfig
-from .configuration_distilbert import DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, DistilBertConfig
-from .configuration_dpr import DPR_PRETRAINED_CONFIG_ARCHIVE_MAP, DPRConfig
-from .configuration_electra import ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP, ElectraConfig
-from .configuration_encoder_decoder import EncoderDecoderConfig
-from .configuration_flaubert import FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, FlaubertConfig
-from .configuration_fsmt import FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP, FSMTConfig
-from .configuration_funnel import FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP, FunnelConfig
-from .configuration_gpt2 import GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP, GPT2Config
-from .configuration_layoutlm import LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP, LayoutLMConfig
-from .configuration_longformer import LONGFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, LongformerConfig
-from .configuration_lxmert import LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP, LxmertConfig
-from .configuration_marian import MarianConfig
-from .configuration_mbart import MBartConfig
-from .configuration_mmbt import MMBTConfig
-from .configuration_mobilebert import MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileBertConfig
-from .configuration_openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig
-from .configuration_pegasus import PegasusConfig
-from .configuration_prophetnet import PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, ProphetNetConfig
-from .configuration_rag import RagConfig
-from .configuration_reformer import REFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, ReformerConfig
-from .configuration_retribert import RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, RetriBertConfig
-from .configuration_roberta import ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, RobertaConfig
-from .configuration_squeezebert import SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, SqueezeBertConfig
-from .configuration_t5 import T5_PRETRAINED_CONFIG_ARCHIVE_MAP, T5Config
-from .configuration_transfo_xl import TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP, TransfoXLConfig
 from .configuration_utils import PretrainedConfig
-from .configuration_xlm import XLM_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMConfig
+
-from .configuration_xlm_prophetnet import XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMProphetNetConfig
+# Data
-from .configuration_xlm_roberta import XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMRobertaConfig
-from .configuration_xlnet import XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLNetConfig
 from .data import (
     DataProcessor,
     InputExample,
@@ -130,6 +94,77 @@ from .modeling_tf_pytorch_utils import (
     load_tf2_model_in_pytorch_model,
     load_tf2_weights_in_pytorch_model,
 )
+from .models.albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig
+from .models.auto import (
+    ALL_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    CONFIG_MAPPING,
+    TOKENIZER_MAPPING,
+    AutoConfig,
+    AutoTokenizer,
+)
+from .models.bart import BartConfig, BartTokenizer
+from .models.bert import (
+    BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    BasicTokenizer,
+    BertConfig,
+    BertTokenizer,
+    WordpieceTokenizer,
+)
+from .models.bert_generation import BertGenerationConfig
+from .models.bert_japanese import BertJapaneseTokenizer, CharacterTokenizer, MecabTokenizer
+from .models.bertweet import BertweetTokenizer
+from .models.blenderbot import (
+    BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    BlenderbotConfig,
+    BlenderbotSmallTokenizer,
+    BlenderbotTokenizer,
+)
+from .models.camembert import CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, CamembertConfig
+from .models.ctrl import CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP, CTRLConfig, CTRLTokenizer
+from .models.deberta import DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, DebertaConfig, DebertaTokenizer
+from .models.distilbert import DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, DistilBertConfig, DistilBertTokenizer
+from .models.dpr import (
+    DPR_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    DPRConfig,
+    DPRContextEncoderTokenizer,
+    DPRQuestionEncoderTokenizer,
+    DPRReaderOutput,
+    DPRReaderTokenizer,
+)
+from .models.electra import ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP, ElectraConfig, ElectraTokenizer
+from .models.encoder_decoder import EncoderDecoderConfig
+from .models.flaubert import FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, FlaubertConfig, FlaubertTokenizer
+from .models.fsmt import FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP, FSMTConfig, FSMTTokenizer
+from .models.funnel import FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP, FunnelConfig, FunnelTokenizer
+from .models.gpt2 import GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP, GPT2Config, GPT2Tokenizer
+from .models.herbert import HerbertTokenizer
+from .models.layoutlm import LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP, LayoutLMConfig, LayoutLMTokenizer
+from .models.longformer import LONGFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, LongformerConfig, LongformerTokenizer
+from .models.lxmert import LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP, LxmertConfig, LxmertTokenizer
+from .models.marian import MarianConfig
+from .models.mbart import MBartConfig
+from .models.mmbt import MMBTConfig
+from .models.mobilebert import MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileBertConfig, MobileBertTokenizer
+from .models.openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig, OpenAIGPTTokenizer
+from .models.pegasus import PegasusConfig
+from .models.phobert import PhobertTokenizer
+from .models.prophetnet import PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, ProphetNetConfig, ProphetNetTokenizer
+from .models.rag import RagConfig, RagRetriever, RagTokenizer
+from .models.reformer import REFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, ReformerConfig
+from .models.retribert import RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, RetriBertConfig, RetriBertTokenizer
+from .models.roberta import ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, RobertaConfig, RobertaTokenizer
+from .models.squeezebert import SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, SqueezeBertConfig, SqueezeBertTokenizer
+from .models.t5 import T5_PRETRAINED_CONFIG_ARCHIVE_MAP, T5Config
+from .models.transfo_xl import (
+    TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    TransfoXLConfig,
+    TransfoXLCorpus,
+    TransfoXLTokenizer,
+)
+from .models.xlm import XLM_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMConfig, XLMTokenizer
+from .models.xlm_prophetnet import XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMProphetNetConfig
+from .models.xlm_roberta import XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMRobertaConfig
+from .models.xlnet import XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLNetConfig
 
 # Pipelines
 from .pipelines import (
@@ -154,43 +189,7 @@ from .pipelines import (
     pipeline,
 )
 
-# Retriever
+# Tokenization
-from .retrieval_rag import RagRetriever
-
-# Tokenizers
-from .tokenization_auto import TOKENIZER_MAPPING, AutoTokenizer
-from .tokenization_bart import BartTokenizer
-from .tokenization_bert import BasicTokenizer, BertTokenizer, WordpieceTokenizer
-from .tokenization_bert_japanese import BertJapaneseTokenizer, CharacterTokenizer, MecabTokenizer
-from .tokenization_bertweet import BertweetTokenizer
-from .tokenization_blenderbot import BlenderbotSmallTokenizer, BlenderbotTokenizer
-from .tokenization_ctrl import CTRLTokenizer
-from .tokenization_deberta import DebertaTokenizer
-from .tokenization_distilbert import DistilBertTokenizer
-from .tokenization_dpr import (
-    DPRContextEncoderTokenizer,
-    DPRQuestionEncoderTokenizer,
-    DPRReaderOutput,
-    DPRReaderTokenizer,
-)
-from .tokenization_electra import ElectraTokenizer
-from .tokenization_flaubert import FlaubertTokenizer
-from .tokenization_fsmt import FSMTTokenizer
-from .tokenization_funnel import FunnelTokenizer
-from .tokenization_gpt2 import GPT2Tokenizer
-from .tokenization_herbert import HerbertTokenizer
-from .tokenization_layoutlm import LayoutLMTokenizer
-from .tokenization_longformer import LongformerTokenizer
-from .tokenization_lxmert import LxmertTokenizer
-from .tokenization_mobilebert import MobileBertTokenizer
-from .tokenization_openai import OpenAIGPTTokenizer
-from .tokenization_phobert import PhobertTokenizer
-from .tokenization_prophetnet import ProphetNetTokenizer
-from .tokenization_rag import RagTokenizer
-from .tokenization_retribert import RetriBertTokenizer
-from .tokenization_roberta import RobertaTokenizer
-from .tokenization_squeezebert import SqueezeBertTokenizer
-from .tokenization_transfo_xl import TransfoXLCorpus, TransfoXLTokenizer
 from .tokenization_utils import PreTrainedTokenizer
 from .tokenization_utils_base import (
     AddedToken,
@@ -201,54 +200,49 @@ from .tokenization_utils_base import (
     TensorType,
     TokenSpan,
 )
-from .tokenization_xlm import XLMTokenizer
 
 
 if is_sentencepiece_available():
-    from .tokenization_albert import AlbertTokenizer
+    from .models.albert import AlbertTokenizer
-    from .tokenization_bert_generation import BertGenerationTokenizer
+    from .models.bert_generation import BertGenerationTokenizer
-    from .tokenization_camembert import CamembertTokenizer
+    from .models.camembert import CamembertTokenizer
-    from .tokenization_marian import MarianTokenizer
+    from .models.marian import MarianTokenizer
-    from .tokenization_mbart import MBartTokenizer
+    from .models.mbart import MBartTokenizer
-    from .tokenization_pegasus import PegasusTokenizer
+    from .models.pegasus import PegasusTokenizer
-    from .tokenization_reformer import ReformerTokenizer
+    from .models.reformer import ReformerTokenizer
-    from .tokenization_t5 import T5Tokenizer
+    from .models.t5 import T5Tokenizer
-    from .tokenization_xlm_prophetnet import XLMProphetNetTokenizer
+    from .models.xlm_prophetnet import XLMProphetNetTokenizer
-    from .tokenization_xlm_roberta import XLMRobertaTokenizer
+    from .models.xlm_roberta import XLMRobertaTokenizer
-    from .tokenization_xlnet import XLNetTokenizer
+    from .models.xlnet import XLNetTokenizer
 else:
     from .utils.dummy_sentencepiece_objects import *
 
 if is_tokenizers_available():
-    from .tokenization_albert_fast import AlbertTokenizerFast
+    from .models.albert import AlbertTokenizerFast
-    from .tokenization_bart_fast import BartTokenizerFast
+    from .models.bart import BartTokenizerFast
-    from .tokenization_bert_fast import BertTokenizerFast
+    from .models.bert import BertTokenizerFast
-    from .tokenization_camembert_fast import CamembertTokenizerFast
+    from .models.camembert import CamembertTokenizerFast
-    from .tokenization_distilbert_fast import DistilBertTokenizerFast
+    from .models.distilbert import DistilBertTokenizerFast
-    from .tokenization_dpr_fast import (
+    from .models.dpr import DPRContextEncoderTokenizerFast, DPRQuestionEncoderTokenizerFast, DPRReaderTokenizerFast
-        DPRContextEncoderTokenizerFast,
+    from .models.electra import ElectraTokenizerFast
-        DPRQuestionEncoderTokenizerFast,
+    from .models.funnel import FunnelTokenizerFast
-        DPRReaderTokenizerFast,
+    from .models.gpt2 import GPT2TokenizerFast
-    )
+    from .models.herbert import HerbertTokenizerFast
-    from .tokenization_electra_fast import ElectraTokenizerFast
+    from .models.layoutlm import LayoutLMTokenizerFast
-    from .tokenization_funnel_fast import FunnelTokenizerFast
+    from .models.longformer import LongformerTokenizerFast
-    from .tokenization_gpt2_fast import GPT2TokenizerFast
+    from .models.lxmert import LxmertTokenizerFast
-    from .tokenization_herbert_fast import HerbertTokenizerFast
+    from .models.mbart import MBartTokenizerFast
-    from .tokenization_layoutlm_fast import LayoutLMTokenizerFast
+    from .models.mobilebert import MobileBertTokenizerFast
-    from .tokenization_longformer_fast import LongformerTokenizerFast
+    from .models.openai import OpenAIGPTTokenizerFast
-    from .tokenization_lxmert_fast import LxmertTokenizerFast
+    from .models.pegasus import PegasusTokenizerFast
-    from .tokenization_mbart_fast import MBartTokenizerFast
+    from .models.reformer import ReformerTokenizerFast
-    from .tokenization_mobilebert_fast import MobileBertTokenizerFast
+    from .models.retribert import RetriBertTokenizerFast
-    from .tokenization_openai_fast import OpenAIGPTTokenizerFast
+    from .models.roberta import RobertaTokenizerFast
-    from .tokenization_pegasus_fast import PegasusTokenizerFast
+    from .models.squeezebert import SqueezeBertTokenizerFast
-    from .tokenization_reformer_fast import ReformerTokenizerFast
+    from .models.t5 import T5TokenizerFast
-    from .tokenization_retribert_fast import RetriBertTokenizerFast
+    from .models.xlm_roberta import XLMRobertaTokenizerFast
-    from .tokenization_roberta_fast import RobertaTokenizerFast
+    from .models.xlnet import XLNetTokenizerFast
-    from .tokenization_squeezebert_fast import SqueezeBertTokenizerFast
-    from .tokenization_t5_fast import T5TokenizerFast
     from .tokenization_utils_fast import PreTrainedTokenizerFast
-    from .tokenization_xlm_roberta_fast import XLMRobertaTokenizerFast
-    from .tokenization_xlnet_fast import XLNetTokenizerFast
 
