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

src/transformers/modeling_tf_xlnet.py → src/transformers/models/xlnet/modeling_tf_xlnet.py

@@ -23,9 +23,8 @@ from typing import List, Optional, Tuple
 
 import tensorflow as tf
 
-from .activations_tf import get_tf_activation
-from .configuration_xlnet import XLNetConfig
-from .file_utils import (
+from ...activations_tf import get_tf_activation
+from ...file_utils import (
     MULTIPLE_CHOICE_DUMMY_INPUTS,
     ModelOutput,
     add_code_sample_docstrings,
@@ -33,7 +32,7 @@ from .file_utils import (
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_tf_utils import (
+from ...modeling_tf_utils import (
     TFCausalLanguageModelingLoss,
     TFMultipleChoiceLoss,
     TFPreTrainedModel,
@@ -46,8 +45,9 @@ from .modeling_tf_utils import (
     keras_serializable,
     shape_list,
 )
-from .tokenization_utils import BatchEncoding
-from .utils import logging
+from ...tokenization_utils import BatchEncoding
+from ...utils import logging
+from .configuration_xlnet import XLNetConfig
 
 
 logger = logging.get_logger(__name__)

src/transformers/modeling_xlnet.py → src/transformers/models/xlnet/modeling_xlnet.py

@@ -24,16 +24,15 @@ from torch import nn
 from torch.nn import CrossEntropyLoss, MSELoss
 from torch.nn import functional as F
 
-from .activations import ACT2FN
-from .configuration_xlnet import XLNetConfig
-from .file_utils import (
+from ...activations import ACT2FN
+from ...file_utils import (
     ModelOutput,
     add_code_sample_docstrings,
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
 )
-from .modeling_utils import (
+from ...modeling_utils import (
     PoolerAnswerClass,
     PoolerEndLogits,
     PoolerStartLogits,
@@ -41,7 +40,8 @@ from .modeling_utils import (
     SequenceSummary,
     apply_chunking_to_forward,
 )
-from .utils import logging
+from ...utils import logging
+from .configuration_xlnet import XLNetConfig
 
 
 logger = logging.get_logger(__name__)

src/transformers/tokenization_xlnet.py → src/transformers/models/xlnet/tokenization_xlnet.py

@@ -22,9 +22,9 @@ from typing import List, Optional, Tuple
 
 import sentencepiece as spm
 
-from .file_utils import SPIECE_UNDERLINE
-from .tokenization_utils import PreTrainedTokenizer
-from .utils import logging
+from ...file_utils import SPIECE_UNDERLINE
+from ...tokenization_utils import PreTrainedTokenizer
+from ...utils import logging
 
 
 logger = logging.get_logger(__name__)

src/transformers/tokenization_xlnet_fast.py → src/transformers/models/xlnet/tokenization_xlnet_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 .tokenization_utils_fast import PreTrainedTokenizerFast
-from .utils import logging
+from ...file_utils import is_sentencepiece_available
+from ...tokenization_utils_fast import PreTrainedTokenizerFast
+from ...utils import logging
 
 
 if is_sentencepiece_available():

src/transformers/pipelines.py

@@ -30,13 +30,13 @@ from uuid import UUID
 
 import numpy as np
 
-from .configuration_auto import AutoConfig
 from .configuration_utils import PretrainedConfig
 from .data import SquadExample, SquadFeatures, squad_convert_examples_to_features
 from .file_utils import add_end_docstrings, is_tf_available, is_torch_available
 from .modelcard import ModelCard
-from .tokenization_auto import AutoTokenizer
-from .tokenization_bert import BasicTokenizer
+from .models.auto.configuration_auto import AutoConfig
+from .models.auto.tokenization_auto import AutoTokenizer
+from .models.bert.tokenization_bert import BasicTokenizer
 from .tokenization_utils import PreTrainedTokenizer
 from .tokenization_utils_base import PaddingStrategy
 from .utils import logging
@@ -45,7 +45,7 @@ from .utils import logging
 if is_tf_available():
     import tensorflow as tf
 
