Add many missing spaces in adjacent strings (#26751)

Add missing spaces in adjacent strings

examples/flax/image-captioning/create_model_from_encoder_decoder_models.py

@@ -37,7 +37,7 @@ class ModelArguments:
     encoder_model_name_or_path: str = field(
         metadata={
             "help": (
-                "The encoder model checkpoint for weights initialization."
+                "The encoder model checkpoint for weights initialization. "
                 "Don't set if you want to train an encoder model from scratch."
             )
         },
@@ -45,7 +45,7 @@ class ModelArguments:
     decoder_model_name_or_path: str = field(
         metadata={
             "help": (
-                "The decoder model checkpoint for weights initialization."
+                "The decoder model checkpoint for weights initialization. "
                 "Don't set if you want to train a decoder model from scratch."
             )
         },

examples/flax/image-captioning/run_image_captioning_flax.py

@@ -203,7 +203,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -256,7 +256,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the `max_length` param of `model.generate`, which is used "
                 "during evaluation."
             )
@@ -423,7 +423,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -685,7 +685,7 @@ def main():
     eval_batch_size = int(training_args.per_device_eval_batch_size) * jax.device_count()
     if training_args.block_size % train_batch_size > 0 or training_args.block_size % eval_batch_size > 0:
         raise ValueError(
-            "`training_args.block_size` needs to be a multiple of the global train/eval batch size."
+            "`training_args.block_size` needs to be a multiple of the global train/eval batch size. "
             f"Got {training_args.block_size}, {train_batch_size} and {eval_batch_size} respectively instead."
         )
 

examples/flax/language-modeling/run_bart_dlm_flax.py

@@ -487,7 +487,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -606,7 +606,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/flax/language-modeling/run_clm_flax.py

@@ -190,7 +190,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -368,7 +368,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -524,7 +524,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -586,7 +586,7 @@ def main():
     else:
         if data_args.block_size > tokenizer.model_max_length:
             logger.warning(
-                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model"
+                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model "
                 f"({tokenizer.model_max_length}). Using block_size={tokenizer.model_max_length}."
             )
         block_size = min(data_args.block_size, tokenizer.model_max_length)

examples/flax/language-modeling/run_mlm_flax.py

@@ -195,7 +195,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -411,7 +411,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -556,7 +556,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/flax/language-modeling/run_t5_mlm_flax.py

@@ -528,7 +528,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -647,7 +647,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/flax/question-answering/run_qa.py

@@ -176,7 +176,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -582,7 +582,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/flax/speech-recognition/run_flax_speech_recognition_seq2seq.py

@@ -415,7 +415,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use `--overwrite_output_dir` to overcome."
         )
 

examples/flax/summarization/run_summarization_flax.py

@@ -209,7 +209,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -268,7 +268,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the `max_length` param of `model.generate`, which is used "
                 "during evaluation."
             )
@@ -451,7 +451,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -558,7 +558,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/flax/text-classification/run_flax_glue.py

@@ -122,7 +122,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/flax/token-classification/run_flax_ner.py

@@ -170,7 +170,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/flax/vision/run_image_classification.py

@@ -180,7 +180,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -291,7 +291,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 

examples/legacy/multiple_choice/utils_multiple_choice.py

@@ -379,7 +379,7 @@ class SwagProcessor(DataProcessor):
         """See base class."""
         logger.info("LOOKING AT {} dev".format(data_dir))
         raise ValueError(
-            "For swag testing, the input file does not contain a label column. It can not be tested in current code"
+            "For swag testing, the input file does not contain a label column. It can not be tested in current code "
             "setting!"
         )
         return self._create_examples(self._read_csv(os.path.join(data_dir, "test.csv")), "test")
@@ -541,7 +541,7 @@ def convert_examples_to_features(
             if "num_truncated_tokens" in inputs and inputs["num_truncated_tokens"] > 0:
                 logger.info(
                     "Attention! you are cropping tokens (swag task is ok). "
-                    "If you are training ARC and RACE and you are poping question + options,"
+                    "If you are training ARC and RACE and you are poping question + options, "
                     "you need to try to use a bigger max seq length!"
                 )
 

examples/legacy/pytorch-lightning/lightning_base.py

@@ -313,7 +313,7 @@ def add_generic_args(parser, root_dir) -> None:
         type=str,
         default="O2",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/legacy/question-answering/run_squad.py

@@ -663,7 +663,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/legacy/run_language_modeling.py

@@ -149,7 +149,7 @@ class DataTrainingArguments:
         default=-1,
         metadata={
             "help": (
-                "Optional input sequence length after tokenization."
+                "Optional input sequence length after tokenization. "
                 "The training dataset will be truncated in block of this size for training."
                 "Default to the model max input length for single sentence inputs (take into account special tokens)."
             )
@@ -283,7 +283,7 @@ def main():
 
     if config.model_type in ["bert", "roberta", "distilbert", "camembert"] and not data_args.mlm:
         raise ValueError(
-            "BERT and RoBERTa-like models do not have LM heads but masked LM heads. They must be run using the"
+            "BERT and RoBERTa-like models do not have LM heads but masked LM heads. They must be run using the "
             "--mlm flag (masked language modeling)."
         )
 

examples/legacy/run_swag.py

@@ -579,7 +579,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/pytorch/audio-classification/run_audio_classification.py