     if is_sentencepiece_available():
         from .convert_slow_tokenizer import SLOW_TO_FAST_CONVERTERS, convert_slow_tokenizer
@@ -313,7 +307,8 @@ if is_torch_available():
         TopPLogitsWarper,
     )
     from .generation_utils import top_k_top_p_filtering
-    from .modeling_albert import (
+    from .modeling_utils import Conv1D, PreTrainedModel, apply_chunking_to_forward, prune_layer
+    from .models.albert import (
         ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         AlbertForMaskedLM,
         AlbertForMultipleChoice,
@@ -325,7 +320,7 @@ if is_torch_available():
         AlbertPreTrainedModel,
         load_tf_weights_in_albert,
     )
-    from .modeling_auto import (
+    from .models.auto import (
         MODEL_FOR_CAUSAL_LM_MAPPING,
         MODEL_FOR_MASKED_LM_MAPPING,
         MODEL_FOR_MULTIPLE_CHOICE_MAPPING,
@@ -349,7 +344,7 @@ if is_torch_available():
         AutoModelForTokenClassification,
         AutoModelWithLMHead,
     )
-    from .modeling_bart import (
+    from .models.bart import (
         BART_PRETRAINED_MODEL_ARCHIVE_LIST,
         BartForConditionalGeneration,
         BartForQuestionAnswering,
@@ -357,7 +352,7 @@ if is_torch_available():
         BartModel,
         PretrainedBartModel,
     )
-    from .modeling_bert import (
+    from .models.bert import (
         BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         BertForMaskedLM,
         BertForMultipleChoice,
@@ -372,13 +367,13 @@ if is_torch_available():
         BertPreTrainedModel,
         load_tf_weights_in_bert,
     )
-    from .modeling_bert_generation import (
+    from .models.bert_generation import (
         BertGenerationDecoder,
         BertGenerationEncoder,
         load_tf_weights_in_bert_generation,
     )
-    from .modeling_blenderbot import BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST, BlenderbotForConditionalGeneration
+    from .models.blenderbot import BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST, BlenderbotForConditionalGeneration
-    from .modeling_camembert import (
+    from .models.camembert import (
         CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         CamembertForCausalLM,
         CamembertForMaskedLM,
@@ -388,14 +383,14 @@ if is_torch_available():
         CamembertForTokenClassification,
         CamembertModel,
     )
-    from .modeling_ctrl import CTRL_PRETRAINED_MODEL_ARCHIVE_LIST, CTRLLMHeadModel, CTRLModel, CTRLPreTrainedModel
+    from .models.ctrl import CTRL_PRETRAINED_MODEL_ARCHIVE_LIST, CTRLLMHeadModel, CTRLModel, CTRLPreTrainedModel
-    from .modeling_deberta import (
+    from .models.deberta import (
         DEBERTA_PRETRAINED_MODEL_ARCHIVE_LIST,
         DebertaForSequenceClassification,
         DebertaModel,
         DebertaPreTrainedModel,
     )
-    from .modeling_distilbert import (
+    from .models.distilbert import (
         DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         DistilBertForMaskedLM,
         DistilBertForMultipleChoice,
@@ -405,7 +400,7 @@ if is_torch_available():
         DistilBertModel,
         DistilBertPreTrainedModel,
     )
-    from .modeling_dpr import (
+    from .models.dpr import (
         DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
         DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
         DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST,
@@ -416,7 +411,7 @@ if is_torch_available():
         DPRQuestionEncoder,
         DPRReader,
     )
-    from .modeling_electra import (
+    from .models.electra import (
         ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST,
         ElectraForMaskedLM,
         ElectraForMultipleChoice,
@@ -428,8 +423,8 @@ if is_torch_available():
         ElectraPreTrainedModel,
         load_tf_weights_in_electra,
     )
-    from .modeling_encoder_decoder import EncoderDecoderModel
+    from .models.encoder_decoder import EncoderDecoderModel
-    from .modeling_flaubert import (
+    from .models.flaubert import (
         FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         FlaubertForMultipleChoice,
         FlaubertForQuestionAnswering,
@@ -439,8 +434,8 @@ if is_torch_available():
         FlaubertModel,
         FlaubertWithLMHeadModel,
     )
-    from .modeling_fsmt import FSMTForConditionalGeneration, FSMTModel, PretrainedFSMTModel
+    from .models.fsmt import FSMTForConditionalGeneration, FSMTModel, PretrainedFSMTModel
-    from .modeling_funnel import (
+    from .models.funnel import (
         FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST,
         FunnelBaseModel,
         FunnelForMaskedLM,
@@ -452,7 +447,7 @@ if is_torch_available():
         FunnelModel,
         load_tf_weights_in_funnel,
     )
-    from .modeling_gpt2 import (
+    from .models.gpt2 import (
         GPT2_PRETRAINED_MODEL_ARCHIVE_LIST,
         GPT2DoubleHeadsModel,
         GPT2ForSequenceClassification,
@@ -461,13 +456,13 @@ if is_torch_available():
         GPT2PreTrainedModel,
         load_tf_weights_in_gpt2,
     )
-    from .modeling_layoutlm import (
+    from .models.layoutlm import (
         LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST,
         LayoutLMForMaskedLM,
         LayoutLMForTokenClassification,
         LayoutLMModel,
     )
-    from .modeling_longformer import (
+    from .models.longformer import (
         LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
         LongformerForMaskedLM,
         LongformerForMultipleChoice,
@@ -477,7 +472,7 @@ if is_torch_available():
         LongformerModel,
         LongformerSelfAttention,
     )
-    from .modeling_lxmert import (
+    from .models.lxmert import (
         LxmertEncoder,
         LxmertForPreTraining,
         LxmertForQuestionAnswering,
@@ -486,10 +481,10 @@ if is_torch_available():
         LxmertVisualFeatureEncoder,
         LxmertXLayer,
     )
-    from .modeling_marian import MarianMTModel
+    from .models.marian import MarianMTModel
-    from .modeling_mbart import MBartForConditionalGeneration
+    from .models.mbart import MBartForConditionalGeneration
-    from .modeling_mmbt import MMBTForClassification, MMBTModel, ModalEmbeddings
+    from .models.mmbt import MMBTForClassification, MMBTModel, ModalEmbeddings
-    from .modeling_mobilebert import (
+    from .models.mobilebert import (
         MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         MobileBertForMaskedLM,
         MobileBertForMultipleChoice,
@@ -503,7 +498,7 @@ if is_torch_available():
         MobileBertPreTrainedModel,
         load_tf_weights_in_mobilebert,
     )
-    from .modeling_openai import (
+    from .models.openai import (
         OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST,
         OpenAIGPTDoubleHeadsModel,
         OpenAIGPTForSequenceClassification,
@@ -512,8 +507,8 @@ if is_torch_available():
         OpenAIGPTPreTrainedModel,
         load_tf_weights_in_openai_gpt,
     )
-    from .modeling_pegasus import PegasusForConditionalGeneration
+    from .models.pegasus import PegasusForConditionalGeneration
-    from .modeling_prophetnet import (
+    from .models.prophetnet import (
         PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST,
         ProphetNetDecoder,
         ProphetNetEncoder,
@@ -522,8 +517,8 @@ if is_torch_available():
         ProphetNetModel,
         ProphetNetPreTrainedModel,
     )
-    from .modeling_rag import RagModel, RagSequenceForGeneration, RagTokenForGeneration
+    from .models.rag import RagModel, RagSequenceForGeneration, RagTokenForGeneration
-    from .modeling_reformer import (
+    from .models.reformer import (
         REFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
         ReformerAttention,
         ReformerForMaskedLM,
@@ -533,8 +528,8 @@ if is_torch_available():
         ReformerModel,
         ReformerModelWithLMHead,
     )
-    from .modeling_retribert import RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST, RetriBertModel, RetriBertPreTrainedModel
+    from .models.retribert import RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST, RetriBertModel, RetriBertPreTrainedModel
-    from .modeling_roberta import (
+    from .models.roberta import (
         ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST,
         RobertaForCausalLM,
         RobertaForMaskedLM,
@@ -544,7 +539,7 @@ if is_torch_available():
         RobertaForTokenClassification,
         RobertaModel,
     )
-    from .modeling_squeezebert import (
+    from .models.squeezebert import (
         SQUEEZEBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         SqueezeBertForMaskedLM,
         SqueezeBertForMultipleChoice,
@@ -555,14 +550,14 @@ if is_torch_available():
         SqueezeBertModule,
         SqueezeBertPreTrainedModel,
     )
-    from .modeling_t5 import (
+    from .models.t5 import (
         T5_PRETRAINED_MODEL_ARCHIVE_LIST,
         T5ForConditionalGeneration,
         T5Model,
         T5PreTrainedModel,
         load_tf_weights_in_t5,
     )
-    from .modeling_transfo_xl import (
+    from .models.transfo_xl import (
         TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST,
         AdaptiveEmbedding,
         TransfoXLLMHeadModel,
@@ -570,8 +565,7 @@ if is_torch_available():
         TransfoXLPreTrainedModel,
         load_tf_weights_in_transfo_xl,
     )
-    from .modeling_utils import Conv1D, PreTrainedModel, apply_chunking_to_forward, prune_layer
+    from .models.xlm import (
-    from .modeling_xlm import (
         XLM_PRETRAINED_MODEL_ARCHIVE_LIST,
         XLMForMultipleChoice,
         XLMForQuestionAnswering,
@@ -582,7 +576,7 @@ if is_torch_available():
         XLMPreTrainedModel,
         XLMWithLMHeadModel,
     )
-    from .modeling_xlm_prophetnet import (
+    from .models.xlm_prophetnet import (
         XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST,
         XLMProphetNetDecoder,
         XLMProphetNetEncoder,
@@ -590,7 +584,7 @@ if is_torch_available():
         XLMProphetNetForConditionalGeneration,
         XLMProphetNetModel,
     )
-    from .modeling_xlm_roberta import (
+    from .models.xlm_roberta import (
         XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST,
         XLMRobertaForCausalLM,
         XLMRobertaForMaskedLM,
@@ -600,7 +594,7 @@ if is_torch_available():
         XLMRobertaForTokenClassification,
         XLMRobertaModel,
     )
-    from .modeling_xlnet import (
+    from .models.xlnet import (
         XLNET_PRETRAINED_MODEL_ARCHIVE_LIST,
         XLNetForMultipleChoice,
         XLNetForQuestionAnswering,
@@ -638,7 +632,8 @@ if is_tf_available():
     # Benchmarks
     from .benchmark.benchmark_tf import TensorFlowBenchmark
     from .generation_tf_utils import tf_top_k_top_p_filtering
-    from .modeling_tf_albert import (
+    from .modeling_tf_utils import TFPreTrainedModel, TFSequenceSummary, TFSharedEmbeddings, shape_list
+    from .models.albert import (
         TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFAlbertForMaskedLM,
         TFAlbertForMultipleChoice,
@@ -650,7 +645,7 @@ if is_tf_available():
         TFAlbertModel,
         TFAlbertPreTrainedModel,
     )
-    from .modeling_tf_auto import (
+    from .models.auto import (
         TF_MODEL_FOR_CAUSAL_LM_MAPPING,
         TF_MODEL_FOR_MASKED_LM_MAPPING,
         TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING,
@@ -673,8 +668,8 @@ if is_tf_available():
         TFAutoModelForTokenClassification,
         TFAutoModelWithLMHead,
     )
-    from .modeling_tf_bart import TFBartForConditionalGeneration, TFBartModel
+    from .models.bart import TFBartForConditionalGeneration, TFBartModel
-    from .modeling_tf_bert import (
+    from .models.bert import (
         TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFBertEmbeddings,
         TFBertForMaskedLM,
@@ -689,8 +684,8 @@ if is_tf_available():
         TFBertModel,
         TFBertPreTrainedModel,
     )
-    from .modeling_tf_blenderbot import TFBlenderbotForConditionalGeneration
+    from .models.blenderbot import TFBlenderbotForConditionalGeneration
-    from .modeling_tf_camembert import (
+    from .models.camembert import (
         TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFCamembertForMaskedLM,
         TFCamembertForMultipleChoice,
@@ -699,13 +694,13 @@ if is_tf_available():
         TFCamembertForTokenClassification,
         TFCamembertModel,
     )
-    from .modeling_tf_ctrl import (
+    from .models.ctrl import (
         TF_CTRL_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFCTRLLMHeadModel,
         TFCTRLModel,
         TFCTRLPreTrainedModel,
     )
-    from .modeling_tf_distilbert import (
+    from .models.distilbert import (
         TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFDistilBertForMaskedLM,
         TFDistilBertForMultipleChoice,
@@ -716,7 +711,7 @@ if is_tf_available():
         TFDistilBertModel,
         TFDistilBertPreTrainedModel,
     )
-    from .modeling_tf_dpr import (
+    from .models.dpr import (
         TF_DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
         TF_DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
         TF_DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST,
@@ -727,7 +722,7 @@ if is_tf_available():
         TFDPRQuestionEncoder,
         TFDPRReader,
     )
-    from .modeling_tf_electra import (
+    from .models.electra import (
         TF_ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFElectraForMaskedLM,
         TFElectraForMultipleChoice,
@@ -738,7 +733,7 @@ if is_tf_available():
         TFElectraModel,
         TFElectraPreTrainedModel,
     )
-    from .modeling_tf_flaubert import (
+    from .models.flaubert import (
         TF_FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFFlaubertForMultipleChoice,
         TFFlaubertForQuestionAnsweringSimple,
@@ -747,7 +742,7 @@ if is_tf_available():
         TFFlaubertModel,
         TFFlaubertWithLMHeadModel,
     )
-    from .modeling_tf_funnel import (
+    from .models.funnel import (
         TF_FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFFunnelBaseModel,
         TFFunnelForMaskedLM,
@@ -758,7 +753,7 @@ if is_tf_available():
         TFFunnelForTokenClassification,
         TFFunnelModel,
     )
-    from .modeling_tf_gpt2 import (
+    from .models.gpt2 import (
         TF_GPT2_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFGPT2DoubleHeadsModel,
         TFGPT2LMHeadModel,
@@ -766,14 +761,14 @@ if is_tf_available():
         TFGPT2Model,
         TFGPT2PreTrainedModel,
     )
-    from .modeling_tf_longformer import (
+    from .models.longformer import (
         TF_LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFLongformerForMaskedLM,
         TFLongformerForQuestionAnswering,
         TFLongformerModel,
         TFLongformerSelfAttention,
     )
-    from .modeling_tf_lxmert import (
+    from .models.lxmert import (
         TF_LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFLxmertForPreTraining,
         TFLxmertMainLayer,
@@ -781,9 +776,9 @@ if is_tf_available():
         TFLxmertPreTrainedModel,
         TFLxmertVisualFeatureEncoder,
     )
-    from .modeling_tf_marian import TFMarianMTModel
+    from .models.marian import TFMarianMTModel
-    from .modeling_tf_mbart import TFMBartForConditionalGeneration
+    from .models.mbart import TFMBartForConditionalGeneration
-    from .modeling_tf_mobilebert import (
+    from .models.mobilebert import (
         TF_MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFMobileBertForMaskedLM,
         TFMobileBertForMultipleChoice,
@@ -796,7 +791,7 @@ if is_tf_available():
         TFMobileBertModel,
         TFMobileBertPreTrainedModel,
     )
-    from .modeling_tf_openai import (
+    from .models.openai import (
         TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFOpenAIGPTDoubleHeadsModel,
         TFOpenAIGPTLMHeadModel,
@@ -804,8 +799,8 @@ if is_tf_available():
         TFOpenAIGPTModel,
         TFOpenAIGPTPreTrainedModel,
     )
-    from .modeling_tf_pegasus import TFPegasusForConditionalGeneration
+    from .models.pegasus import TFPegasusForConditionalGeneration
-    from .modeling_tf_roberta import (
+    from .models.roberta import (
         TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFRobertaForMaskedLM,
         TFRobertaForMultipleChoice,
@@ -816,13 +811,13 @@ if is_tf_available():
         TFRobertaModel,
         TFRobertaPreTrainedModel,
     )
-    from .modeling_tf_t5 import (
+    from .models.t5 import (
         TF_T5_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFT5ForConditionalGeneration,
         TFT5Model,
         TFT5PreTrainedModel,
     )
-    from .modeling_tf_transfo_xl import (
+    from .models.transfo_xl import (
         TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFAdaptiveEmbedding,
         TFTransfoXLLMHeadModel,
@@ -830,8 +825,7 @@ if is_tf_available():
         TFTransfoXLModel,
         TFTransfoXLPreTrainedModel,
     )
-    from .modeling_tf_utils import TFPreTrainedModel, TFSequenceSummary, TFSharedEmbeddings, shape_list
+    from .models.xlm import (
-    from .modeling_tf_xlm import (
         TF_XLM_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFXLMForMultipleChoice,
         TFXLMForQuestionAnsweringSimple,
@@ -842,7 +836,7 @@ if is_tf_available():
         TFXLMPreTrainedModel,
         TFXLMWithLMHeadModel,
     )
-    from .modeling_tf_xlm_roberta import (
+    from .models.xlm_roberta import (
         TF_XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFXLMRobertaForMaskedLM,
         TFXLMRobertaForMultipleChoice,
@@ -851,7 +845,7 @@ if is_tf_available():
         TFXLMRobertaForTokenClassification,
         TFXLMRobertaModel,
     )
-    from .modeling_tf_xlnet import (
+    from .models.xlnet import (
         TF_XLNET_PRETRAINED_MODEL_ARCHIVE_LIST,
         TFXLNetForMultipleChoice,
         TFXLNetForQuestionAnsweringSimple,
@@ -876,8 +870,8 @@ else:
 
