Reorganize file utils (#16264)

* Split file_utils in several submodules

* Fixes

* Add back more objects

* More fixes

* Who exactly decided to import that from there?

* Second suggestion to code with code review

* Revert wront move

* Fix imports

* Adapt all imports

* Adapt all imports everywhere

* Revert this import, will fix in a separate commit

examples/legacy/seq2seq/old_test_tatoeba_conversion.py

@@ -16,9 +16,9 @@ import os
 import tempfile
 import unittest
 
-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 slow
+from transformers.utils import cached_property
 
 
 @unittest.skipUnless(os.path.exists(DEFAULT_REPO), "Tatoeba directory does not exist.")

examples/legacy/seq2seq/seq2seq_trainer.py

@@ -19,7 +19,6 @@ from torch import nn
 from torch.utils.data import DistributedSampler, RandomSampler
 
 from transformers import PreTrainedModel, Trainer, logging
-from transformers.file_utils import is_torch_tpu_available
 from transformers.integrations import is_fairscale_available
 from transformers.models.fsmt.configuration_fsmt import FSMTConfig
 from transformers.optimization import (
@@ -34,6 +33,7 @@ from transformers.optimization import (
 )
 from transformers.trainer_pt_utils import get_tpu_sampler
 from transformers.training_args import ParallelMode
+from transformers.utils import is_torch_tpu_available
 
 
 if is_fairscale_available():

examples/legacy/seq2seq/utils.py

@@ -34,8 +34,8 @@ 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.models.bart.modeling_bart import shift_tokens_right
+from transformers.utils import cached_property
 
 
 try: