Sort imports with isort.

This is the result of:

    $ isort --recursive examples templates transformers utils hubconf.py setup.py

examples/benchmarks.py

@@ -18,12 +18,14 @@
 # If checking the tensors placement
 # tf.debugging.set_log_device_placement(True)
 
-from typing import List
-import timeit
-from transformers import is_tf_available, is_torch_available
-from time import time
 import argparse
 import csv
+import timeit
+from time import time
+from typing import List
+
+from transformers import AutoConfig, AutoTokenizer, is_tf_available, is_torch_available
+
 
 if is_tf_available():
     import tensorflow as tf
@@ -33,7 +35,6 @@ if is_torch_available():
     import torch
     from transformers import AutoModel
 
-from transformers import AutoConfig, AutoTokenizer
 
 input_text = """Bent over their instruments, three hundred Fertilizers were plunged, as 
 the Director of Hatcheries and Conditioning entered the room, in the 

examples/contrib/run_camembert.py

-from pathlib import Path
 import tarfile
 import urllib.request
+from pathlib import Path
 
 import torch
 
-from transformers.tokenization_camembert import CamembertTokenizer
 from transformers.modeling_camembert import CamembertForMaskedLM
+from transformers.tokenization_camembert import CamembertTokenizer
 
 
 def fill_mask(masked_input, model, tokenizer, topk=5):

examples/contrib/run_openai_gpt.py

@@ -28,26 +28,27 @@
           --train_batch_size 16 \
 """
 import argparse
-import os
 import csv
-import random
 import logging
-from tqdm import tqdm, trange
+import os
+import random
 
 import numpy as np
 import torch
 from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
+from tqdm import tqdm, trange
 
 from transformers import (
+    CONFIG_NAME,
+    WEIGHTS_NAME,
+    AdamW,
     OpenAIGPTDoubleHeadsModel,
     OpenAIGPTTokenizer,
-    AdamW,
     cached_path,
-    WEIGHTS_NAME,
-    CONFIG_NAME,
     get_linear_schedule_with_warmup,
 )
 
+
 ROCSTORIES_URL = "https://s3.amazonaws.com/datasets.huggingface.co/ROCStories.tar.gz"
 
 logging.basicConfig(

examples/contrib/run_swag.py

@@ -19,28 +19,34 @@
 from __future__ import absolute_import, division, print_function
 
 import argparse
-import logging
 import csv
+import glob
+import logging
 import os
 import random
 import sys
-import glob
 
 import numpy as np
 import torch
 from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
 from torch.utils.data.distributed import DistributedSampler
+from tqdm import tqdm, trange
+
+from transformers import (
+    WEIGHTS_NAME,
+    AdamW,
+    BertConfig,
+    BertForMultipleChoice,
+    BertTokenizer,
+    get_linear_schedule_with_warmup,
+)
+
 
 try:
     from torch.utils.tensorboard import SummaryWriter
 except:
     from tensorboardX import SummaryWriter
 
-from tqdm import tqdm, trange
-
-from transformers import WEIGHTS_NAME, BertConfig, BertForMultipleChoice, BertTokenizer
-
-from transformers import AdamW, get_linear_schedule_with_warmup
 
 logger = logging.getLogger(__name__)
 

examples/contrib/run_transfo_xl.py

@@ -23,12 +23,13 @@ from __future__ import absolute_import, division, print_function, unicode_litera
 
 import argparse
 import logging
-import time
 import math
+import time
 
 import torch
 
-from transformers import TransfoXLLMHeadModel, TransfoXLCorpus, TransfoXLTokenizer
+from transformers import TransfoXLCorpus, TransfoXLLMHeadModel, TransfoXLTokenizer
+
 
 logging.basicConfig(
     format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO

examples/distillation/distiller.py

@@ -15,31 +15,31 @@
 """ The distiller to distil the student.
     Adapted in part from Facebook, Inc XLM model (https://github.com/facebookresearch/XLM)
 """
-import os
 import math
-import psutil
+import os
 import time
-from tqdm import trange, tqdm
-import numpy as np
 