-    from .modeling_tf_auto import (
+    from .models.auto.modeling_tf_auto import (
         TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
         TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
         TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
@@ -63,7 +63,7 @@ if is_tf_available():
 if is_torch_available():
     import torch
 
-    from .modeling_auto import (
+    from .models.auto.modeling_auto import (
         MODEL_FOR_MASKED_LM_MAPPING,
         MODEL_FOR_QUESTION_ANSWERING_MAPPING,
         MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,

src/transformers/trainer.py

@@ -53,8 +53,8 @@ from torch.utils.data.sampler import RandomSampler, SequentialSampler
 
 from .data.data_collator import DataCollator, DataCollatorWithPadding, default_data_collator
 from .file_utils import WEIGHTS_NAME, is_datasets_available, is_in_notebook, is_torch_tpu_available
-from .modeling_auto import MODEL_FOR_QUESTION_ANSWERING_MAPPING
 from .modeling_utils import PreTrainedModel
+from .models.auto.modeling_auto import MODEL_FOR_QUESTION_ANSWERING_MAPPING
 from .optimization import AdamW, get_linear_schedule_with_warmup
 from .tokenization_utils_base import PreTrainedTokenizerBase
 from .trainer_callback import (

src/transformers/utils/dummy_pt_objects.py

@@ -172,6 +172,28 @@ def top_k_top_p_filtering(*args, **kwargs):
     requires_pytorch(top_k_top_p_filtering)
 
 
+class Conv1D:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class PreTrainedModel:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+def apply_chunking_to_forward(*args, **kwargs):
+    requires_pytorch(apply_chunking_to_forward)
+
+
+def prune_layer(*args, **kwargs):
+    requires_pytorch(prune_layer)
+
+
 ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 
@@ -1749,28 +1771,6 @@ def load_tf_weights_in_transfo_xl(*args, **kwargs):
     requires_pytorch(load_tf_weights_in_transfo_xl)
 
 
-class Conv1D:
-    def __init__(self, *args, **kwargs):
-        requires_pytorch(self)
-
-
-class PreTrainedModel:
-    def __init__(self, *args, **kwargs):
-        requires_pytorch(self)
-
-    @classmethod
-    def from_pretrained(self, *args, **kwargs):
-        requires_pytorch(self)
-
-
-def apply_chunking_to_forward(*args, **kwargs):
-    requires_pytorch(apply_chunking_to_forward)
-
-
-def prune_layer(*args, **kwargs):
-    requires_pytorch(prune_layer)
-
-
 XLM_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 

src/transformers/utils/dummy_tf_objects.py

@@ -16,6 +16,29 @@ def tf_top_k_top_p_filtering(*args, **kwargs):
     requires_tf(tf_top_k_top_p_filtering)
 
 
+class TFPreTrainedModel:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFSequenceSummary:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFSharedEmbeddings:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+
+def shape_list(*args, **kwargs):
+    requires_tf(shape_list)
+
+
 TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 
@@ -1141,29 +1164,6 @@ class TFTransfoXLPreTrainedModel:
         requires_tf(self)
 
 
-class TFPreTrainedModel:
-    def __init__(self, *args, **kwargs):
-        requires_tf(self)
-
-    @classmethod
-    def from_pretrained(self, *args, **kwargs):
-        requires_tf(self)
-
-
-class TFSequenceSummary:
-    def __init__(self, *args, **kwargs):
-        requires_tf(self)
-
-
-class TFSharedEmbeddings:
-    def __init__(self, *args, **kwargs):
-        requires_tf(self)
-
-
-def shape_list(*args, **kwargs):
-    requires_tf(shape_list)
-
-
 TF_XLM_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 

src/transformers/utils/dummy_tokenizers_objects.py

@@ -218,7 +218,7 @@ class T5TokenizerFast:
         requires_tokenizers(self)
 
 
-class PreTrainedTokenizerFast:
+class XLMRobertaTokenizerFast:
     def __init__(self, *args, **kwargs):
         requires_tokenizers(self)
 
