examples: add keep_linebreaks option to CLM examples (#13150)

* examples: add keep_linebreaks option to text dataset loader for all CLM examples

* examples: introduce new keep_linebreaks option as data argument in CLM examples

Author: stefan-it

examples/flax/language-modeling/run_clm_flax.py

@@ -156,6 +156,9 @@ class DataTrainingArguments:
         default=None,
         metadata={"help": "The number of processes to use for the preprocessing."},
     )
+    keep_linebreaks: bool = field(
+        default=True, metadata={"help": "Whether to keep line breaks when using CSV/JSON/TXT files or not."}
+    )
 
     def __post_init__(self):
         if self.dataset_name is None and self.train_file is None and self.validation_file is None:
@@ -314,12 +317,14 @@ def main():
         if "validation" not in dataset.keys():
             dataset["validation"] = load_dataset(
                 extension,
+                keep_linebreaks=data_args.keep_linebreaks,
                 data_files=data_files,
                 split=f"train[:{data_args.validation_split_percentage}%]",
                 cache_dir=model_args.cache_dir,
             )
             dataset["train"] = load_dataset(
                 extension,
+                keep_linebreaks=data_args.keep_linebreaks,
                 data_files=data_files,
                 split=f"train[{data_args.validation_split_percentage}%:]",
                 cache_dir=model_args.cache_dir,

examples/pytorch/language-modeling/run_clm.py

@@ -172,6 +172,9 @@ class DataTrainingArguments:
         default=None,
         metadata={"help": "The number of processes to use for the preprocessing."},
     )
+    keep_linebreaks: bool = field(
+        default=True, metadata={"help": "Whether to keep line breaks when using CSV/JSON/TXT files or not."}
+    )
 
     def __post_init__(self):
         if self.dataset_name is None and self.train_file is None and self.validation_file is None:
@@ -282,12 +285,14 @@ def main():
         if "validation" not in raw_datasets.keys():
             raw_datasets["validation"] = load_dataset(
                 extension,
+                keep_linebreaks=data_args.keep_linebreaks,
                 data_files=data_files,
                 split=f"train[:{data_args.validation_split_percentage}%]",
                 cache_dir=model_args.cache_dir,
             )
             raw_datasets["train"] = load_dataset(
                 extension,
+                keep_linebreaks=data_args.keep_linebreaks,
                 data_files=data_files,
                 split=f"train[{data_args.validation_split_percentage}%:]",
                 cache_dir=model_args.cache_dir,

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -173,6 +173,9 @@ def parse_args():
     parser.add_argument(
         "--overwrite_cache", type=bool, default=False, help="Overwrite the cached training and evaluation sets"
     )
+    parser.add_argument(
+        "--no_keep_linebreaks", action="store_true", help="Do not keep line breaks when using CSV/JSON/TXT files."
+    )
 
     args = parser.parse_args()
 
@@ -257,11 +260,13 @@ def main():
         if "validation" not in raw_datasets.keys():
             raw_datasets["validation"] = load_dataset(
                 extension,
+                keep_linebreaks=not args.no_keep_linebreaks,
                 data_files=data_files,
                 split=f"train[:{args.validation_split_percentage}%]",
             )
             raw_datasets["train"] = load_dataset(
                 extension,
+                keep_linebreaks=not args.no_keep_linebreaks,
                 data_files=data_files,
                 split=f"train[{args.validation_split_percentage}%:]",
             )

examples/tensorflow/language-modeling/run_clm.py

@@ -186,6 +186,9 @@ class DataTrainingArguments:
             "value if set."
         },
     )
+    keep_linebreaks: bool = field(
+        default=True, metadata={"help": "Whether to keep line breaks when using CSV/JSON/TXT files or not."}
+    )
 
     def __post_init__(self):
         if self.dataset_name is None and self.train_file is None and self.validation_file is None:
@@ -325,7 +328,7 @@ def main():
         extension = data_args.train_file.split(".")[-1]
         if extension == "txt":
             extension = "text"
-        raw_datasets = load_dataset(extension, data_files=data_files)
+        raw_datasets = load_dataset(extension, keep_linebreaks=data_args.keep_linebreaks, data_files=data_files)
     # See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at
     # https://huggingface.co/docs/datasets/loading_datasets.html.
     # endregion