@@ -172,7 +172,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -189,14 +189,14 @@ class ModelArguments:
         if not self.freeze_feature_extractor and self.freeze_feature_encoder:
             warnings.warn(
                 "The argument `--freeze_feature_extractor` is deprecated and "
-                "will be removed in a future version. Use `--freeze_feature_encoder`"
+                "will be removed in a future version. Use `--freeze_feature_encoder` "
                 "instead. Setting `freeze_feature_encoder==True`.",
                 FutureWarning,
             )
         if self.freeze_feature_extractor and not self.freeze_feature_encoder:
             raise ValueError(
                 "The argument `--freeze_feature_extractor` is deprecated and "
-                "should not be used in combination with `--freeze_feature_encoder`."
+                "should not be used in combination with `--freeze_feature_encoder`. "
                 "Only make use of `--freeze_feature_encoder`."
             )
 

examples/pytorch/contrastive-image-text/run_clip.py

@@ -107,7 +107,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -358,7 +358,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/pytorch/image-classification/run_image_classification.py

@@ -163,7 +163,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -152,7 +152,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -179,7 +179,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )

examples/pytorch/image-pretraining/run_mim.py

@@ -174,7 +174,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/pytorch/image-pretraining/run_mim_no_trainer.py

@@ -208,7 +208,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -246,7 +246,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )

examples/pytorch/language-modeling/run_clm.py

@@ -132,7 +132,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -151,7 +151,7 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded."
+                "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded. "
                 "set True will benefit LLM loading time and RAM consumption."
             )
         },
@@ -424,7 +424,7 @@ def main():
         tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path, **tokenizer_kwargs)
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -506,7 +506,7 @@ def main():
     else:
         if data_args.block_size > tokenizer.model_max_length:
             logger.warning(
-                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model"
+                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model "
                 f"({tokenizer.model_max_length}). Using block_size={tokenizer.model_max_length}."
             )
         block_size = min(data_args.block_size, tokenizer.model_max_length)

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -199,7 +199,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -226,7 +226,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )
@@ -234,7 +234,7 @@ def parse_args():
         "--low_cpu_mem_usage",
         action="store_true",
         help=(
-            "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded."
+            "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded. "
             "If passed, LLM loading time and RAM consumption will be benefited."
         ),
     )
@@ -398,7 +398,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -449,7 +449,7 @@ def main():
     else:
         if args.block_size > tokenizer.model_max_length:
             logger.warning(
-                f"The block_size passed ({args.block_size}) is larger than the maximum length for the model"
+                f"The block_size passed ({args.block_size}) is larger than the maximum length for the model "
                 f"({tokenizer.model_max_length}). Using block_size={tokenizer.model_max_length}."
             )
         block_size = min(args.block_size, tokenizer.model_max_length)

examples/pytorch/language-modeling/run_mlm.py

@@ -128,7 +128,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -137,7 +137,7 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded."
+                "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded. "
                 "set True will benefit LLM loading time and RAM consumption."
             )
         },
@@ -417,7 +417,7 @@ def main():
         tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path, **tokenizer_kwargs)
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -462,7 +462,7 @@ def main():
     else:
         if data_args.max_seq_length > tokenizer.model_max_length:
             logger.warning(
-                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
                 f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
             )
         max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/language-modeling/run_mlm_no_trainer.py

@@ -206,7 +206,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -233,7 +233,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )
@@ -241,7 +241,7 @@ def parse_args():
         "--low_cpu_mem_usage",
         action="store_true",
         help=(
-            "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded."
+            "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded. "
             "If passed, LLM loading time and RAM consumption will be benefited."
         ),
     )
@@ -395,7 +395,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -434,7 +434,7 @@ def main():
     else:
         if args.max_seq_length > tokenizer.model_max_length:
             logger.warning(
-                f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the"
+                f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the "
                 f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
             )
         max_seq_length = min(args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/language-modeling/run_plm.py

@@ -115,7 +115,7 @@ class ModelArguments:
         default=False,
         metadata={
             "help": (
-                "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded."
+                "It is an option to create the model as an empty shell, then only materialize its parameters when the pretrained weights are loaded. "
                 "set True will benefit LLM loading time and RAM consumption."
             )
         },
@@ -385,7 +385,7 @@ def main():
         tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path, **tokenizer_kwargs)
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -419,7 +419,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/multiple-choice/run_swag.py

@@ -100,7 +100,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -375,7 +375,7 @@ def main():
     else:
         if data_args.max_seq_length > tokenizer.model_max_length:
             logger.warning(
-                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
                 f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
             )
         max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/multiple-choice/run_swag_no_trainer.py

@@ -188,7 +188,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -215,7 +215,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )
@@ -401,7 +401,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/pytorch/question-answering/run_qa.py

@@ -100,7 +100,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -391,7 +391,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/question-answering/run_qa_beam_search.py

@@ -367,7 +367,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/question-answering/run_qa_beam_search_no_trainer.py