+import numpy as np
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
 from torch.optim import AdamW
+from torch.utils.data import BatchSampler, DataLoader, RandomSampler
 from torch.utils.data.distributed import DistributedSampler
-from torch.utils.data import RandomSampler, BatchSampler, DataLoader
+from tqdm import tqdm, trange
+
+import psutil
+from grouped_batch_sampler import GroupedBatchSampler, create_lengths_groups
+from lm_seqs_dataset import LmSeqsDataset
+from transformers import get_linear_schedule_with_warmup
+from utils import logger
+
 
 try:
     from torch.utils.tensorboard import SummaryWriter
 except:
     from tensorboardX import SummaryWriter
 
-from transformers import get_linear_schedule_with_warmup
-
-from utils import logger
-from lm_seqs_dataset import LmSeqsDataset
-from grouped_batch_sampler import GroupedBatchSampler, create_lengths_groups
-
 
 class Distiller:
     def __init__(

examples/distillation/grouped_batch_sampler.py

@@ -17,8 +17,8 @@
 import bisect
 import copy
 from collections import defaultdict
-import numpy as np
 
+import numpy as np
 from torch.utils.data.sampler import BatchSampler, Sampler
 
 from utils import logger

examples/distillation/lm_seqs_dataset.py

@@ -15,10 +15,10 @@
 """ Dataset to distilled models
     adapted in part from Facebook, Inc XLM model (https://github.com/facebookresearch/XLM)
 """
+import numpy as np
 import torch
 from torch.utils.data import Dataset
 
-import numpy as np
 from utils import logger
 
 

examples/distillation/run_squad_w_distillation.py

@@ -18,56 +18,58 @@
 from __future__ import absolute_import, division, print_function
 
 import argparse
+import glob
 import logging
 import os
 import random
-import glob
 
 import numpy as np
 import torch
+import torch.nn as nn
+import torch.nn.functional as F
 from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
 from torch.utils.data.distributed import DistributedSampler
-import torch.nn.functional as F
-import torch.nn as nn
-
-try:
-    from torch.utils.tensorboard import SummaryWriter
-except:
-    from tensorboardX import SummaryWriter
-
 from tqdm import tqdm, trange
 
 from transformers import (
     WEIGHTS_NAME,
+    AdamW,
     BertConfig,
     BertForQuestionAnswering,
     BertTokenizer,
+    DistilBertConfig,
+    DistilBertForQuestionAnswering,
+    DistilBertTokenizer,
     XLMConfig,
     XLMForQuestionAnswering,
     XLMTokenizer,
     XLNetConfig,
     XLNetForQuestionAnswering,
     XLNetTokenizer,
-    DistilBertConfig,
-    DistilBertForQuestionAnswering,
-    DistilBertTokenizer,
+    get_linear_schedule_with_warmup,
 )
 
-from transformers import AdamW, get_linear_schedule_with_warmup
-
 from ..utils_squad import (
-    read_squad_examples,
-    convert_examples_to_features,
     RawResult,
-    write_predictions,
     RawResultExtended,
+    convert_examples_to_features,
+    read_squad_examples,
+    write_predictions,
     write_predictions_extended,
 )
 
 # The follwing import is the official SQuAD evaluation script (2.0).
 # You can remove it from the dependencies if you are using this script outside of the library
 # We've added it here for automated tests (see examples/test_examples.py file)
-from ..utils_squad_evaluate import EVAL_OPTS, main as evaluate_on_squad
+from ..utils_squad_evaluate import EVAL_OPTS
+from ..utils_squad_evaluate import main as evaluate_on_squad
+
+
+try:
+    from torch.utils.tensorboard import SummaryWriter
+except:
+    from tensorboardX import SummaryWriter
+
 
 logger = logging.getLogger(__name__)
 

examples/distillation/scripts/binarized_data.py

@@ -16,12 +16,15 @@
 Preprocessing script before distillation.
 """
 import argparse
+import logging
 import pickle
 import random
 import time
+
 import numpy as np
-from transformers import BertTokenizer, RobertaTokenizer, GPT2Tokenizer
-import logging
+
+from transformers import BertTokenizer, GPT2Tokenizer, RobertaTokenizer
+
 
 logging.basicConfig(
     format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO

examples/distillation/scripts/extract.py

@@ -16,10 +16,13 @@
 Preprocessing script before training the distilled model.
 Specific to RoBERTa -> DistilRoBERTa and GPT2 -> DistilGPT2.
 """
-from transformers import BertForMaskedLM, RobertaForMaskedLM, GPT2LMHeadModel
-import torch
 import argparse
 
+import torch
+
+from transformers import BertForMaskedLM, GPT2LMHeadModel, RobertaForMaskedLM
+
+
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(
         description="Extraction some layers of the full RobertaForMaskedLM or GPT2LMHeadModel for Transfer Learned Distillation"

examples/distillation/scripts/extract_distilbert.py

@@ -16,10 +16,13 @@
 Preprocessing script before training DistilBERT.
 Specific to BERT -> DistilBERT.
 """
-from transformers import BertForMaskedLM, RobertaForMaskedLM
-import torch
 import argparse
 
+import torch
+
+from transformers import BertForMaskedLM, RobertaForMaskedLM
+
+
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(
         description="Extraction some layers of the full BertForMaskedLM or RObertaForMaskedLM for Transfer Learned Distillation"

examples/distillation/scripts/token_counts.py

@@ -15,10 +15,11 @@
 """
 Preprocessing script before training the distilled model.
 """
-from collections import Counter
 import argparse
-import pickle
 import logging
+import pickle
+from collections import Counter
+
 
 logging.basicConfig(
     format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO

examples/distillation/train.py

@@ -16,22 +16,32 @@
 Training the distilled model.
 Supported architectures include: BERT -> DistilBERT, RoBERTa -> DistilRoBERTa, GPT2 -> DistilGPT2.
 """
-import os
 import argparse
-import pickle
 import json
+import os
+import pickle
 import shutil
+
 import numpy as np
 import torch
 
-from transformers import BertConfig, BertForMaskedLM, BertTokenizer
-from transformers import RobertaConfig, RobertaForMaskedLM, RobertaTokenizer
-from transformers import DistilBertConfig, DistilBertForMaskedLM, DistilBertTokenizer
-from transformers import GPT2Config, GPT2LMHeadModel, GPT2Tokenizer
-
 from distiller import Distiller
-from utils import git_log, logger, init_gpu_params, set_seed
 from lm_seqs_dataset import LmSeqsDataset
+from transformers import (
+    BertConfig,
+    BertForMaskedLM,
+    BertTokenizer,
+    DistilBertConfig,
+    DistilBertForMaskedLM,
+    DistilBertTokenizer,
+    GPT2Config,
+    GPT2LMHeadModel,
+    GPT2Tokenizer,
+    RobertaConfig,
+    RobertaForMaskedLM,
+    RobertaTokenizer,
+)
+from utils import git_log, init_gpu_params, logger, set_seed
 
 
 MODEL_CLASSES = {

examples/distillation/utils.py

@@ -15,14 +15,16 @@
 """ Utils to train DistilBERT
     adapted in part from Facebook, Inc XLM model (https://github.com/facebookresearch/XLM)
 """
-import git
 import json
+import logging
 import os
 import socket
-import torch
+
 import numpy as np
+import torch
+
+import git
 
-import logging
 
 logging.basicConfig(
     format="%(asctime)s - %(levelname)s - %(name)s - PID: %(process)d -  %(message)s",

examples/mm-imdb/run_mmimdb.py

@@ -19,32 +19,33 @@ from __future__ import absolute_import, division, print_function
 
 import argparse
 import glob
+import json
 import logging
 import os
 import random
-import json
-from sklearn.metrics import f1_score
 
 import numpy as np
 import torch
 import torch.nn as nn
+from sklearn.metrics import f1_score
 from torch.utils.data import DataLoader, RandomSampler, SequentialSampler
 from torch.utils.data.distributed import DistributedSampler
-
-try:
-    from torch.utils.tensorboard import SummaryWriter
-except:
-    from tensorboardX import SummaryWriter
-
 from tqdm import tqdm, trange
 