 
 if is_flax_available():
-    from .modeling_flax_bert import FlaxBertModel
+    from .models.bert import FlaxBertModel
-    from .modeling_flax_roberta import FlaxRobertaModel
+    from .models.roberta import FlaxRobertaModel
 else:
     # Import the same objects as dummies to get them in the namespace.
     # They will raise an import error if the user tries to instantiate / use them.

src/transformers/benchmark/benchmark.py

@@ -23,7 +23,7 @@ from typing import Callable, Optional
 
 from ..configuration_utils import PretrainedConfig
 from ..file_utils import is_py3nvml_available, is_torch_available
-from ..modeling_auto import MODEL_MAPPING, MODEL_WITH_LM_HEAD_MAPPING
+from ..models.auto.modeling_auto import MODEL_MAPPING, MODEL_WITH_LM_HEAD_MAPPING
 from ..utils import logging
 from .benchmark_utils import (
     Benchmark,

src/transformers/benchmark/benchmark_tf.py

@@ -25,7 +25,7 @@ from typing import Callable, Optional
 
 from ..configuration_utils import PretrainedConfig
 from ..file_utils import is_py3nvml_available, is_tf_available
-from ..modeling_tf_auto import TF_MODEL_MAPPING, TF_MODEL_WITH_LM_HEAD_MAPPING
+from ..models.auto.modeling_tf_auto import TF_MODEL_MAPPING, TF_MODEL_WITH_LM_HEAD_MAPPING
 from ..utils import logging
 from .benchmark_utils import (
     Benchmark,

src/transformers/benchmark/benchmark_utils.py

@@ -327,7 +327,7 @@ def start_memory_tracing(
 
         - `modules_to_trace`: (None, string, list/tuple of string) if None, all events are recorded if string or list
           of strings: only events from the listed module/sub-module will be recorded (e.g. 'fairseq' or
-          'transformers.modeling_gpt2')
+          'transformers.models.gpt2.modeling_gpt2')
         - `modules_not_to_trace`: (None, string, list/tuple of string) if None, no module is avoided if string or list
           of strings: events from the listed module/sub-module will not be recorded (e.g. 'torch')
         - `events_to_trace`: string or list of string of events to be recorded (see official python doc for

src/transformers/commands/convert.py

@@ -73,7 +73,7 @@ class ConvertCommand(BaseTransformersCLICommand):
     def run(self):
         if self._model_type == "albert":
             try:
-                from transformers.convert_albert_original_tf_checkpoint_to_pytorch import (
+                from transformers.models.albert.convert_albert_original_tf_checkpoint_to_pytorch import (
                     convert_tf_checkpoint_to_pytorch,
                 )
             except ImportError:
@@ -82,7 +82,7 @@ class ConvertCommand(BaseTransformersCLICommand):
             convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output)
         elif self._model_type == "bert":
             try:
-                from transformers.convert_bert_original_tf_checkpoint_to_pytorch import (
+                from transformers.models.bert.convert_bert_original_tf_checkpoint_to_pytorch import (
                     convert_tf_checkpoint_to_pytorch,
                 )
             except ImportError:
@@ -91,7 +91,7 @@ class ConvertCommand(BaseTransformersCLICommand):
             convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output)
         elif self._model_type == "funnel":
             try:
-                from transformers.convert_funnel_original_tf_checkpoint_to_pytorch import (
+                from transformers.models.funnel.convert_funnel_original_tf_checkpoint_to_pytorch import (
                     convert_tf_checkpoint_to_pytorch,
                 )
             except ImportError:
@@ -99,14 +99,14 @@ class ConvertCommand(BaseTransformersCLICommand):
 
             convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output)
         elif self._model_type == "gpt":
-            from transformers.convert_openai_original_tf_checkpoint_to_pytorch import (
+            from transformers.models.openai.convert_openai_original_tf_checkpoint_to_pytorch import (
                 convert_openai_checkpoint_to_pytorch,
             )
 
             convert_openai_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output)
         elif self._model_type == "transfo_xl":
             try:
-                from transformers.convert_transfo_xl_original_tf_checkpoint_to_pytorch import (
+                from transformers.models.transfo_xl.convert_transfo_xl_original_tf_checkpoint_to_pytorch import (
                     convert_transfo_xl_checkpoint_to_pytorch,
                 )
             except ImportError:
@@ -123,7 +123,7 @@ class ConvertCommand(BaseTransformersCLICommand):
             )
         elif self._model_type == "gpt2":
             try:
-                from transformers.convert_gpt2_original_tf_checkpoint_to_pytorch import (
+                from transformers.models.gpt2.convert_gpt2_original_tf_checkpoint_to_pytorch import (
                     convert_gpt2_checkpoint_to_pytorch,
                 )
             except ImportError:
@@ -132,7 +132,7 @@ class ConvertCommand(BaseTransformersCLICommand):
             convert_gpt2_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output)
         elif self._model_type == "xlnet":
             try:
-                from transformers.convert_xlnet_original_tf_checkpoint_to_pytorch import (
+                from transformers.models.xlnet.convert_xlnet_original_tf_checkpoint_to_pytorch import (
                     convert_xlnet_checkpoint_to_pytorch,
                 )
             except ImportError:
@@ -142,13 +142,13 @@ class ConvertCommand(BaseTransformersCLICommand):
                 self._tf_checkpoint, self._config, self._pytorch_dump_output, self._finetuning_task_name
             )
         elif self._model_type == "xlm":
-            from transformers.convert_xlm_original_pytorch_checkpoint_to_pytorch import (
+            from transformers.models.xlm.convert_xlm_original_pytorch_checkpoint_to_pytorch import (
                 convert_xlm_checkpoint_to_pytorch,
             )
 
             convert_xlm_checkpoint_to_pytorch(self._tf_checkpoint, self._pytorch_dump_output)
         elif self._model_type == "lxmert":
-            from transformers.convert_lxmert_original_pytorch_checkpoint_to_pytorch import (
+            from transformers.models.lxmert.convert_lxmert_original_pytorch_checkpoint_to_pytorch import (
                 convert_lxmert_checkpoint_to_pytorch,
             )
 

src/transformers/data/datasets/squad.py

@@ -9,7 +9,7 @@ from torch.utils.data.dataset import Dataset
 
 from filelock import FileLock
 
-from ...modeling_auto import MODEL_FOR_QUESTION_ANSWERING_MAPPING
+from ...models.auto.modeling_auto import MODEL_FOR_QUESTION_ANSWERING_MAPPING
 from ...tokenization_utils import PreTrainedTokenizer
 from ...utils import logging
 from ..processors.squad import SquadFeatures, SquadV1Processor, SquadV2Processor, squad_convert_examples_to_features

src/transformers/data/metrics/squad_metrics.py

@@ -14,7 +14,7 @@ import math
 import re
 import string
 
-from transformers.tokenization_bert import BasicTokenizer
+from transformers import BasicTokenizer
 
 from ...utils import logging
 

src/transformers/data/processors/squad.py

@@ -7,7 +7,7 @@ import numpy as np
 from tqdm import tqdm
 
 from ...file_utils import is_tf_available, is_torch_available
-from ...tokenization_bert import whitespace_tokenize
+from ...models.bert.tokenization_bert import whitespace_tokenize
 from ...tokenization_utils_base import BatchEncoding, PreTrainedTokenizerBase, TruncationStrategy
 from ...utils import logging
 from .utils import DataProcessor

src/transformers/modelcard.py

@@ -19,7 +19,6 @@ import copy
 import json
 import os
 
-from .configuration_auto import ALL_PRETRAINED_CONFIG_ARCHIVE_MAP
 from .file_utils import (
     CONFIG_NAME,
     MODEL_CARD_NAME,
@@ -29,6 +28,7 @@ from .file_utils import (
     hf_bucket_url,
     is_remote_url,
 )
+from .models.auto.configuration_auto import ALL_PRETRAINED_CONFIG_ARCHIVE_MAP
 from .utils import logging
 