@@ -395,7 +395,7 @@ def main():
 
     if args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
 

examples/pytorch/question-answering/run_qa_no_trainer.py

@@ -279,7 +279,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -306,7 +306,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )
@@ -442,7 +442,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -471,7 +471,7 @@ def main():
 
     if args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
 

examples/pytorch/question-answering/run_seq2seq_qa.py

@@ -101,7 +101,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -171,7 +171,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_answer_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_answer_length`. "
                 "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
                 "during ``evaluate`` and ``predict``."
             )
@@ -465,13 +465,13 @@ def main():
 
     if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
         logger.warning(
-            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for"
+            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for "
             f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
         )
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/semantic-segmentation/run_semantic_segmentation.py

@@ -262,7 +262,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py

@@ -279,7 +279,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -307,7 +307,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )

examples/pytorch/speech-recognition/run_speech_recognition_ctc.py

@@ -104,8 +104,8 @@ class ModelArguments:
         default=0.05,
         metadata={
             "help": (
-                "Probability of each feature vector along the time axis to be chosen as the start of the vector"
-                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
+                "Probability of each feature vector along the time axis to be chosen as the start of the vector "
+                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
                 "vectors will be masked along the time axis."
             )
         },
@@ -249,7 +249,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -430,7 +430,7 @@ def main():
 
     # Log on each process the small summary:
     logger.warning(
-        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu} "
         f"distributed training: {training_args.parallel_mode.value == 'distributed'}, 16-bits training: {training_args.fp16}"
     )
     # Set the verbosity to info of the Transformers logger (on main process only):

examples/pytorch/speech-recognition/run_speech_recognition_ctc_adapter.py

@@ -90,8 +90,8 @@ class ModelArguments:
         default=0.05,
         metadata={
             "help": (
-                "Probability of each feature vector along the time axis to be chosen as the start of the vector"
-                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
+                "Probability of each feature vector along the time axis to be chosen as the start of the vector "
+                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
                 "vectors will be masked along the time axis."
             )
         },
@@ -252,7 +252,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -426,7 +426,7 @@ def main():
 
     # Log on each process the small summary:
     logger.warning(
-        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu} "
         f"distributed training: {training_args.parallel_mode.value == 'distributed'}, 16-bits training: {training_args.fp16}"
     )
     # Set the verbosity to info of the Transformers logger (on main process only):

examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py

@@ -106,7 +106,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -322,7 +322,7 @@ def main():
 
     # Log on each process the small summary:
     logger.warning(
-        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu} "
         f"distributed training: {training_args.parallel_mode.value == 'distributed'}, 16-bits training: {training_args.fp16}"
     )
     logger.info(f"Training/evaluation parameters {training_args}")

examples/pytorch/summarization/run_summarization.py

@@ -120,7 +120,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -205,7 +205,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
                 "during ``evaluate`` and ``predict``."
             )
@@ -271,7 +271,7 @@ class DataTrainingArguments:
         default=None,
         metadata={
             "help": (
-                "The token to force as the first generated token after the decoder_start_token_id."
+                "The token to force as the first generated token after the decoder_start_token_id. "
                 "Useful for multilingual models like mBART where the first generated token"
                 "needs to be the target language token (Usually it is the target language token)"
             )
@@ -556,7 +556,7 @@ def main():
 
     if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
         logger.warning(
-            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for"
+            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for "
             f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
         )
 

examples/pytorch/summarization/run_summarization_no_trainer.py

@@ -146,7 +146,7 @@ def parse_args():
         default=128,
         help=(
             "The maximum total sequence length for target text after "
-            "tokenization. Sequences longer than this will be truncated, sequences shorter will be padded."
+            "tokenization. Sequences longer than this will be truncated, sequences shorter will be padded. "
             "during ``evaluate`` and ``predict``."
         ),
     )
@@ -272,7 +272,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -299,7 +299,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )
@@ -433,7 +433,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/pytorch/text-classification/run_classification.py

@@ -82,7 +82,7 @@ class DataTrainingArguments:
         default=None,
         metadata={
             "help": (
-                "The name of the text column in the input dataset or a CSV/JSON file."
+                "The name of the text column in the input dataset or a CSV/JSON file. "
                 'If not specified, will use the "sentence" column for single/multi-label classifcation task.'
             )
         },
@@ -120,7 +120,7 @@ class DataTrainingArguments:
         default=None,
         metadata={
             "help": (
-                "The name of the label column in the input dataset or a CSV/JSON file."
+                "The name of the label column in the input dataset or a CSV/JSON file. "
                 'If not specified, will use the "label" column for single/multi-label classifcation task'
             )
         },
@@ -248,7 +248,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -552,7 +552,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/text-classification/run_glue.py

@@ -209,7 +209,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -456,7 +456,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/pytorch/text-classification/run_glue_no_trainer.py

@@ -162,7 +162,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -189,7 +189,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )

examples/pytorch/text-classification/run_xnli.py

@@ -173,7 +173,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/pytorch/token-classification/run_ner.py

@@ -100,7 +100,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/pytorch/token-classification/run_ner_no_trainer.py

@@ -216,7 +216,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -243,7 +243,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )
@@ -412,7 +412,7 @@ def main():
     tokenizer_name_or_path = args.tokenizer_name if args.tokenizer_name else args.model_name_or_path
     if not tokenizer_name_or_path:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/pytorch/translation/run_translation.py