@@ -227,7 +227,7 @@ class PreTrainedTokenizerFast:
         requires_tokenizers(self)
 
 
-class XLMRobertaTokenizerFast:
+class XLNetTokenizerFast:
     def __init__(self, *args, **kwargs):
         requires_tokenizers(self)
 
@@ -236,7 +236,7 @@ class XLMRobertaTokenizerFast:
         requires_tokenizers(self)
 
 
-class XLNetTokenizerFast:
+class PreTrainedTokenizerFast:
     def __init__(self, *args, **kwargs):
         requires_tokenizers(self)
 

templates/adding_a_new_model/README.md

@@ -66,10 +66,10 @@ Choose from 1, 2 [1]:
 Once the command has finished, you should have a total of 7 new files spread across the repository:
 ```
 docs/source/model_doc/<model_name>.rst
-src/transformers/configuration_<model_name>.py
-src/transformers/modeling_<model_name>.py
-src/transformers/modeling_tf_<model_name>.py
-src/transformers/tokenization_<model_name>.py
+src/transformers/models/<model_name>/configuration_<model_name>.py
+src/transformers/models/<model_name>/modeling_<model_name>.py
+src/transformers/models/<model_name>/modeling_tf_<model_name>.py
+src/transformers/models/<model_name>/tokenization_<model_name>.py
 tests/test_modeling_<model_name>.py
 tests/test_modeling_tf_<model_name>.py
 ```

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_tf_{{cookiecutter.lowercase_modelname}}.py

@@ -61,7 +61,7 @@ TF_{{cookiecutter.uppercase_modelname}}_PRETRAINED_MODEL_ARCHIVE_LIST = [
 ]
 
 
-# Copied from transformers.modeling_tf_bert.TFBertEmbeddings with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertEmbeddings with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}Embeddings(tf.keras.layers.Layer):
     """Construct the embeddings from word, position and token_type embeddings."""
 
@@ -175,7 +175,7 @@ class TF{{cookiecutter.camelcase_modelname}}Embeddings(tf.keras.layers.Layer):
         return tf.reshape(logits, [batch_size, length, self.vocab_size])
 
 
-# Copied from transformers.modeling_tf_bert.TFBertSelfAttention with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertSelfAttention with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}SelfAttention(tf.keras.layers.Layer):
     def __init__(self, config, **kwargs):
         super().__init__(**kwargs)
@@ -247,7 +247,7 @@ class TF{{cookiecutter.camelcase_modelname}}SelfAttention(tf.keras.layers.Layer)
         return outputs
 
 
-# Copied from transformers.modeling_tf_bert.TFBertSelfOutput with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertSelfOutput with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}SelfOutput(tf.keras.layers.Layer):
     def __init__(self, config, **kwargs):
         super().__init__(**kwargs)
@@ -266,7 +266,7 @@ class TF{{cookiecutter.camelcase_modelname}}SelfOutput(tf.keras.layers.Layer):
         return hidden_states
 
 
-# Copied from transformers.modeling_tf_bert.TFBertAttention with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertAttention with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}Attention(tf.keras.layers.Layer):
     def __init__(self, config, **kwargs):
         super().__init__(**kwargs)
@@ -287,7 +287,7 @@ class TF{{cookiecutter.camelcase_modelname}}Attention(tf.keras.layers.Layer):
         return outputs
 
 
-# Copied from transformers.modeling_tf_bert.TFBertIntermediate with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertIntermediate with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}Intermediate(tf.keras.layers.Layer):
     def __init__(self, config, **kwargs):
         super().__init__(**kwargs)
@@ -308,7 +308,7 @@ class TF{{cookiecutter.camelcase_modelname}}Intermediate(tf.keras.layers.Layer):
         return hidden_states
 
 
-# Copied from transformers.modeling_tf_bert.TFBertOutput with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertOutput with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}Output(tf.keras.layers.Layer):
     def __init__(self, config, **kwargs):
         super().__init__(**kwargs)
@@ -327,7 +327,7 @@ class TF{{cookiecutter.camelcase_modelname}}Output(tf.keras.layers.Layer):
         return hidden_states
 
 
-# Copied from transformers.modeling_tf_bert.TFBertLayer with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertLayer with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}Layer(tf.keras.layers.Layer):
     def __init__(self, config, **kwargs):
         super().__init__(**kwargs)
@@ -391,7 +391,7 @@ class TF{{cookiecutter.camelcase_modelname}}Encoder(tf.keras.layers.Layer):
         )
 
 
-# Copied from transformers.modeling_tf_bert.TFBertPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}PredictionHeadTransform(tf.keras.layers.Layer):
     def __init__(self, config, **kwargs):
         super().__init__(**kwargs)
@@ -415,7 +415,7 @@ class TF{{cookiecutter.camelcase_modelname}}PredictionHeadTransform(tf.keras.lay
         return hidden_states
 
 
-# Copied from transformers.modeling_tf_bert.TFBertLMPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertLMPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}LMPredictionHead(tf.keras.layers.Layer):
     def __init__(self, config, input_embeddings, **kwargs):
         super().__init__(**kwargs)
@@ -440,7 +440,7 @@ class TF{{cookiecutter.camelcase_modelname}}LMPredictionHead(tf.keras.layers.Lay
         return hidden_states
 
 
-# Copied from transformers.modeling_tf_bert.TFBertMLMHead with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertMLMHead with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}MLMHead(tf.keras.layers.Layer):
     def __init__(self, config, input_embeddings, **kwargs):
         super().__init__(**kwargs)
@@ -606,7 +606,7 @@ class TF{{cookiecutter.camelcase_modelname}}MainLayer(tf.keras.layers.Layer):
         )
 
 
-# Copied from transformers.modeling_tf_bert.TFBertPreTrainedModel with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertPreTrainedModel with Bert->{{cookiecutter.camelcase_modelname}}
 class TF{{cookiecutter.camelcase_modelname}}PreTrainedModel(TFPreTrainedModel):
     """An abstract class to handle weights initialization and
     a simple interface for downloading and loading pretrained models.

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_{{cookiecutter.lowercase_modelname}}.py

@@ -140,7 +140,7 @@ def mish(x):
     return x * torch.tanh(nn.functional.softplus(x))
 
 
-# Copied from transformers.modeling_bert.BertEmbeddings with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertEmbeddings with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}Embeddings(nn.Module):
     """Construct the embeddings from word, position and token_type embeddings."""
 
@@ -183,7 +183,7 @@ class {{cookiecutter.camelcase_modelname}}Embeddings(nn.Module):
         return embeddings
 
 
-# Copied from transformers.modeling_bert.BertSelfAttention with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertSelfAttention with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}SelfAttention(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -262,7 +262,7 @@ class {{cookiecutter.camelcase_modelname}}SelfAttention(nn.Module):
         return outputs
 
 
-# Copied from transformers.modeling_bert.BertSelfOutput with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertSelfOutput with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}SelfOutput(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -277,7 +277,7 @@ class {{cookiecutter.camelcase_modelname}}SelfOutput(nn.Module):
         return hidden_states
 
 
-# Copied from transformers.modeling_bert.BertAttention with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertAttention with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}Attention(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -325,7 +325,7 @@ class {{cookiecutter.camelcase_modelname}}Attention(nn.Module):
         return outputs
 
 
-# Copied from transformers.modeling_bert.BertIntermediate with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertIntermediate with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}Intermediate(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -341,7 +341,7 @@ class {{cookiecutter.camelcase_modelname}}Intermediate(nn.Module):
         return hidden_states
 
 
-# Copied from transformers.modeling_bert.BertOutput with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertOutput with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}Output(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -356,7 +356,7 @@ class {{cookiecutter.camelcase_modelname}}Output(nn.Module):
         return hidden_states
 
 
-# Copied from transformers.modeling_bert.BertLayer with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertLayer with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}Layer(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -416,7 +416,7 @@ class {{cookiecutter.camelcase_modelname}}Layer(nn.Module):
         return layer_output
 