 

src/transformers/models/albert/__init__.py

@@ -0,0 +1,41 @@
+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from ...file_utils import is_sentencepiece_available, is_tf_available, is_tokenizers_available, is_torch_available
+from .configuration_albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig
+
+
+if is_sentencepiece_available():
+    from .tokenization_albert import AlbertTokenizer
+
+if is_tokenizers_available():
+    from .tokenization_albert_fast import AlbertTokenizerFast
+
+if is_torch_available():
+    from .modeling_albert import (
+        ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
+        AlbertForMaskedLM,
+        AlbertForMultipleChoice,
+        AlbertForPreTraining,
+        AlbertForQuestionAnswering,
+        AlbertForSequenceClassification,
+        AlbertForTokenClassification,
+        AlbertModel,
+        AlbertPreTrainedModel,
+        load_tf_weights_in_albert,
+    )
+
+if is_tf_available():
+    from .modeling_tf_albert import (
+        TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
+        TFAlbertForMaskedLM,
+        TFAlbertForMultipleChoice,
+        TFAlbertForPreTraining,
+        TFAlbertForQuestionAnswering,
+        TFAlbertForSequenceClassification,
+        TFAlbertForTokenClassification,
+        TFAlbertMainLayer,
+        TFAlbertModel,
+        TFAlbertPreTrainedModel,
+    )

src/transformers/models/albert/configuration_albert.py

@@ -15,7 +15,7 @@
 # limitations under the License.
 """ ALBERT model configuration """
 
-from .configuration_utils import PretrainedConfig
+from ...configuration_utils import PretrainedConfig
 
 
 ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP = {

src/transformers/models/albert/modeling_albert.py

@@ -24,16 +24,15 @@ import torch
 import torch.nn as nn
 from torch.nn import CrossEntropyLoss, MSELoss
 
-from .activations import ACT2FN
+from ...activations import ACT2FN
-from .configuration_albert import AlbertConfig
+from ...file_utils import (
-from .file_utils import (
     ModelOutput,
     add_code_sample_docstrings,
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_outputs import (
+from ...modeling_outputs import (
     BaseModelOutput,
     BaseModelOutputWithPooling,
     MaskedLMOutput,
@@ -42,13 +41,14 @@ from .modeling_outputs import (
     SequenceClassifierOutput,
     TokenClassifierOutput,
 )
-from .modeling_utils import (
+from ...modeling_utils import (
     PreTrainedModel,
     apply_chunking_to_forward,
     find_pruneable_heads_and_indices,
     prune_linear_layer,
 )
-from .utils import logging
+from ...utils import logging
+from .configuration_albert import AlbertConfig
 
 
 logger = logging.get_logger(__name__)
@@ -216,7 +216,7 @@ class AlbertEmbeddings(nn.Module):
         # position_ids (1, len position emb) is contiguous in memory and exported when serialized
         self.register_buffer("position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)))
 
-    # Copied from transformers.modeling_bert.BertEmbeddings.forward
+    # Copied from transformers.models.bert.modeling_bert.BertEmbeddings.forward
     def forward(self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None):
         if input_ids is not None:
             input_shape = input_ids.size()
@@ -266,7 +266,7 @@ class AlbertAttention(nn.Module):
         self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
         self.pruned_heads = set()
 
-    # Copied from transformers.modeling_bert.BertSelfAttention.transpose_for_scores
+    # Copied from transformers.models.bert.modeling_bert.BertSelfAttention.transpose_for_scores
     def transpose_for_scores(self, x):
         new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
         x = x.view(*new_x_shape)

src/transformers/models/albert/modeling_tf_albert.py

@@ -21,9 +21,8 @@ from typing import Optional, Tuple
 
 import tensorflow as tf
 
-from .activations_tf import get_tf_activation
+from ...activations_tf import get_tf_activation
-from .configuration_albert import AlbertConfig
+from ...file_utils import (
-from .file_utils import (
     MULTIPLE_CHOICE_DUMMY_INPUTS,
     ModelOutput,
     add_code_sample_docstrings,
@@ -31,7 +30,7 @@ from .file_utils import (
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_tf_outputs import (
+from ...modeling_tf_outputs import (
     TFBaseModelOutput,
     TFBaseModelOutputWithPooling,
     TFMaskedLMOutput,
@@ -40,7 +39,7 @@ from .modeling_tf_outputs import (
     TFSequenceClassifierOutput,
     TFTokenClassifierOutput,
 )
-from .modeling_tf_utils import (
+from ...modeling_tf_utils import (
     TFMaskedLanguageModelingLoss,
     TFMultipleChoiceLoss,
     TFPreTrainedModel,
@@ -51,8 +50,9 @@ from .modeling_tf_utils import (
     keras_serializable,
     shape_list,
 )
-from .tokenization_utils import BatchEncoding
+from ...tokenization_utils import BatchEncoding
-from .utils import logging
+from ...utils import logging
+from .configuration_albert import AlbertConfig
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/albert/tokenization_albert.py

@@ -22,8 +22,8 @@ from typing import List, Optional, Tuple
 
 import sentencepiece as spm
 
-from .tokenization_utils import PreTrainedTokenizer
+from ...tokenization_utils import PreTrainedTokenizer
-from .utils import logging
+from ...utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/albert/tokenization_albert_fast.py

@@ -19,9 +19,9 @@ import os
 from shutil import copyfile
 from typing import List, Optional, Tuple
 
-from .file_utils import is_sentencepiece_available
+from ...file_utils import is_sentencepiece_available
-from .tokenization_utils_fast import PreTrainedTokenizerFast
+from ...tokenization_utils_fast import PreTrainedTokenizerFast
-from .utils import logging
+from ...utils import logging
 
 
 if is_sentencepiece_available():

src/transformers/models/auto/__init__.py

@@ -0,0 +1,59 @@
+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from ...file_utils import is_tf_available, is_torch_available
+from .configuration_auto import ALL_PRETRAINED_CONFIG_ARCHIVE_MAP, CONFIG_MAPPING, AutoConfig
+from .tokenization_auto import TOKENIZER_MAPPING, AutoTokenizer
+
+
+if is_torch_available():
+    from .modeling_auto import (
+        MODEL_FOR_CAUSAL_LM_MAPPING,
+        MODEL_FOR_MASKED_LM_MAPPING,
+        MODEL_FOR_MULTIPLE_CHOICE_MAPPING,
+        MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING,
+        MODEL_FOR_PRETRAINING_MAPPING,
+        MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+        MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
+        MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
+        MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
+        MODEL_MAPPING,
+        MODEL_WITH_LM_HEAD_MAPPING,
+        AutoModel,
+        AutoModelForCausalLM,
+        AutoModelForMaskedLM,
+        AutoModelForMultipleChoice,
+        AutoModelForNextSentencePrediction,
+        AutoModelForPreTraining,
+        AutoModelForQuestionAnswering,
+        AutoModelForSeq2SeqLM,
+        AutoModelForSequenceClassification,
+        AutoModelForTokenClassification,
+        AutoModelWithLMHead,
+    )
+
+if is_tf_available():
+    from .modeling_tf_auto import (
+        TF_MODEL_FOR_CAUSAL_LM_MAPPING,
+        TF_MODEL_FOR_MASKED_LM_MAPPING,
+        TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING,
+        TF_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING,
+        TF_MODEL_FOR_PRETRAINING_MAPPING,
+        TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+        TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
+        TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
+        TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
+        TF_MODEL_MAPPING,
+        TF_MODEL_WITH_LM_HEAD_MAPPING,
+        TFAutoModel,
+        TFAutoModelForCausalLM,
+        TFAutoModelForMaskedLM,
+        TFAutoModelForMultipleChoice,
+        TFAutoModelForPreTraining,
+        TFAutoModelForQuestionAnswering,
+        TFAutoModelForSeq2SeqLM,
+        TFAutoModelForSequenceClassification,
+        TFAutoModelForTokenClassification,
+        TFAutoModelWithLMHead,
+    )

src/transformers/models/auto/configuration_auto.py

@@ -17,43 +17,46 @@
 import re
 from collections import OrderedDict
 
-from .configuration_albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig
+from ...configuration_utils import PretrainedConfig
-from .configuration_bart import BART_PRETRAINED_CONFIG_ARCHIVE_MAP, BartConfig
+from ..albert.configuration_albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig
-from .configuration_bert import BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, BertConfig
+from ..bart.configuration_bart import BART_PRETRAINED_CONFIG_ARCHIVE_MAP, BartConfig
-from .configuration_bert_generation import BertGenerationConfig
+from ..bert.configuration_bert import BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, BertConfig
-from .configuration_blenderbot import BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP, BlenderbotConfig
+from ..bert_generation.configuration_bert_generation import BertGenerationConfig
-from .configuration_camembert import CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, CamembertConfig
+from ..blenderbot.configuration_blenderbot import BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP, BlenderbotConfig
-from .configuration_ctrl import CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP, CTRLConfig
+from ..camembert.configuration_camembert import CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, CamembertConfig
-from .configuration_deberta import DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, DebertaConfig
+from ..ctrl.configuration_ctrl import CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP, CTRLConfig
-from .configuration_distilbert import DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, DistilBertConfig
+from ..deberta.configuration_deberta import DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, DebertaConfig
-from .configuration_dpr import DPR_PRETRAINED_CONFIG_ARCHIVE_MAP, DPRConfig
+from ..distilbert.configuration_distilbert import DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, DistilBertConfig
-from .configuration_electra import ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP, ElectraConfig
+from ..dpr.configuration_dpr import DPR_PRETRAINED_CONFIG_ARCHIVE_MAP, DPRConfig
-from .configuration_encoder_decoder import EncoderDecoderConfig
+from ..electra.configuration_electra import ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP, ElectraConfig
-from .configuration_flaubert import FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, FlaubertConfig
+from ..encoder_decoder.configuration_encoder_decoder import EncoderDecoderConfig
-from .configuration_fsmt import FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP, FSMTConfig
+from ..flaubert.configuration_flaubert import FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, FlaubertConfig
-from .configuration_funnel import FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP, FunnelConfig
+from ..fsmt.configuration_fsmt import FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP, FSMTConfig
-from .configuration_gpt2 import GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP, GPT2Config
+from ..funnel.configuration_funnel import FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP, FunnelConfig
-from .configuration_layoutlm import LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP, LayoutLMConfig
+from ..gpt2.configuration_gpt2 import GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP, GPT2Config
-from .configuration_longformer import LONGFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, LongformerConfig
+from ..layoutlm.configuration_layoutlm import LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP, LayoutLMConfig
-from .configuration_lxmert import LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP, LxmertConfig
+from ..longformer.configuration_longformer import LONGFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, LongformerConfig
-from .configuration_marian import MarianConfig
+from ..lxmert.configuration_lxmert import LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP, LxmertConfig
-from .configuration_mbart import MBART_PRETRAINED_CONFIG_ARCHIVE_MAP, MBartConfig
+from ..marian.configuration_marian import MarianConfig
-from .configuration_mobilebert import MobileBertConfig
+from ..mbart.configuration_mbart import MBART_PRETRAINED_CONFIG_ARCHIVE_MAP, MBartConfig
-from .configuration_openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig
+from ..mobilebert.configuration_mobilebert import MobileBertConfig
-from .configuration_pegasus import PegasusConfig
+from ..openai.configuration_openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig
-from .configuration_prophetnet import PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, ProphetNetConfig
+from ..pegasus.configuration_pegasus import PegasusConfig
-from .configuration_rag import RagConfig
+from ..prophetnet.configuration_prophetnet import PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, ProphetNetConfig
-from .configuration_reformer import ReformerConfig
+from ..rag.configuration_rag import RagConfig
-from .configuration_retribert import RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, RetriBertConfig
+from ..reformer.configuration_reformer import ReformerConfig
-from .configuration_roberta import ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, RobertaConfig
+from ..retribert.configuration_retribert import RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, RetriBertConfig
-from .configuration_squeezebert import SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, SqueezeBertConfig
+from ..roberta.configuration_roberta import ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, RobertaConfig
-from .configuration_t5 import T5_PRETRAINED_CONFIG_ARCHIVE_MAP, T5Config
+from ..squeezebert.configuration_squeezebert import SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, SqueezeBertConfig
-from .configuration_transfo_xl import TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP, TransfoXLConfig
+from ..t5.configuration_t5 import T5_PRETRAINED_CONFIG_ARCHIVE_MAP, T5Config
-from .configuration_utils import PretrainedConfig
+from ..transfo_xl.configuration_transfo_xl import TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP, TransfoXLConfig
-from .configuration_xlm import XLM_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMConfig
+from ..xlm.configuration_xlm import XLM_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMConfig
-from .configuration_xlm_prophetnet import XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMProphetNetConfig
+from ..xlm_prophetnet.configuration_xlm_prophetnet import (
-from .configuration_xlm_roberta import XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMRobertaConfig
+    XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
-from .configuration_xlnet import XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLNetConfig
+    XLMProphetNetConfig,
+)
+from ..xlm_roberta.configuration_xlm_roberta import XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMRobertaConfig
+from ..xlnet.configuration_xlnet import XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLNetConfig
 
 
 ALL_PRETRAINED_CONFIG_ARCHIVE_MAP = dict(

src/transformers/models/auto/modeling_auto.py

@@ -18,6 +18,173 @@
 import warnings
 from collections import OrderedDict
 
+from ...configuration_utils import PretrainedConfig
+from ...file_utils import add_start_docstrings
+from ...utils import logging
+from ..albert.modeling_albert import (
+    AlbertForMaskedLM,
+    AlbertForMultipleChoice,
+    AlbertForPreTraining,
+    AlbertForQuestionAnswering,
+    AlbertForSequenceClassification,
+    AlbertForTokenClassification,
+    AlbertModel,
+)
+from ..bart.modeling_bart import (
+    BartForConditionalGeneration,
+    BartForQuestionAnswering,
+    BartForSequenceClassification,
+    BartModel,
+)
+from ..bert.modeling_bert import (
+    BertForMaskedLM,
+    BertForMultipleChoice,
+    BertForNextSentencePrediction,
+    BertForPreTraining,
+    BertForQuestionAnswering,
+    BertForSequenceClassification,
+    BertForTokenClassification,
+    BertLMHeadModel,
+    BertModel,
+)
+from ..bert_generation.modeling_bert_generation import BertGenerationDecoder, BertGenerationEncoder
+from ..blenderbot.modeling_blenderbot import BlenderbotForConditionalGeneration
+from ..camembert.modeling_camembert import (
+    CamembertForCausalLM,
+    CamembertForMaskedLM,
+    CamembertForMultipleChoice,
+    CamembertForQuestionAnswering,
+    CamembertForSequenceClassification,
+    CamembertForTokenClassification,
+    CamembertModel,
+)
+from ..ctrl.modeling_ctrl import CTRLLMHeadModel, CTRLModel
+from ..deberta.modeling_deberta import DebertaForSequenceClassification, DebertaModel
+from ..distilbert.modeling_distilbert import (
+    DistilBertForMaskedLM,
+    DistilBertForMultipleChoice,
+    DistilBertForQuestionAnswering,
+    DistilBertForSequenceClassification,
+    DistilBertForTokenClassification,
+    DistilBertModel,
+)
+from ..dpr.modeling_dpr import DPRQuestionEncoder
+from ..electra.modeling_electra import (
+    ElectraForMaskedLM,
+    ElectraForMultipleChoice,
+    ElectraForPreTraining,
+    ElectraForQuestionAnswering,
+    ElectraForSequenceClassification,
+    ElectraForTokenClassification,
+    ElectraModel,
+)
+from ..encoder_decoder.modeling_encoder_decoder import EncoderDecoderModel
+from ..flaubert.modeling_flaubert import (
+    FlaubertForMultipleChoice,
+    FlaubertForQuestionAnsweringSimple,
+    FlaubertForSequenceClassification,
+    FlaubertForTokenClassification,
+    FlaubertModel,
+    FlaubertWithLMHeadModel,
+)
+from ..fsmt.modeling_fsmt import FSMTForConditionalGeneration, FSMTModel
+from ..funnel.modeling_funnel import (
+    FunnelForMaskedLM,
+    FunnelForMultipleChoice,
+    FunnelForPreTraining,
+    FunnelForQuestionAnswering,
+    FunnelForSequenceClassification,
+    FunnelForTokenClassification,
+    FunnelModel,
+)
+from ..gpt2.modeling_gpt2 import GPT2ForSequenceClassification, GPT2LMHeadModel, GPT2Model
+from ..layoutlm.modeling_layoutlm import LayoutLMForMaskedLM, LayoutLMForTokenClassification, LayoutLMModel
+from ..longformer.modeling_longformer import (
+    LongformerForMaskedLM,
+    LongformerForMultipleChoice,
+    LongformerForQuestionAnswering,
+    LongformerForSequenceClassification,
+    LongformerForTokenClassification,
+    LongformerModel,
+)
+from ..lxmert.modeling_lxmert import LxmertForPreTraining, LxmertForQuestionAnswering, LxmertModel
+from ..marian.modeling_marian import MarianMTModel
+from ..mbart.modeling_mbart import MBartForConditionalGeneration
+from ..mobilebert.modeling_mobilebert import (
+    MobileBertForMaskedLM,
+    MobileBertForMultipleChoice,
+    MobileBertForNextSentencePrediction,
+    MobileBertForPreTraining,
+    MobileBertForQuestionAnswering,
+    MobileBertForSequenceClassification,
+    MobileBertForTokenClassification,
+    MobileBertModel,
+)
+from ..openai.modeling_openai import OpenAIGPTForSequenceClassification, OpenAIGPTLMHeadModel, OpenAIGPTModel
+from ..pegasus.modeling_pegasus import PegasusForConditionalGeneration
+from ..prophetnet.modeling_prophetnet import ProphetNetForCausalLM, ProphetNetForConditionalGeneration, ProphetNetModel
+from ..rag.modeling_rag import (  # noqa: F401 - need to import all RagModels to be in globals() function
+    RagModel,
+    RagSequenceForGeneration,
+    RagTokenForGeneration,
+)
+from ..reformer.modeling_reformer import (
+    ReformerForMaskedLM,
+    ReformerForQuestionAnswering,
+    ReformerForSequenceClassification,
+    ReformerModel,
+    ReformerModelWithLMHead,
+)
+from ..retribert.modeling_retribert import RetriBertModel
+from ..roberta.modeling_roberta import (
+    RobertaForCausalLM,
+    RobertaForMaskedLM,
+    RobertaForMultipleChoice,
+    RobertaForQuestionAnswering,
+    RobertaForSequenceClassification,
+    RobertaForTokenClassification,
+    RobertaModel,
+)
+from ..squeezebert.modeling_squeezebert import (
+    SqueezeBertForMaskedLM,
+    SqueezeBertForMultipleChoice,
+    SqueezeBertForQuestionAnswering,
+    SqueezeBertForSequenceClassification,
+    SqueezeBertForTokenClassification,
+    SqueezeBertModel,
+)
+from ..t5.modeling_t5 import T5ForConditionalGeneration, T5Model
+from ..transfo_xl.modeling_transfo_xl import TransfoXLLMHeadModel, TransfoXLModel
+from ..xlm.modeling_xlm import (
+    XLMForMultipleChoice,
+    XLMForQuestionAnsweringSimple,
+    XLMForSequenceClassification,
+    XLMForTokenClassification,
+    XLMModel,
+    XLMWithLMHeadModel,
+)
+from ..xlm_prophetnet.modeling_xlm_prophetnet import (
+    XLMProphetNetForCausalLM,
+    XLMProphetNetForConditionalGeneration,
+    XLMProphetNetModel,
+)
+from ..xlm_roberta.modeling_xlm_roberta import (
+    XLMRobertaForCausalLM,
+    XLMRobertaForMaskedLM,
+    XLMRobertaForMultipleChoice,
+    XLMRobertaForQuestionAnswering,
+    XLMRobertaForSequenceClassification,
+    XLMRobertaForTokenClassification,
+    XLMRobertaModel,
+)
+from ..xlnet.modeling_xlnet import (
+    XLNetForMultipleChoice,
+    XLNetForQuestionAnsweringSimple,
+    XLNetForSequenceClassification,
+    XLNetForTokenClassification,
+    XLNetLMHeadModel,
+    XLNetModel,
+)
 from .configuration_auto import (
     AlbertConfig,
     AutoConfig,
@@ -39,6 +206,7 @@ from .configuration_auto import (
     LayoutLMConfig,
     LongformerConfig,
     LxmertConfig,
+    MarianConfig,
     MBartConfig,
     MobileBertConfig,
     OpenAIGPTConfig,
@@ -56,174 +224,6 @@ from .configuration_auto import (
     XLNetConfig,
     replace_list_option_in_docstrings,
 )
-from .configuration_marian import MarianConfig
-from .configuration_utils import PretrainedConfig
-from .file_utils import add_start_docstrings
-from .modeling_albert import (
-    AlbertForMaskedLM,
-    AlbertForMultipleChoice,
-    AlbertForPreTraining,
-    AlbertForQuestionAnswering,
-    AlbertForSequenceClassification,
-    AlbertForTokenClassification,
-    AlbertModel,
-)
-from .modeling_bart import (
-    BartForConditionalGeneration,
-    BartForQuestionAnswering,
-    BartForSequenceClassification,
-    BartModel,
-)
-from .modeling_bert import (
-    BertForMaskedLM,
-    BertForMultipleChoice,
-    BertForNextSentencePrediction,
-    BertForPreTraining,
-    BertForQuestionAnswering,
-    BertForSequenceClassification,
-    BertForTokenClassification,
-    BertLMHeadModel,
-    BertModel,
-)
-from .modeling_bert_generation import BertGenerationDecoder, BertGenerationEncoder
-from .modeling_blenderbot import BlenderbotForConditionalGeneration
-from .modeling_camembert import (
-    CamembertForCausalLM,
-    CamembertForMaskedLM,
-    CamembertForMultipleChoice,
-    CamembertForQuestionAnswering,
-    CamembertForSequenceClassification,
-    CamembertForTokenClassification,
-    CamembertModel,
-)
-from .modeling_ctrl import CTRLLMHeadModel, CTRLModel
-from .modeling_deberta import DebertaForSequenceClassification, DebertaModel
-from .modeling_distilbert import (
-    DistilBertForMaskedLM,
-    DistilBertForMultipleChoice,
-    DistilBertForQuestionAnswering,
-    DistilBertForSequenceClassification,
-    DistilBertForTokenClassification,
-    DistilBertModel,
-)
-from .modeling_dpr import DPRQuestionEncoder
-from .modeling_electra import (
-    ElectraForMaskedLM,
-    ElectraForMultipleChoice,
-    ElectraForPreTraining,
-    ElectraForQuestionAnswering,
-    ElectraForSequenceClassification,
-    ElectraForTokenClassification,
-    ElectraModel,
-)
-from .modeling_encoder_decoder import EncoderDecoderModel
-from .modeling_flaubert import (
-    FlaubertForMultipleChoice,
-    FlaubertForQuestionAnsweringSimple,
-    FlaubertForSequenceClassification,
-    FlaubertForTokenClassification,
-    FlaubertModel,
-    FlaubertWithLMHeadModel,
-)
-from .modeling_fsmt import FSMTForConditionalGeneration, FSMTModel
-from .modeling_funnel import (
-    FunnelForMaskedLM,
-    FunnelForMultipleChoice,
-    FunnelForPreTraining,
-    FunnelForQuestionAnswering,
-    FunnelForSequenceClassification,
-    FunnelForTokenClassification,
-    FunnelModel,
-)
-from .modeling_gpt2 import GPT2ForSequenceClassification, GPT2LMHeadModel, GPT2Model
-from .modeling_layoutlm import LayoutLMForMaskedLM, LayoutLMForTokenClassification, LayoutLMModel
-from .modeling_longformer import (
-    LongformerForMaskedLM,
-    LongformerForMultipleChoice,
-    LongformerForQuestionAnswering,
-    LongformerForSequenceClassification,
-    LongformerForTokenClassification,
-    LongformerModel,
-)
-from .modeling_lxmert import LxmertForPreTraining, LxmertForQuestionAnswering, LxmertModel
-from .modeling_marian import MarianMTModel
-from .modeling_mbart import MBartForConditionalGeneration
-from .modeling_mobilebert import (
-    MobileBertForMaskedLM,
-    MobileBertForMultipleChoice,
-    MobileBertForNextSentencePrediction,
-    MobileBertForPreTraining,
-    MobileBertForQuestionAnswering,
-    MobileBertForSequenceClassification,
-    MobileBertForTokenClassification,
-    MobileBertModel,
-)
-from .modeling_openai import OpenAIGPTForSequenceClassification, OpenAIGPTLMHeadModel, OpenAIGPTModel
-from .modeling_pegasus import PegasusForConditionalGeneration
-from .modeling_prophetnet import ProphetNetForCausalLM, ProphetNetForConditionalGeneration, ProphetNetModel
-from .modeling_rag import (  # noqa: F401 - need to import all RagModels to be in globals() function
-    RagModel,
-    RagSequenceForGeneration,
-    RagTokenForGeneration,
-)
-from .modeling_reformer import (
-    ReformerForMaskedLM,
-    ReformerForQuestionAnswering,
-    ReformerForSequenceClassification,
-    ReformerModel,
-    ReformerModelWithLMHead,
-)
-from .modeling_retribert import RetriBertModel
-from .modeling_roberta import (
-    RobertaForCausalLM,
-    RobertaForMaskedLM,
-    RobertaForMultipleChoice,
-    RobertaForQuestionAnswering,
-    RobertaForSequenceClassification,
-    RobertaForTokenClassification,
-    RobertaModel,
-)
-from .modeling_squeezebert import (
-    SqueezeBertForMaskedLM,
-    SqueezeBertForMultipleChoice,
-    SqueezeBertForQuestionAnswering,
-    SqueezeBertForSequenceClassification,
-    SqueezeBertForTokenClassification,
-    SqueezeBertModel,
-)
-from .modeling_t5 import T5ForConditionalGeneration, T5Model
-from .modeling_transfo_xl import TransfoXLLMHeadModel, TransfoXLModel
-from .modeling_xlm import (
-    XLMForMultipleChoice,
-    XLMForQuestionAnsweringSimple,
-    XLMForSequenceClassification,
-    XLMForTokenClassification,
-    XLMModel,
-    XLMWithLMHeadModel,
-)
-from .modeling_xlm_prophetnet import (
-    XLMProphetNetForCausalLM,
-    XLMProphetNetForConditionalGeneration,
-    XLMProphetNetModel,
-)
-from .modeling_xlm_roberta import (
-    XLMRobertaForCausalLM,
-    XLMRobertaForMaskedLM,
-    XLMRobertaForMultipleChoice,
-    XLMRobertaForQuestionAnswering,
-    XLMRobertaForSequenceClassification,
-    XLMRobertaForTokenClassification,
-    XLMRobertaModel,
-)
-from .modeling_xlnet import (
-    XLNetForMultipleChoice,
-    XLNetForQuestionAnsweringSimple,
-    XLNetForSequenceClassification,
-    XLNetForTokenClassification,
-    XLNetLMHeadModel,
-    XLNetModel,
-)
-from .utils import logging
 
 
 # Add modeling imports here

src/transformers/models/auto/modeling_flax_auto.py

@@ -17,11 +17,11 @@
 
 from collections import OrderedDict
 
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+from ..bert.modeling_flax_bert import FlaxBertModel
+from ..roberta.modeling_flax_roberta import FlaxRobertaModel
 from .configuration_auto import AutoConfig, BertConfig, RobertaConfig
-from .configuration_utils import PretrainedConfig
-from .modeling_flax_bert import FlaxBertModel
-from .modeling_flax_roberta import FlaxRobertaModel
-from .utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/auto/modeling_tf_auto.py

@@ -18,38 +18,10 @@
 import warnings
 from collections import OrderedDict
 
-from .configuration_auto import (
+from ...configuration_utils import PretrainedConfig
-    AlbertConfig,
+from ...file_utils import add_start_docstrings
-    AutoConfig,
+from ...utils import logging
-    BartConfig,
+from ..albert.modeling_tf_albert import (
-    BertConfig,
-    CamembertConfig,
-    CTRLConfig,
-    DistilBertConfig,
-    ElectraConfig,
-    FlaubertConfig,
-    FunnelConfig,
-    GPT2Config,
-    LongformerConfig,
-    LxmertConfig,
-    MobileBertConfig,
-    OpenAIGPTConfig,
-    RobertaConfig,
-    T5Config,
-    TransfoXLConfig,
-    XLMConfig,
-    XLMRobertaConfig,
-    XLNetConfig,
-    replace_list_option_in_docstrings,
-)
-from .configuration_blenderbot import BlenderbotConfig
-from .configuration_dpr import DPRConfig
-from .configuration_marian import MarianConfig
-from .configuration_mbart import MBartConfig
-from .configuration_pegasus import PegasusConfig
-from .configuration_utils import PretrainedConfig
-from .file_utils import add_start_docstrings
-from .modeling_tf_albert import (
     TFAlbertForMaskedLM,
     TFAlbertForMultipleChoice,
     TFAlbertForPreTraining,
@@ -58,8 +30,8 @@ from .modeling_tf_albert import (
     TFAlbertForTokenClassification,
     TFAlbertModel,
 )
-from .modeling_tf_bart import TFBartForConditionalGeneration, TFBartModel
+from ..bart.modeling_tf_bart import TFBartForConditionalGeneration, TFBartModel
-from .modeling_tf_bert import (
+from ..bert.modeling_tf_bert import (
     TFBertForMaskedLM,
     TFBertForMultipleChoice,
     TFBertForNextSentencePrediction,
@@ -70,8 +42,8 @@ from .modeling_tf_bert import (
     TFBertLMHeadModel,
     TFBertModel,
 )
-from .modeling_tf_blenderbot import TFBlenderbotForConditionalGeneration
+from ..blenderbot.modeling_tf_blenderbot import TFBlenderbotForConditionalGeneration
-from .modeling_tf_camembert import (
+from ..camembert.modeling_tf_camembert import (
     TFCamembertForMaskedLM,
     TFCamembertForMultipleChoice,
     TFCamembertForQuestionAnswering,
@@ -79,8 +51,8 @@ from .modeling_tf_camembert import (
     TFCamembertForTokenClassification,
     TFCamembertModel,
 )
-from .modeling_tf_ctrl import TFCTRLLMHeadModel, TFCTRLModel
+from ..ctrl.modeling_tf_ctrl import TFCTRLLMHeadModel, TFCTRLModel
-from .modeling_tf_distilbert import (
+from ..distilbert.modeling_tf_distilbert import (
     TFDistilBertForMaskedLM,
     TFDistilBertForMultipleChoice,
     TFDistilBertForQuestionAnswering,
@@ -88,8 +60,8 @@ from .modeling_tf_distilbert import (
     TFDistilBertForTokenClassification,
     TFDistilBertModel,
 )
-from .modeling_tf_dpr import TFDPRQuestionEncoder
+from ..dpr.modeling_tf_dpr import TFDPRQuestionEncoder
-from .modeling_tf_electra import (
+from ..electra.modeling_tf_electra import (
     TFElectraForMaskedLM,
     TFElectraForMultipleChoice,
     TFElectraForPreTraining,
@@ -98,7 +70,7 @@ from .modeling_tf_electra import (
     TFElectraForTokenClassification,
     TFElectraModel,
 )
-from .modeling_tf_flaubert import (
+from ..flaubert.modeling_tf_flaubert import (
     TFFlaubertForMultipleChoice,
     TFFlaubertForQuestionAnsweringSimple,
     TFFlaubertForSequenceClassification,
@@ -106,7 +78,7 @@ from .modeling_tf_flaubert import (
     TFFlaubertModel,
     TFFlaubertWithLMHeadModel,
 )
-from .modeling_tf_funnel import (
+from ..funnel.modeling_tf_funnel import (
     TFFunnelForMaskedLM,
     TFFunnelForMultipleChoice,
     TFFunnelForPreTraining,
@@ -115,12 +87,16 @@ from .modeling_tf_funnel import (
     TFFunnelForTokenClassification,
     TFFunnelModel,
 )
-from .modeling_tf_gpt2 import TFGPT2LMHeadModel, TFGPT2Model
+from ..gpt2.modeling_tf_gpt2 import TFGPT2LMHeadModel, TFGPT2Model
-from .modeling_tf_longformer import TFLongformerForMaskedLM, TFLongformerForQuestionAnswering, TFLongformerModel
+from ..longformer.modeling_tf_longformer import (
-from .modeling_tf_lxmert import TFLxmertForPreTraining, TFLxmertModel
+    TFLongformerForMaskedLM,
-from .modeling_tf_marian import TFMarianMTModel
+    TFLongformerForQuestionAnswering,
-from .modeling_tf_mbart import TFMBartForConditionalGeneration
+    TFLongformerModel,
-from .modeling_tf_mobilebert import (
+)
+from ..lxmert.modeling_tf_lxmert import TFLxmertForPreTraining, TFLxmertModel
+from ..marian.modeling_tf_marian import TFMarianMTModel
+from ..mbart.modeling_tf_mbart import TFMBartForConditionalGeneration
+from ..mobilebert.modeling_tf_mobilebert import (
     TFMobileBertForMaskedLM,
     TFMobileBertForMultipleChoice,
     TFMobileBertForNextSentencePrediction,
@@ -130,9 +106,9 @@ from .modeling_tf_mobilebert import (
     TFMobileBertForTokenClassification,
     TFMobileBertModel,
 )
-from .modeling_tf_openai import TFOpenAIGPTLMHeadModel, TFOpenAIGPTModel
+from ..openai.modeling_tf_openai import TFOpenAIGPTLMHeadModel, TFOpenAIGPTModel
-from .modeling_tf_pegasus import TFPegasusForConditionalGeneration
+from ..pegasus.modeling_tf_pegasus import TFPegasusForConditionalGeneration
-from .modeling_tf_roberta import (
+from ..roberta.modeling_tf_roberta import (
     TFRobertaForMaskedLM,
     TFRobertaForMultipleChoice,
     TFRobertaForQuestionAnswering,
@@ -140,9 +116,9 @@ from .modeling_tf_roberta import (
     TFRobertaForTokenClassification,
     TFRobertaModel,
 )
-from .modeling_tf_t5 import TFT5ForConditionalGeneration, TFT5Model
+from ..t5.modeling_tf_t5 import TFT5ForConditionalGeneration, TFT5Model
-from .modeling_tf_transfo_xl import TFTransfoXLLMHeadModel, TFTransfoXLModel
+from ..transfo_xl.modeling_tf_transfo_xl import TFTransfoXLLMHeadModel, TFTransfoXLModel
-from .modeling_tf_xlm import (
+from ..xlm.modeling_tf_xlm import (
     TFXLMForMultipleChoice,
     TFXLMForQuestionAnsweringSimple,
     TFXLMForSequenceClassification,
@@ -150,7 +126,7 @@ from .modeling_tf_xlm import (
     TFXLMModel,
     TFXLMWithLMHeadModel,
 )
-from .modeling_tf_xlm_roberta import (
+from ..xlm_roberta.modeling_tf_xlm_roberta import (
     TFXLMRobertaForMaskedLM,
     TFXLMRobertaForMultipleChoice,
     TFXLMRobertaForQuestionAnswering,
@@ -158,7 +134,7 @@ from .modeling_tf_xlm_roberta import (
     TFXLMRobertaForTokenClassification,
     TFXLMRobertaModel,
 )
-from .modeling_tf_xlnet import (
+from ..xlnet.modeling_tf_xlnet import (
     TFXLNetForMultipleChoice,
     TFXLNetForQuestionAnsweringSimple,
     TFXLNetForSequenceClassification,
@@ -166,7 +142,35 @@ from .modeling_tf_xlnet import (
     TFXLNetLMHeadModel,
     TFXLNetModel,
 )
-from .utils import logging
+from .configuration_auto import (
+    AlbertConfig,
+    AutoConfig,
+    BartConfig,
+    BertConfig,
+    BlenderbotConfig,
+    CamembertConfig,
+    CTRLConfig,
+    DistilBertConfig,
+    DPRConfig,
+    ElectraConfig,
+    FlaubertConfig,
+    FunnelConfig,
+    GPT2Config,
+    LongformerConfig,
+    LxmertConfig,
+    MarianConfig,
+    MBartConfig,
+    MobileBertConfig,
+    OpenAIGPTConfig,
+    PegasusConfig,
+    RobertaConfig,
+    T5Config,
+    TransfoXLConfig,
+    XLMConfig,
+    XLMRobertaConfig,
+    XLNetConfig,
+    replace_list_option_in_docstrings,
+)
 