@@ -110,7 +110,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -173,7 +173,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
                 "during ``evaluate`` and ``predict``."
             )
@@ -468,7 +468,7 @@ def main():
 
     if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
         logger.warning(
-            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for"
+            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for "
             f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
         )
 

examples/pytorch/translation/run_translation_no_trainer.py

@@ -118,7 +118,7 @@ def parse_args():
         default=128,
         help=(
             "The maximum total sequence length for target text after "
-            "tokenization. Sequences longer than this will be truncated, sequences shorter will be padded."
+            "tokenization. Sequences longer than this will be truncated, sequences shorter will be padded "
             "during ``evaluate`` and ``predict``."
         ),
     )
@@ -139,7 +139,7 @@ def parse_args():
         default=False,
         help=(
             "Whether to pad all samples to model maximum sentence "
-            "length. If False, will pad the samples dynamically when batching to the maximum length in the batch. More"
+            "length. If False, will pad the samples dynamically when batching to the maximum length in the batch. More "
             "efficient on GPU but very bad for TPU."
         ),
     )
@@ -263,7 +263,7 @@ def parse_args():
         default=False,
         help=(
             "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-            "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+            "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
             "execute code present on the Hub on your local machine."
         ),
     )
@@ -290,7 +290,7 @@ def parse_args():
         default="all",
         help=(
             'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
-            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations.'
+            ' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
             "Only applicable when `--with_tracking` is passed."
         ),
     )
@@ -413,7 +413,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/research_projects/bert-loses-patience/run_glue_with_pabee.py

@@ -575,7 +575,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/deebert/run_glue_deebert.py

@@ -532,7 +532,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/distillation/run_squad_w_distillation.py

@@ -696,7 +696,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/distillation/train.py

@@ -208,7 +208,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/information-gain-filtration/run_clm_igf.py

@@ -354,7 +354,7 @@ def main():
         default=10,
         type=int,
         help=(
-            "decay the selectivity of our secondary learner filter from"
+            "decay the selectivity of our secondary learner filter from "
             "1 standard deviation above average to 1 below average after 10 batches"
         ),
     )

examples/research_projects/jax-projects/dataset-streaming/run_mlm_flax_stream.py

@@ -341,7 +341,7 @@ if __name__ == "__main__":
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -399,7 +399,7 @@ if __name__ == "__main__":
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/research_projects/jax-projects/hybrid_clip/run_hybrid_clip.py

@@ -78,7 +78,7 @@ class ModelArguments:
     text_model_name_or_path: str = field(
         metadata={
             "help": (
-                "The text model checkpoint for weights initialization."
+                "The text model checkpoint for weights initialization. "
                 "Don't set if you want to train a model from scratch."
             )
         },
@@ -86,7 +86,7 @@ class ModelArguments:
     vision_model_name_or_path: str = field(
         metadata={
             "help": (
-                "The vision model checkpoint for weights initialization."
+                "The vision model checkpoint for weights initialization. "
                 "Don't set if you want to train a model from scratch."
             )
         },
@@ -311,7 +311,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -341,7 +341,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/research_projects/jax-projects/model_parallel/run_clm_mp.py

@@ -246,7 +246,7 @@ def main():
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -325,7 +325,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 
@@ -368,7 +368,7 @@ def main():
     else:
         if data_args.block_size > tokenizer.model_max_length:
             logger.warning(
-                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model"
+                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model "
                 f"({tokenizer.model_max_length}). Using block_size={tokenizer.model_max_length}."
             )
         block_size = min(data_args.block_size, tokenizer.model_max_length)

examples/research_projects/luke/run_luke_ner_no_trainer.py

@@ -357,7 +357,7 @@ def main():
     tokenizer_name_or_path = args.tokenizer_name if args.tokenizer_name else args.model_name_or_path
     if not tokenizer_name_or_path:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/research_projects/mlm_wwm/run_mlm_wwm.py

@@ -314,7 +314,7 @@ def main():
         tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path, **tokenizer_kwargs)
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/research_projects/mm-imdb/run_mmimdb.py

@@ -426,7 +426,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/movement-pruning/bertarize.py

@@ -112,8 +112,8 @@ if __name__ == "__main__":
         type=float,
         required=False,
         help=(
-            "For `magnitude` and `topK`, it is the level of remaining weights (in %) in the fine-pruned model."
-            "For `sigmoied_threshold`, it is the threshold \tau against which the (sigmoied) scores are compared."
+            "For `magnitude` and `topK`, it is the level of remaining weights (in %) in the fine-pruned model. "
+            "For `sigmoied_threshold`, it is the threshold \tau against which the (sigmoied) scores are compared. "
             "Not needed for `l0`"
         ),
     )

examples/research_projects/movement-pruning/counts_parameters.py

@@ -79,8 +79,8 @@ if __name__ == "__main__":
         type=float,
         required=False,
         help=(
-            "For `topK`, it is the level of remaining weights (in %) in the fine-pruned model."
-            "For `sigmoied_threshold`, it is the threshold \tau against which the (sigmoied) scores are compared."
+            "For `topK`, it is the level of remaining weights (in %) in the fine-pruned model. "
+            "For `sigmoied_threshold`, it is the threshold \tau against which the (sigmoied) scores are compared. "
             "Not needed for `l0`"
         ),
     )