-from utils_mmimdb import ImageEncoder, JsonlDataset, collate_fn, get_mmimdb_labels, get_image_transforms
-
 from transformers import (
     WEIGHTS_NAME,
+    AdamW,
+    AlbertConfig,
+    AlbertModel,
+    AlbertTokenizer,
     BertConfig,
     BertModel,
     BertTokenizer,
+    DistilBertConfig,
+    DistilBertModel,
+    DistilBertTokenizer,
+    MMBTConfig,
+    MMBTForClassification,
     RobertaConfig,
     RobertaModel,
     RobertaTokenizer,
@@ -54,17 +55,16 @@ from transformers import (
     XLNetConfig,
     XLNetModel,
     XLNetTokenizer,
-    DistilBertConfig,
-    DistilBertModel,
-    DistilBertTokenizer,
-    AlbertConfig,
-    AlbertModel,
-    AlbertTokenizer,
-    MMBTForClassification,
-    MMBTConfig,
+    get_linear_schedule_with_warmup,
 )
+from utils_mmimdb import ImageEncoder, JsonlDataset, collate_fn, get_image_transforms, get_mmimdb_labels
+
+
+try:
+    from torch.utils.tensorboard import SummaryWriter
+except:
+    from tensorboardX import SummaryWriter
 
-from transformers import AdamW, get_linear_schedule_with_warmup
 
 logger = logging.getLogger(__name__)
 

examples/mm-imdb/utils_mmimdb.py

@@ -17,13 +17,15 @@
 import json
 import os
 from collections import Counter
-from PIL import Image
 
 import torch
 import torch.nn as nn
+from torch.utils.data import Dataset
+
 import torchvision
 import torchvision.transforms as transforms
-from torch.utils.data import Dataset
+from PIL import Image
+
 
 POOLING_BREAKDOWN = {1: (1, 1), 2: (2, 1), 3: (3, 1), 4: (2, 2), 5: (5, 1), 6: (3, 2), 7: (7, 1), 8: (4, 2), 9: (3, 3)}
 

examples/pplm/run_pplm.py

@@ -34,10 +34,11 @@ import torch.nn.functional as F
 from torch.autograd import Variable
 from tqdm import trange
 
+from pplm_classification_head import ClassificationHead
 from transformers import GPT2Tokenizer
 from transformers.file_utils import cached_path
 from transformers.modeling_gpt2 import GPT2LMHeadModel
-from pplm_classification_head import ClassificationHead
+
 
 PPLM_BOW = 1
 PPLM_DISCRIM = 2

examples/pplm/run_pplm_discrim_train.py

@@ -24,16 +24,16 @@ import time
 import numpy as np
 import torch
 import torch.nn.functional as F
-import torch.optim
 import torch.optim as optim
 import torch.utils.data as data
+from tqdm import tqdm, trange
+
 from nltk.tokenize.treebank import TreebankWordDetokenizer
+from pplm_classification_head import ClassificationHead
 from torchtext import data as torchtext_data
 from torchtext import datasets
-from tqdm import tqdm, trange
+from transformers import GPT2LMHeadModel, GPT2Tokenizer
 
-from transformers import GPT2Tokenizer, GPT2LMHeadModel
-from pplm_classification_head import ClassificationHead
 
 torch.manual_seed(0)
 np.random.seed(0)

examples/run_bertology.py

@@ -19,19 +19,19 @@
     Some parts of this script are adapted from the code of Michel et al. (http://arxiv.org/abs/1905.10650)
     which is available at https://github.com/pmichel31415/are-16-heads-really-better-than-1
 """
-import os
 import argparse
 import logging
-from datetime import timedelta, datetime
-from tqdm import tqdm
+import os
+from datetime import datetime, timedelta
 
 import numpy as np
-
 import torch
-from torch.utils.data import DataLoader, SequentialSampler, TensorDataset, Subset
-from torch.utils.data.distributed import DistributedSampler
 from torch.nn import CrossEntropyLoss, MSELoss
+from torch.utils.data import DataLoader, SequentialSampler, Subset, TensorDataset
+from torch.utils.data.distributed import DistributedSampler
+from tqdm import tqdm
 
+from run_glue import ALL_MODELS, MODEL_CLASSES, load_and_cache_examples, set_seed
 from transformers import (
     WEIGHTS_NAME,
     BertConfig,
@@ -44,13 +44,11 @@ from transformers import (
     XLNetForSequenceClassification,
     XLNetTokenizer,
 )
-
-from run_glue import set_seed, load_and_cache_examples, ALL_MODELS, MODEL_CLASSES
-
 from transformers import glue_compute_metrics as compute_metrics
 from transformers import glue_output_modes as output_modes
 from transformers import glue_processors as processors
 
+
 logger = logging.getLogger(__name__)