 
 # Add modeling imports here

src/transformers/models/auto/tokenization_auto.py

@@ -17,6 +17,37 @@
 
 from collections import OrderedDict
 
+from ...configuration_utils import PretrainedConfig
+from ...file_utils import is_sentencepiece_available, is_tokenizers_available
+from ...utils import logging
+from ..bart.tokenization_bart import BartTokenizer
+from ..bert.tokenization_bert import BertTokenizer
+from ..bert_japanese.tokenization_bert_japanese import BertJapaneseTokenizer
+from ..bertweet.tokenization_bertweet import BertweetTokenizer
+from ..blenderbot.tokenization_blenderbot import BlenderbotSmallTokenizer
+from ..ctrl.tokenization_ctrl import CTRLTokenizer
+from ..deberta.tokenization_deberta import DebertaTokenizer
+from ..distilbert.tokenization_distilbert import DistilBertTokenizer
+from ..dpr.tokenization_dpr import DPRQuestionEncoderTokenizer
+from ..electra.tokenization_electra import ElectraTokenizer
+from ..flaubert.tokenization_flaubert import FlaubertTokenizer
+from ..fsmt.tokenization_fsmt import FSMTTokenizer
+from ..funnel.tokenization_funnel import FunnelTokenizer
+from ..gpt2.tokenization_gpt2 import GPT2Tokenizer
+from ..herbert.tokenization_herbert import HerbertTokenizer
+from ..layoutlm.tokenization_layoutlm import LayoutLMTokenizer
+from ..longformer.tokenization_longformer import LongformerTokenizer
+from ..lxmert.tokenization_lxmert import LxmertTokenizer
+from ..mobilebert.tokenization_mobilebert import MobileBertTokenizer
+from ..openai.tokenization_openai import OpenAIGPTTokenizer
+from ..phobert.tokenization_phobert import PhobertTokenizer
+from ..prophetnet.tokenization_prophetnet import ProphetNetTokenizer
+from ..rag.tokenization_rag import RagTokenizer
+from ..retribert.tokenization_retribert import RetriBertTokenizer
+from ..roberta.tokenization_roberta import RobertaTokenizer
+from ..squeezebert.tokenization_squeezebert import SqueezeBertTokenizer
+from ..transfo_xl.tokenization_transfo_xl import TransfoXLTokenizer
+from ..xlm.tokenization_xlm import XLMTokenizer
 from .configuration_auto import (
     AlbertConfig,
     AutoConfig,
@@ -57,51 +88,20 @@ from .configuration_auto import (
     XLNetConfig,
     replace_list_option_in_docstrings,
 )
-from .configuration_utils import PretrainedConfig
-from .file_utils import is_sentencepiece_available, is_tokenizers_available
-from .tokenization_bart import BartTokenizer
-from .tokenization_bert import BertTokenizer
-from .tokenization_bert_japanese import BertJapaneseTokenizer
-from .tokenization_bertweet import BertweetTokenizer
-from .tokenization_blenderbot import BlenderbotSmallTokenizer
-from .tokenization_ctrl import CTRLTokenizer
-from .tokenization_deberta import DebertaTokenizer
-from .tokenization_distilbert import DistilBertTokenizer
-from .tokenization_dpr import DPRQuestionEncoderTokenizer
-from .tokenization_electra import ElectraTokenizer
-from .tokenization_flaubert import FlaubertTokenizer
-from .tokenization_fsmt import FSMTTokenizer
-from .tokenization_funnel import FunnelTokenizer
-from .tokenization_gpt2 import GPT2Tokenizer
-from .tokenization_herbert import HerbertTokenizer
-from .tokenization_layoutlm import LayoutLMTokenizer
-from .tokenization_longformer import LongformerTokenizer
-from .tokenization_lxmert import LxmertTokenizer
-from .tokenization_mobilebert import MobileBertTokenizer
-from .tokenization_openai import OpenAIGPTTokenizer
-from .tokenization_phobert import PhobertTokenizer
-from .tokenization_prophetnet import ProphetNetTokenizer
-from .tokenization_rag import RagTokenizer
-from .tokenization_retribert import RetriBertTokenizer
-from .tokenization_roberta import RobertaTokenizer
-from .tokenization_squeezebert import SqueezeBertTokenizer
-from .tokenization_transfo_xl import TransfoXLTokenizer
-from .tokenization_xlm import XLMTokenizer
-from .utils import logging
 
 
 if is_sentencepiece_available():
-    from .tokenization_albert import AlbertTokenizer
+    from ..albert.tokenization_albert import AlbertTokenizer
-    from .tokenization_bert_generation import BertGenerationTokenizer
+    from ..bert_generation.tokenization_bert_generation import BertGenerationTokenizer
-    from .tokenization_camembert import CamembertTokenizer
+    from ..camembert.tokenization_camembert import CamembertTokenizer
-    from .tokenization_marian import MarianTokenizer
+    from ..marian.tokenization_marian import MarianTokenizer
-    from .tokenization_mbart import MBartTokenizer
+    from ..mbart.tokenization_mbart import MBartTokenizer
-    from .tokenization_pegasus import PegasusTokenizer
+    from ..pegasus.tokenization_pegasus import PegasusTokenizer
-    from .tokenization_reformer import ReformerTokenizer
+    from ..reformer.tokenization_reformer import ReformerTokenizer
-    from .tokenization_t5 import T5Tokenizer
+    from ..t5.tokenization_t5 import T5Tokenizer
-    from .tokenization_xlm_prophetnet import XLMProphetNetTokenizer
+    from ..xlm_prophetnet.tokenization_xlm_prophetnet import XLMProphetNetTokenizer
-    from .tokenization_xlm_roberta import XLMRobertaTokenizer
+    from ..xlm_roberta.tokenization_xlm_roberta import XLMRobertaTokenizer
-    from .tokenization_xlnet import XLNetTokenizer
+    from ..xlnet.tokenization_xlnet import XLNetTokenizer
 else:
     AlbertTokenizer = None
     BertGenerationTokenizer = None
@@ -116,30 +116,30 @@ else:
     XLMProphetNetTokenizer = None
 
 if is_tokenizers_available():
-    from .tokenization_albert_fast import AlbertTokenizerFast
+    from ..albert.tokenization_albert_fast import AlbertTokenizerFast
-    from .tokenization_bart_fast import BartTokenizerFast
+    from ..bart.tokenization_bart_fast import BartTokenizerFast
-    from .tokenization_bert_fast import BertTokenizerFast
+    from ..bert.tokenization_bert_fast import BertTokenizerFast
-    from .tokenization_camembert_fast import CamembertTokenizerFast
+    from ..camembert.tokenization_camembert_fast import CamembertTokenizerFast
-    from .tokenization_distilbert_fast import DistilBertTokenizerFast
+    from ..distilbert.tokenization_distilbert_fast import DistilBertTokenizerFast
-    from .tokenization_dpr_fast import DPRQuestionEncoderTokenizerFast
+    from ..dpr.tokenization_dpr_fast import DPRQuestionEncoderTokenizerFast
-    from .tokenization_electra_fast import ElectraTokenizerFast
+    from ..electra.tokenization_electra_fast import ElectraTokenizerFast
-    from .tokenization_funnel_fast import FunnelTokenizerFast
+    from ..funnel.tokenization_funnel_fast import FunnelTokenizerFast
-    from .tokenization_gpt2_fast import GPT2TokenizerFast
+    from ..gpt2.tokenization_gpt2_fast import GPT2TokenizerFast
-    from .tokenization_herbert_fast import HerbertTokenizerFast
+    from ..herbert.tokenization_herbert_fast import HerbertTokenizerFast
-    from .tokenization_layoutlm_fast import LayoutLMTokenizerFast
+    from ..layoutlm.tokenization_layoutlm_fast import LayoutLMTokenizerFast
-    from .tokenization_longformer_fast import LongformerTokenizerFast
+    from ..longformer.tokenization_longformer_fast import LongformerTokenizerFast
-    from .tokenization_lxmert_fast import LxmertTokenizerFast
+    from ..lxmert.tokenization_lxmert_fast import LxmertTokenizerFast
-    from .tokenization_mbart_fast import MBartTokenizerFast
+    from ..mbart.tokenization_mbart_fast import MBartTokenizerFast
-    from .tokenization_mobilebert_fast import MobileBertTokenizerFast
+    from ..mobilebert.tokenization_mobilebert_fast import MobileBertTokenizerFast
-    from .tokenization_openai_fast import OpenAIGPTTokenizerFast
+    from ..openai.tokenization_openai_fast import OpenAIGPTTokenizerFast
-    from .tokenization_pegasus_fast import PegasusTokenizerFast
+    from ..pegasus.tokenization_pegasus_fast import PegasusTokenizerFast
-    from .tokenization_reformer_fast import ReformerTokenizerFast
+    from ..reformer.tokenization_reformer_fast import ReformerTokenizerFast
-    from .tokenization_retribert_fast import RetriBertTokenizerFast
+    from ..retribert.tokenization_retribert_fast import RetriBertTokenizerFast
-    from .tokenization_roberta_fast import RobertaTokenizerFast
+    from ..roberta.tokenization_roberta_fast import RobertaTokenizerFast
-    from .tokenization_squeezebert_fast import SqueezeBertTokenizerFast
+    from ..squeezebert.tokenization_squeezebert_fast import SqueezeBertTokenizerFast
-    from .tokenization_t5_fast import T5TokenizerFast
+    from ..t5.tokenization_t5_fast import T5TokenizerFast
-    from .tokenization_xlm_roberta_fast import XLMRobertaTokenizerFast
+    from ..xlm_roberta.tokenization_xlm_roberta_fast import XLMRobertaTokenizerFast
-    from .tokenization_xlnet_fast import XLNetTokenizerFast
+    from ..xlnet.tokenization_xlnet_fast import XLNetTokenizerFast
 else:
     AlbertTokenizerFast = None
     BartTokenizerFast = None

src/transformers/models/bart/__init__.py

@@ -0,0 +1,24 @@
+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from ...file_utils import is_tf_available, is_tokenizers_available, is_torch_available
+from .configuration_bart import BartConfig
+from .tokenization_bart import BartTokenizer
+
+
+if is_tokenizers_available():
+    from .tokenization_bart_fast import BartTokenizerFast
+
+if is_torch_available():
+    from .modeling_bart import (
+        BART_PRETRAINED_MODEL_ARCHIVE_LIST,
+        BartForConditionalGeneration,
+        BartForQuestionAnswering,
+        BartForSequenceClassification,
+        BartModel,
+        PretrainedBartModel,
+    )
+
+if is_tf_available():
+    from .modeling_tf_bart import TFBartForConditionalGeneration, TFBartModel

src/transformers/models/bart/configuration_bart.py

@@ -14,8 +14,8 @@
 # limitations under the License.
 """ BART configuration """
 