examples/research_projects/movement-pruning/masked_run_glue.py

@@ -671,7 +671,7 @@ def main():
         default=1,
         type=int,
         help=(
-            "Run `initial_warmup` * `warmup_steps` steps of threshold warmup during which threshold stays"
+            "Run `initial_warmup` * `warmup_steps` steps of threshold warmup during which threshold stays "
             "at its `initial_threshold` value (sparsity schedule)."
         ),
     )
@@ -680,7 +680,7 @@ def main():
         default=2,
         type=int,
         help=(
-            "Run `final_warmup` * `warmup_steps` steps of threshold cool-down during which threshold stays"
+            "Run `final_warmup` * `warmup_steps` steps of threshold cool-down during which threshold stays "
             "at its final_threshold value (sparsity schedule)."
         ),
     )
@@ -799,7 +799,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/movement-pruning/masked_run_squad.py

@@ -789,7 +789,7 @@ def main():
         default=1,
         type=int,
         help=(
-            "Run `initial_warmup` * `warmup_steps` steps of threshold warmup during which threshold stays"
+            "Run `initial_warmup` * `warmup_steps` steps of threshold warmup during which threshold stays "
             "at its `initial_threshold` value (sparsity schedule)."
         ),
     )
@@ -798,7 +798,7 @@ def main():
         default=2,
         type=int,
         help=(
-            "Run `final_warmup` * `warmup_steps` steps of threshold cool-down during which threshold stays"
+            "Run `final_warmup` * `warmup_steps` steps of threshold cool-down during which threshold stays "
             "at its final_threshold value (sparsity schedule)."
         ),
     )
@@ -946,7 +946,7 @@ def main():
         type=str,
         default="O1",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/performer/run_mlm_performer.py

@@ -466,7 +466,7 @@ if __name__ == "__main__":
         and not training_args.overwrite_output_dir
     ):
         raise ValueError(
-            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            f"Output directory ({training_args.output_dir}) already exists and is not empty. "
             "Use --overwrite_output_dir to overcome."
         )
 
@@ -558,7 +558,7 @@ if __name__ == "__main__":
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/research_projects/pplm/run_pplm_discrim_train.py

@@ -490,8 +490,8 @@ if __name__ == "__main__":
         default="SST",
         choices=("SST", "clickbait", "toxic", "generic"),
         help=(
-            "dataset to train the discriminator on."
-            "In case of generic, the dataset is expected"
+            "dataset to train the discriminator on. "
+            "In case of generic, the dataset is expected "
             "to be a TSBV file with structure: class \\t text"
         ),
     )

examples/research_projects/quantization-qdqbert/evaluate-hf-trt-qa.py

@@ -153,7 +153,7 @@ if args.tokenizer_name:
     tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name, use_fast=True)
 else:
     raise ValueError(
-        "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+        "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
         "You can do it from another script, save it, and load it from here, using --tokenizer_name."
     )
 
@@ -288,7 +288,7 @@ pad_on_right = tokenizer.padding_side == "right"
 
 if args.max_seq_length > tokenizer.model_max_length:
     logger.warning(
-        f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the"
+        f"The max_seq_length passed ({args.max_seq_length}) is larger than the maximum length for the "
         f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
     )
 

examples/research_projects/quantization-qdqbert/run_quant_qa.py

@@ -365,7 +365,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/research_projects/rag-end2end-retriever/finetune_rag.py

@@ -680,7 +680,7 @@ class GenerativeQAModule(BaseTransformer):
             type=int,
             default=1,
             help=(
-                "The number of retrieval actors to use when Ray is selected"
+                "The number of retrieval actors to use when Ray is selected "
                 "for the distributed retriever. Has no effect when "
                 "distributed_retriever is set to pytorch."
             ),
@@ -719,7 +719,7 @@ def main(args=None, model=None) -> GenerativeQAModule:
             ray.init(address=args.ray_address, namespace="rag")
         except (ConnectionError, ValueError):
             logger.warning(
-                "Connection to Ray cluster failed. Make sure a Ray"
+                "Connection to Ray cluster failed. Make sure a Ray "
                 "cluster is running by either using Ray's cluster "
                 "launcher (`ray up`) or by manually starting Ray on "
                 "each node via `ray start --head` for the head node "

examples/research_projects/rag-end2end-retriever/lightning_base.py

@@ -333,7 +333,7 @@ def add_generic_args(parser, root_dir) -> None:
         type=str,
         default="O2",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/rag/finetune_rag.py