 
-# Copied from transformers.modeling_bert.BertEncoder with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertEncoder with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}Encoder(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -481,7 +481,7 @@ class {{cookiecutter.camelcase_modelname}}Encoder(nn.Module):
         )
 
 
-# Copied from transformers.modeling_bert.BertPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}PredictionHeadTransform(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -499,7 +499,7 @@ class {{cookiecutter.camelcase_modelname}}PredictionHeadTransform(nn.Module):
         return hidden_states
 
 
-# Copied from transformers.modeling_bert.BertLMPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertLMPredictionHead with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}LMPredictionHead(nn.Module):
     def __init__(self, config):
         super().__init__()
@@ -520,7 +520,7 @@ class {{cookiecutter.camelcase_modelname}}LMPredictionHead(nn.Module):
         return hidden_states
 
 
-# Copied from transformers.modeling_bert.BertOnlyMLMHead with Bert->{{cookiecutter.camelcase_modelname}}
+# Copied from transformers.models.bert.modeling_bert.BertOnlyMLMHead with Bert->{{cookiecutter.camelcase_modelname}}
 class {{cookiecutter.camelcase_modelname}}OnlyMLMHead(nn.Module):
     def __init__(self, config):
         super().__init__()

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/test_modeling_tf_{{cookiecutter.lowercase_modelname}}.py

@@ -26,7 +26,7 @@ from .test_modeling_tf_common import TFModelTesterMixin, ids_tensor
 if is_tf_available():
     import tensorflow as tf
 
-    from transformers.modeling_tf_{{cookiecutter.lowercase_modelname}} import (
+    from transformers import (
         TF{{cookiecutter.camelcase_modelname}}ForMaskedLM,
         TF{{cookiecutter.camelcase_modelname}}ForMultipleChoice,
         TF{{cookiecutter.camelcase_modelname}}ForQuestionAnswering,

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/test_modeling_{{cookiecutter.lowercase_modelname}}.py

@@ -34,7 +34,7 @@ if is_torch_available():
         {{cookiecutter.camelcase_modelname}}ForTokenClassification,
         {{cookiecutter.camelcase_modelname}}Model,
     )
-    from transformers.modeling_{{cookiecutter.lowercase_modelname}} import {{cookiecutter.uppercase_modelname}}_PRETRAINED_MODEL_ARCHIVE_LIST
+    from transformers.models.{{cookiecutter.lowercase_modelname}}.modeling_{{cookiecutter.lowercase_modelname}} import {{cookiecutter.uppercase_modelname}}_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
 class {{cookiecutter.camelcase_modelname}}ModelTester:

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/to_replace_{{cookiecutter.lowercase_modelname}}.py

@@ -74,7 +74,7 @@ from .configuration_{{cookiecutter.lowercase_modelname}} import {{cookiecutter.u
 
 
 
-# To replace in: "src/transformers/modeling_auto.py" if generating PyTorch
+# To replace in: "src/transformers/models/auto/modeling_auto.py" if generating PyTorch
 # Below: "from .configuration_auto import ("
 # Replace with:
     {{cookiecutter.camelcase_modelname}}Config,
@@ -129,7 +129,7 @@ from .modeling_{{cookiecutter.lowercase_modelname}} import (
 # End.
 
 
-# To replace in: "src/transformers/modeling_tf_auto.py" if generating TensorFlow
+# To replace in: "src/transformers/models/auto/modeling_tf_auto.py" if generating TensorFlow
 # Below: "from .configuration_auto import ("
 # Replace with:
     {{cookiecutter.camelcase_modelname}}Config,

tests/test_configuration_auto.py

@@ -16,9 +16,9 @@
 import os
 import unittest
 
-from transformers.configuration_auto import CONFIG_MAPPING, AutoConfig
-from transformers.configuration_bert import BertConfig
-from transformers.configuration_roberta import RobertaConfig
+from transformers.models.auto.configuration_auto import CONFIG_MAPPING, AutoConfig
+from transformers.models.bert.configuration_bert import BertConfig
+from transformers.models.roberta.configuration_roberta import RobertaConfig
 from transformers.testing_utils import DUMMY_UNKWOWN_IDENTIFIER
 