-from .configuration_utils import PretrainedConfig
+from ...configuration_utils import PretrainedConfig
-from .utils import logging
+from ...utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bart/convert_bart_original_pytorch_checkpoint_to_pytorch.py

@@ -30,7 +30,7 @@ from transformers import (
     BartModel,
     BartTokenizer,
 )
-from transformers.modeling_bart import _make_linear_from_emb
+from transformers.models.bart.modeling_bart import _make_linear_from_emb
 from transformers.utils import logging
 
 

src/transformers/models/bart/modeling_bart.py

@@ -24,16 +24,15 @@ import torch.nn.functional as F
 from torch import Tensor, nn
 from torch.nn import CrossEntropyLoss
 
-from .activations import ACT2FN
+from ...activations import ACT2FN
-from .configuration_bart import BartConfig
+from ...file_utils import (
-from .file_utils import (
     add_code_sample_docstrings,
     add_end_docstrings,
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_outputs import (
+from ...modeling_outputs import (
     BaseModelOutput,
     BaseModelOutputWithPastAndCrossAttentions,
     Seq2SeqLMOutput,
@@ -41,8 +40,9 @@ from .modeling_outputs import (
     Seq2SeqQuestionAnsweringModelOutput,
     Seq2SeqSequenceClassifierOutput,
 )
-from .modeling_utils import PreTrainedModel
+from ...modeling_utils import PreTrainedModel
-from .utils import logging
+from ...utils import logging
+from .configuration_bart import BartConfig
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bart/modeling_tf_bart.py

@@ -24,13 +24,17 @@ import tensorflow as tf
 from tensorflow import Tensor
 from tensorflow.keras.layers import Dense, Layer, LayerNormalization
 
-from .activations_tf import ACT2FN
+from ...activations_tf import ACT2FN
-from .configuration_bart import BartConfig
+from ...file_utils import add_start_docstrings, add_start_docstrings_to_model_forward, replace_return_docstrings
-from .file_utils import add_start_docstrings, add_start_docstrings_to_model_forward, replace_return_docstrings
+from ...modeling_tf_outputs import (
-from .modeling_tf_outputs import TFBaseModelOutput, TFBaseModelOutputWithPast, TFSeq2SeqLMOutput, TFSeq2SeqModelOutput
+    TFBaseModelOutput,
+    TFBaseModelOutputWithPast,
+    TFSeq2SeqLMOutput,
+    TFSeq2SeqModelOutput,
+)
 
 # Public API
-from .modeling_tf_utils import (
+from ...modeling_tf_utils import (
     DUMMY_INPUTS,
     TFPreTrainedModel,
     TFSharedEmbeddings,
@@ -39,8 +43,9 @@ from .modeling_tf_utils import (
     keras_serializable,
     shape_list,
 )
-from .tokenization_utils_base import BatchEncoding
+from ...tokenization_utils_base import BatchEncoding
-from .utils import logging
+from ...utils import logging
+from .configuration_bart import BartConfig
 
 
 _CONFIG_FOR_DOC = "BartConfig"

src/transformers/models/bart/tokenization_bart.py

@@ -15,9 +15,9 @@
 
 from typing import List, Optional
 
-from .tokenization_roberta import RobertaTokenizer
+from ...tokenization_utils_base import BatchEncoding
-from .tokenization_utils_base import BatchEncoding
+from ...utils import logging
-from .utils import logging
+from ..roberta.tokenization_roberta import RobertaTokenizer
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bart/tokenization_bart_fast.py

@@ -15,10 +15,10 @@
 
 from typing import List, Optional
 
+from ...tokenization_utils_base import BatchEncoding
+from ...utils import logging
+from ..roberta.tokenization_roberta_fast import RobertaTokenizerFast
 from .tokenization_bart import BartTokenizer
-from .tokenization_roberta_fast import RobertaTokenizerFast
-from .tokenization_utils_base import BatchEncoding
-from .utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert/__init__.py

@@ -0,0 +1,48 @@
+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from ...file_utils import is_flax_available, is_tf_available, is_tokenizers_available, is_torch_available
+from .configuration_bert import BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, BertConfig
+from .tokenization_bert import BasicTokenizer, BertTokenizer, WordpieceTokenizer
+
+
+if is_tokenizers_available():
+    from .tokenization_bert_fast import BertTokenizerFast
+
+if is_torch_available():
+    from .modeling_bert import (
+        BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
+        BertForMaskedLM,
+        BertForMultipleChoice,
+        BertForNextSentencePrediction,
+        BertForPreTraining,
+        BertForQuestionAnswering,
+        BertForSequenceClassification,
+        BertForTokenClassification,
+        BertLayer,
+        BertLMHeadModel,
+        BertModel,
+        BertPreTrainedModel,
+        load_tf_weights_in_bert,
+    )
+
+if is_tf_available():
+    from .modeling_tf_bert import (
+        TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST,
+        TFBertEmbeddings,
+        TFBertForMaskedLM,
+        TFBertForMultipleChoice,
+        TFBertForNextSentencePrediction,
+        TFBertForPreTraining,
+        TFBertForQuestionAnswering,
+        TFBertForSequenceClassification,
+        TFBertForTokenClassification,
+        TFBertLMHeadModel,
+        TFBertMainLayer,
+        TFBertModel,
+        TFBertPreTrainedModel,
+    )
+
+if is_flax_available():
+    from .modeling_flax_bert import FlaxBertModel

src/transformers/models/bert/configuration_bert.py

@@ -15,8 +15,8 @@
 # limitations under the License.
 """ BERT model configuration """
 
-from .configuration_utils import PretrainedConfig
+from ...configuration_utils import PretrainedConfig
-from .utils import logging
+from ...utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert/modeling_bert.py

@@ -27,16 +27,15 @@ import torch.utils.checkpoint
 from torch import nn
 from torch.nn import CrossEntropyLoss, MSELoss
 
-from .activations import ACT2FN
+from ...activations import ACT2FN
-from .configuration_bert import BertConfig
+from ...file_utils import (
-from .file_utils import (
     ModelOutput,
     add_code_sample_docstrings,
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_outputs import (
+from ...modeling_outputs import (
     BaseModelOutputWithCrossAttentions,
     BaseModelOutputWithPoolingAndCrossAttentions,
     CausalLMOutputWithCrossAttentions,
@@ -47,13 +46,14 @@ from .modeling_outputs import (
     SequenceClassifierOutput,
     TokenClassifierOutput,
 )
-from .modeling_utils import (
+from ...modeling_utils import (
     PreTrainedModel,
     apply_chunking_to_forward,
     find_pruneable_heads_and_indices,
     prune_linear_layer,
 )
-from .utils import logging
+from ...utils import logging
+from .configuration_bert import BertConfig
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert/modeling_flax_bert.py

@@ -21,10 +21,10 @@ import flax.linen as nn
 import jax
 import jax.numpy as jnp
 
+from ...file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
+from ...modeling_flax_utils import FlaxPreTrainedModel, gelu
+from ...utils import logging
 from .configuration_bert import BertConfig
-from .file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
-from .modeling_flax_utils import FlaxPreTrainedModel, gelu
-from .utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert/modeling_tf_bert.py

@@ -21,9 +21,8 @@ from typing import Optional, Tuple
 
 import tensorflow as tf
 
-from .activations_tf import get_tf_activation
+from ...activations_tf import get_tf_activation
-from .configuration_bert import BertConfig
+from ...file_utils import (
-from .file_utils import (
     MULTIPLE_CHOICE_DUMMY_INPUTS,
     ModelOutput,
     add_code_sample_docstrings,
@@ -31,7 +30,7 @@ from .file_utils import (
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_tf_outputs import (
+from ...modeling_tf_outputs import (
     TFBaseModelOutput,
     TFBaseModelOutputWithPooling,
     TFCausalLMOutput,
@@ -42,7 +41,7 @@ from .modeling_tf_outputs import (
     TFSequenceClassifierOutput,
     TFTokenClassifierOutput,
 )
-from .modeling_tf_utils import (
+from ...modeling_tf_utils import (
     TFCausalLanguageModelingLoss,
     TFMaskedLanguageModelingLoss,
     TFMultipleChoiceLoss,
@@ -55,8 +54,9 @@ from .modeling_tf_utils import (
     keras_serializable,
     shape_list,
 )
-from .tokenization_utils import BatchEncoding
+from ...tokenization_utils import BatchEncoding
-from .utils import logging
+from ...utils import logging
+from .configuration_bert import BertConfig
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert/tokenization_bert.py

@@ -20,8 +20,8 @@ import os
 import unicodedata
 from typing import List, Optional, Tuple
 
-from .tokenization_utils import PreTrainedTokenizer, _is_control, _is_punctuation, _is_whitespace
+from ...tokenization_utils import PreTrainedTokenizer, _is_control, _is_punctuation, _is_whitespace
-from .utils import logging
+from ...utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert/tokenization_bert_fast.py

@@ -19,9 +19,9 @@ from typing import List, Optional, Tuple
 
 from tokenizers import normalizers
 
+from ...tokenization_utils_fast import PreTrainedTokenizerFast
+from ...utils import logging
 from .tokenization_bert import BertTokenizer
-from .tokenization_utils_fast import PreTrainedTokenizerFast
-from .utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert_generation/__init__.py

@@ -0,0 +1,17 @@
+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from ...file_utils import is_sentencepiece_available, is_torch_available
+from .configuration_bert_generation import BertGenerationConfig
+
+
+if is_sentencepiece_available():
+    from .tokenization_bert_generation import BertGenerationTokenizer
+
+if is_torch_available():
+    from .modeling_bert_generation import (
+        BertGenerationDecoder,
+        BertGenerationEncoder,
+        load_tf_weights_in_bert_generation,
+    )

src/transformers/models/bert_generation/configuration_bert_generation.py

@@ -14,7 +14,7 @@
 # limitations under the License.
 """  BertGeneration model configuration """
 
-from .configuration_utils import PretrainedConfig
+from ...configuration_utils import PretrainedConfig
 
 
 class BertGenerationConfig(PretrainedConfig):

src/transformers/models/bert_generation/modeling_bert_generation.py

@@ -20,17 +20,17 @@ import torch.utils.checkpoint
 from torch import nn
 from torch.nn import CrossEntropyLoss
 
-from .configuration_bert_generation import BertGenerationConfig
+from ...file_utils import (
-from .file_utils import (
     add_code_sample_docstrings,
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_bert import BertEncoder
+from ...modeling_outputs import BaseModelOutputWithCrossAttentions, CausalLMOutputWithCrossAttentions
-from .modeling_outputs import BaseModelOutputWithCrossAttentions, CausalLMOutputWithCrossAttentions
+from ...modeling_utils import PreTrainedModel
-from .modeling_utils import PreTrainedModel
+from ...utils import logging
-from .utils import logging
+from ..bert.modeling_bert import BertEncoder
+from .configuration_bert_generation import BertGenerationConfig
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert_generation/tokenization_bert_generation.py

@@ -21,8 +21,8 @@ from typing import List, Optional, Tuple
 
 import sentencepiece as spm
 
-from .tokenization_utils import PreTrainedTokenizer
+from ...tokenization_utils import PreTrainedTokenizer
-from .utils import logging
+from ...utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bert_japanese/__init__.py

@@ -0,0 +1,5 @@
+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from .tokenization_bert_japanese import BertJapaneseTokenizer, CharacterTokenizer, MecabTokenizer

src/transformers/models/bert_japanese/tokenization_bert_japanese.py

@@ -21,8 +21,8 @@ import os
 import unicodedata
 from typing import Optional
 
-from .tokenization_bert import BasicTokenizer, BertTokenizer, WordpieceTokenizer, load_vocab
+from ...utils import logging
-from .utils import logging
+from ..bert.tokenization_bert import BasicTokenizer, BertTokenizer, WordpieceTokenizer, load_vocab
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/bertweet/__init__.py

@@ -0,0 +1,5 @@
+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from .tokenization_bertweet import BertweetTokenizer

src/transformers/models/bertweet/tokenization_bertweet.py

@@ -24,8 +24,8 @@ from typing import List, Optional, Tuple
 
 import regex
 
-from .tokenization_utils import PreTrainedTokenizer
+from ...tokenization_utils import PreTrainedTokenizer
-from .utils import logging
+from ...utils import logging
 
 
 logger = logging.get_logger(__name__)