@@ -525,7 +525,7 @@ class GenerativeQAModule(BaseTransformer):
             type=int,
             default=1,
             help=(
-                "The number of retrieval actors to use when Ray is selected"
+                "The number of retrieval actors to use when Ray is selected "
                 "for the distributed retriever. Has no effect when "
                 "distributed_retriever is set to pytorch."
             ),
@@ -552,7 +552,7 @@ def main(args=None, model=None) -> GenerativeQAModule:
             ray.init(address=args.ray_address, namespace="rag")
         except (ConnectionError, ValueError):
             logger.warning(
-                "Connection to Ray cluster failed. Make sure a Ray"
+                "Connection to Ray cluster failed. Make sure a Ray "
                 "cluster is running by either using Ray's cluster "
                 "launcher (`ray up`) or by manually starting Ray on "
                 "each node via `ray start --head` for the head node "

examples/research_projects/rag/lightning_base.py

@@ -322,7 +322,7 @@ def add_generic_args(parser, root_dir) -> None:
         type=str,
         default="O2",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/robust-speech-event/run_speech_recognition_ctc_bnb.py

@@ -104,8 +104,8 @@ class ModelArguments:
         default=0.05,
         metadata={
             "help": (
-                "Probability of each feature vector along the time axis to be chosen as the start of the vector"
-                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
+                "Probability of each feature vector along the time axis to be chosen as the start of the vector "
+                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
                 "vectors will be masked along the time axis."
             )
         },
@@ -399,7 +399,7 @@ def main():
 
     # Log on each process the small summary:
     logger.warning(
-        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}, "
         f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
     )
     # Set the verbosity to info of the Transformers logger (on main process only):

examples/research_projects/robust-speech-event/run_speech_recognition_ctc_streaming.py

@@ -103,8 +103,8 @@ class ModelArguments:
         default=0.05,
         metadata={
             "help": (
-                "Probability of each feature vector along the time axis to be chosen as the start of the vector"
-                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
+                "Probability of each feature vector along the time axis to be chosen as the start of the vector "
+                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
                 "vectors will be masked along the time axis."
             )
         },
@@ -354,7 +354,7 @@ def main():
 
     # Log on each process the small summary:
     logger.warning(
-        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}, "
         f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
     )
     # Set the verbosity to info of the Transformers logger (on main process only):

examples/research_projects/seq2seq-distillation/lightning_base.py

@@ -313,7 +313,7 @@ def add_generic_args(parser, root_dir) -> None:
         type=str,
         default="O2",
         help=(
-            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
+            "For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
             "See details at https://nvidia.github.io/apex/amp.html"
         ),
     )

examples/research_projects/tapex/run_tabfact_with_tapex.py

@@ -325,7 +325,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/research_projects/tapex/run_wikisql_with_tapex.py

@@ -170,7 +170,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
                 "during ``evaluate`` and ``predict``."
             )
@@ -379,7 +379,7 @@ def main():
 
     if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
         logger.warning(
-            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for"
+            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for "
             f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
         )
 

examples/research_projects/tapex/run_wikitablequestions_with_tapex.py

@@ -168,7 +168,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
                 "during ``evaluate`` and ``predict``."
             )
@@ -377,7 +377,7 @@ def main():
 
     if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
         logger.warning(
-            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for"
+            "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for "
             f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
         )
 

examples/research_projects/wav2vec2/run_common_voice.py

@@ -80,8 +80,8 @@ class ModelArguments:
         default=0.05,
         metadata={
             "help": (
-                "Propability of each feature vector along the time axis to be chosen as the start of the vector"
-                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
+                "Propability of each feature vector along the time axis to be chosen as the start of the vector "
+                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
                 "vectors will be masked along the time axis. This is only relevant if ``apply_spec_augment is True``."
             )
         },

examples/research_projects/xtreme-s/run_xtreme_s.py

@@ -116,8 +116,8 @@ class ModelArguments:
         default=0.05,
         metadata={
             "help": (
-                "Probability of each feature vector along the time axis to be chosen as the start of the vector"
-                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
+                "Probability of each feature vector along the time axis to be chosen as the start of the vector "
+                "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
                 "vectors will be masked along the time axis."
             )
         },
@@ -455,7 +455,7 @@ def main():
 
     # Log on each process the small summary:
     logger.warning(
-        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}, "
         f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
     )
     # Set the verbosity to info of the Transformers logger (on main process only):

examples/research_projects/zero-shot-distillation/distill_classifier.py

@@ -41,7 +41,7 @@ class TeacherModelArguments:
         default="This example is {}.",
         metadata={
             "help": (
-                "Template used to turn class names into mock hypotheses for teacher NLI model. Must include {{}}"
+                "Template used to turn class names into mock hypotheses for teacher NLI model. Must include {{}} "
                 "where class name is inserted."
             )
         },
@@ -53,7 +53,7 @@ class TeacherModelArguments:
         default=False,
         metadata={
             "help": (
-                "Allow multiple classes to be true rather than forcing them to sum to 1 (sometimes called"
+                "Allow multiple classes to be true rather than forcing them to sum to 1 (sometimes called "
                 "multi-class multi-label classification)."
             )
         },
@@ -98,7 +98,7 @@ class DistillTrainingArguments(TrainingArguments):
         default=True,
         metadata={
             "help": (
-                "Whether to evaluate the agreement of the final student predictions and the teacher predictions"
+                "Whether to evaluate the agreement of the final student predictions and the teacher predictions "
                 "after training."
             )
         },
@@ -107,7 +107,7 @@ class DistillTrainingArguments(TrainingArguments):
         default=0,
         metadata={
             "help": (
-                "Limit the total amount of checkpoints."
+                "Limit the total amount of checkpoints. "
                 "Deletes the older checkpoints in the output_dir. Default is 0 (no checkpoints)."
             )
         },

examples/tensorflow/contrastive-image-text/run_clip.py

@@ -113,7 +113,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -387,7 +387,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/tensorflow/image-classification/run_image_classification.py