 

tests/test_flax_auto.py

@@ -6,9 +6,9 @@ from transformers.testing_utils import require_flax, slow
 
 if is_flax_available():
     import jax
-    from transformers.modeling_flax_auto import FlaxAutoModel
-    from transformers.modeling_flax_bert import FlaxBertModel
-    from transformers.modeling_flax_roberta import FlaxRobertaModel
+    from transformers.models.auto.modeling_flax_auto import FlaxAutoModel
+    from transformers.models.bert.modeling_flax_bert import FlaxBertModel
+    from transformers.models.roberta.modeling_flax_roberta import FlaxRobertaModel
 
 
 @require_flax

tests/test_logging.py

 import os
 import unittest
 
-import transformers.tokenization_bart
+import transformers.models.bart.tokenization_bart
 from transformers import logging
 from transformers.testing_utils import CaptureLogger, mockenv
 
@@ -31,7 +31,7 @@ class HfArgumentParserTest(unittest.TestCase):
     def test_integration(self):
         level_origin = logging.get_verbosity()
 
-        logger = logging.get_logger("transformers.tokenization_bart")
+        logger = logging.get_logger("transformers.models.bart.tokenization_bart")
         msg = "Testing 1, 2, 3"
 
         # should be able to log warnings (if default settings weren't overridden by `pytest --log-level-all`)
@@ -62,7 +62,7 @@ class HfArgumentParserTest(unittest.TestCase):
         # reset for the env var to take effect, next time some logger call is made
         transformers.utils.logging._reset_library_root_logger()
         # this action activates the env var
-        _ = logging.get_logger("transformers.tokenization_bart")
+        _ = logging.get_logger("transformers.models.bart.tokenization_bart")
 
         env_level_str = os.getenv("TRANSFORMERS_VERBOSITY", None)
         env_level = logging.log_levels[env_level_str]
@@ -85,7 +85,7 @@ class HfArgumentParserTest(unittest.TestCase):
         logger = logging.logging.getLogger()
         with CaptureLogger(logger) as cl:
             # this action activates the env var
-            logging.get_logger("transformers.tokenization_bart")
+            logging.get_logger("transformers.models.bart.tokenization_bart")
         self.assertIn("Unknown option TRANSFORMERS_VERBOSITY=super-error", cl.out)
 
         # no need to restore as nothing was changed

tests/test_modeling_albert.py

@@ -37,7 +37,7 @@ if is_torch_available():
         AlbertForTokenClassification,
         AlbertModel,
     )
-    from transformers.modeling_albert import ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST
+    from transformers.models.albert.modeling_albert import ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
 class AlbertModelTester:

tests/test_modeling_auto.py

@@ -45,7 +45,7 @@ if is_torch_available():
         T5Config,
         T5ForConditionalGeneration,
     )
-    from transformers.modeling_auto import (
+    from transformers.models.auto.modeling_auto import (
         MODEL_FOR_CAUSAL_LM_MAPPING,
         MODEL_FOR_MASKED_LM_MAPPING,
         MODEL_FOR_PRETRAINING_MAPPING,
@@ -56,9 +56,9 @@ if is_torch_available():
         MODEL_MAPPING,
         MODEL_WITH_LM_HEAD_MAPPING,
     )
-    from transformers.modeling_bert import BERT_PRETRAINED_MODEL_ARCHIVE_LIST
-    from transformers.modeling_gpt2 import GPT2_PRETRAINED_MODEL_ARCHIVE_LIST
-    from transformers.modeling_t5 import T5_PRETRAINED_MODEL_ARCHIVE_LIST
+    from transformers.models.bert.modeling_bert import BERT_PRETRAINED_MODEL_ARCHIVE_LIST
+    from transformers.models.gpt2.modeling_gpt2 import GPT2_PRETRAINED_MODEL_ARCHIVE_LIST
+    from transformers.models.t5.modeling_t5 import T5_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
 @require_torch