@@ -178,7 +178,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },

examples/tensorflow/language-modeling/run_clm.py

@@ -133,7 +133,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -399,7 +399,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
     # endregion
@@ -432,7 +432,7 @@ def main():
     else:
         if data_args.block_size > tokenizer.model_max_length:
             logger.warning(
-                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model"
+                f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model "
                 f"({tokenizer.model_max_length}). Using block_size={tokenizer.model_max_length}."
             )
         block_size = min(data_args.block_size, tokenizer.model_max_length)

examples/tensorflow/language-modeling/run_mlm.py

@@ -131,7 +131,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -383,7 +383,7 @@ def main():
         )
     else:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
     # endregion
@@ -404,7 +404,7 @@ def main():
     else:
         if data_args.max_seq_length > tokenizer.model_max_length:
             logger.warning(
-                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
                 f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
             )
         max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/tensorflow/multiple-choice/run_swag.py

@@ -167,7 +167,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -383,7 +383,7 @@ def main():
     else:
         if data_args.max_seq_length > tokenizer.model_max_length:
             logger.warning(
-                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+                f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
                 f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
             )
         max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/tensorflow/question-answering/run_qa.py

@@ -98,7 +98,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -400,7 +400,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/tensorflow/summarization/run_summarization.py

@@ -120,7 +120,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -194,7 +194,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
                 "during ``evaluate`` and ``predict``."
             )

examples/tensorflow/text-classification/run_glue.py

@@ -185,7 +185,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -353,7 +353,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/tensorflow/text-classification/run_text_classification.py

@@ -101,7 +101,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "Whether to pad all samples to `max_seq_length`. "
-                "If False, will pad the samples dynamically when batching to the maximum length in the batch."
+                "If False, will pad the samples dynamically when batching to the maximum length in the batch. "
                 "Data will always be padded when using TPUs."
             )
         },
@@ -191,7 +191,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -360,7 +360,7 @@ def main():
 
     if data_args.max_seq_length > tokenizer.model_max_length:
         logger.warning(
-            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
+            f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the "
             f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
         )
     max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)

examples/tensorflow/token-classification/run_ner.py

@@ -96,7 +96,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -334,7 +334,7 @@ def main():
     tokenizer_name_or_path = model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path
     if not tokenizer_name_or_path:
         raise ValueError(
-            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script. "
             "You can do it from another script, save it, and load it from here, using --tokenizer_name."
         )
 

examples/tensorflow/translation/run_translation.py

@@ -114,7 +114,7 @@ class ModelArguments:
         metadata={
             "help": (
                 "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
-                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will "
                 "execute code present on the Hub on your local machine."
             )
         },
@@ -182,7 +182,7 @@ class DataTrainingArguments:
         metadata={
             "help": (
                 "The maximum total sequence length for validation target text after tokenization. Sequences longer "
-                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
+                "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
                 "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
                 "during ``evaluate`` and ``predict``."
             )

src/transformers/commands/serving.py

@@ -131,7 +131,7 @@ class ServeCommand(BaseTransformersCLICommand):
         if not _serve_dependencies_installed:
             raise RuntimeError(
                 "Using serve command requires FastAPI and uvicorn. "
-                'Please install transformers with [serving]: pip install "transformers[serving]".'
+                'Please install transformers with [serving]: pip install "transformers[serving]". '
                 "Or install FastAPI and uvicorn separately."
             )
         else:

src/transformers/dynamic_module_utils.py

@@ -575,7 +575,7 @@ def custom_object_save(obj: Any, folder: Union[str, os.PathLike], config: Option
 
 def _raise_timeout_error(signum, frame):
     raise ValueError(
-        "Loading this model requires you to execute custom code contained in the model repository on your local"
+        "Loading this model requires you to execute custom code contained in the model repository on your local "
         "machine. Please set the option `trust_remote_code=True` to permit loading of this model."
     )
 
@@ -593,7 +593,7 @@ def resolve_trust_remote_code(trust_remote_code, model_name, has_local_code, has
                 signal.alarm(TIME_OUT_REMOTE_CODE)
                 while trust_remote_code is None:
                     answer = input(
-                        f"The repository for {model_name} contains custom code which must be executed to correctly"
+                        f"The repository for {model_name} contains custom code which must be executed to correctly "
                         f"load the model. You can inspect the repository content at https://hf.co/{model_name}.\n"
                         f"You can avoid this prompt in future by passing the argument `trust_remote_code=True`.\n\n"
                         f"Do you wish to run the custom code? [y/N] "
@@ -606,7 +606,7 @@ def resolve_trust_remote_code(trust_remote_code, model_name, has_local_code, has
             except Exception:
                 # OS which does not support signal.SIGALRM
                 raise ValueError(
-                    f"The repository for {model_name} contains custom code which must be executed to correctly"
+                    f"The repository for {model_name} contains custom code which must be executed to correctly "
                     f"load the model. You can inspect the repository content at https://hf.co/{model_name}.\n"
                     f"Please pass the argument `trust_remote_code=True` to allow custom code to be run."
                 )

src/transformers/generation/tf_utils.py

@@ -1198,7 +1198,7 @@ class TFGenerationMixin:
         inputs_kwarg = model_kwargs.pop(input_name, None)
         if inputs_kwarg is not None and inputs is not None:
             raise ValueError(
-                f"`inputs`: {inputs}` were passed alongside {input_name} which is not allowed."
+                f"`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. "
                 f"Make sure to either pass {inputs} or {input_name}=..."
             )
         elif inputs_kwarg is not None:

src/transformers/generation/utils.py

@@ -544,7 +544,7 @@ class GenerationMixin:
         inputs_kwarg = model_kwargs.pop(input_name, None)
         if inputs_kwarg is not None and inputs is not None:
             raise ValueError(
-                f"`inputs`: {inputs}` were passed alongside {input_name} which is not allowed."
+                f"`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. "
                 f"Make sure to either pass {inputs} or {input_name}=..."
             )
         elif inputs_kwarg is not None:
@@ -1816,7 +1816,7 @@ class GenerationMixin:
 
                 def typeerror():
                     raise ValueError(
-                        "`force_words_ids` has to either be a `List[List[List[int]]]` or `List[List[int]]`"
+                        "`force_words_ids` has to either be a `List[List[List[int]]]` or `List[List[int]]` "
                         f"of positive integers, but is {generation_config.force_words_ids}."
                     )
 

src/transformers/integrations/integration_utils.py

@@ -1312,7 +1312,7 @@ class NeptuneCallback(TrainerCallback):
                 target_path = consistent_checkpoint_path
             except IOError as e:
                 logger.warning(
-                    "NeptuneCallback was unable to made a copy of checkpoint due to I/O exception: '{}'."
+                    "NeptuneCallback was unable to made a copy of checkpoint due to I/O exception: '{}'. "
                     "Could fail trying to upload.".format(e)
                 )
 

src/transformers/modeling_tf_utils.py

@@ -850,7 +850,7 @@ def load_tf_shard(model, model_layer_map, resolved_archive_file, ignore_mismatch
             raise OSError(
                 f"Unable to load weights from TF checkpoint file for '{resolved_archive_file}' "
                 f"at '{resolved_archive_file}'. "
-                "If you tried to load a TF model from a sharded checkpoint, you should try converting the model"
+                "If you tried to load a TF model from a sharded checkpoint, you should try converting the model "
                 "by loading it in pytorch and saving it localy. A convertion script should be realeased soon."
             )
 

src/transformers/modeling_utils.py

@@ -2635,8 +2635,8 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                 else:
                     raise RuntimeError("No GPU found. A GPU is needed for quantization.")
                 logger.info(
-                    "The device_map was not initialized."
-                    "Setting device_map to {'':torch.cuda.current_device()}."
+                    "The device_map was not initialized. "
+                    "Setting device_map to {'':torch.cuda.current_device()}. "
                     "If you want to use the model for inference, please set device_map ='auto' "
                 )
                 if low_cpu_mem_usage is None:
@@ -2762,8 +2762,8 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                     else:
                         raise RuntimeError("No GPU found. A GPU is needed for quantization.")
                     logger.info(
-                        "The device_map was not initialized."
-                        "Setting device_map to {'':torch.cuda.current_device()}."
+                        "The device_map was not initialized. "
+                        "Setting device_map to {'':torch.cuda.current_device()}. "
                         "If you want to use the model for inference, please set device_map ='auto' "
                     )
                     if low_cpu_mem_usage is None:
@@ -3149,7 +3149,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
 
         if load_in_8bit and torch_dtype is None:
             logger.warning(
-                "You are loading your model in 8bit but you did not specify a `torch_dtype` attribute."
+                "You are loading your model in 8bit but you did not specify a `torch_dtype` attribute. "
                 "All non-linear modules will be loaded in full precision."
                 " If you want to load the other modules in other precision, please specify a `torch_dtype` attribute."
             )
@@ -3193,8 +3193,8 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                 else:
                     raise ValueError(
                         "You are using `device_map='auto'` on a 4bit loaded version of the model. To automatically compute"
-                        " the appropriate device map, you should upgrade your `accelerate` library,"
-                        "`pip install --upgrade accelerate` or install it from source to support fp4 auto device map"
+                        " the appropriate device map, you should upgrade your `accelerate` library, "
+                        "`pip install --upgrade accelerate` or install it from source to support fp4 auto device map "
                         "calculation. You may encounter unexpected behavior, or pass your own device map"
                     )
             elif load_in_8bit:
@@ -3202,7 +3202,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
 
             if model._no_split_modules is None:
                 raise ValueError(
-                    f"{model.__class__.__name__} does not support `device_map='{device_map}'`. To implement support, the model"
+                    f"{model.__class__.__name__} does not support `device_map='{device_map}'`. To implement support, the model "
                     "class needs to implement the `_no_split_modules` attribute."
                 )
             no_split_modules = model._no_split_modules