Fix multiple deletions of the same files in save_pretrained

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -444,7 +444,11 @@ def collate_fn(examples):
                     if args.push_to_hub and epoch < args.num_train_epochs - 1:
                         accelerator.wait_for_everyone()
                         unwrapped_model = accelerator.unwrap_model(model)
-                        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+                        unwrapped_model.save_pretrained(
+                            args.output_dir,
+                            is_main_process=accelerator.is_main_process,
+                            save_function=accelerator.save,
+                        )
                         if accelerator.is_main_process:
                             feature_extractor.save_pretrained(args.output_dir)
                             repo.push_to_hub(
@@ -490,7 +494,9 @@ def collate_fn(examples):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 feature_extractor.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -506,7 +512,9 @@ def collate_fn(examples):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             feature_extractor.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -580,7 +580,9 @@ def group_texts(examples):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -596,7 +598,9 @@ def group_texts(examples):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/language-modeling/run_mlm_no_trainer.py

@@ -627,7 +627,9 @@ def group_texts(examples):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -643,7 +645,9 @@ def group_texts(examples):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/multiple-choice/run_swag_no_trainer.py

@@ -588,7 +588,9 @@ def preprocess_function(examples):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -604,7 +606,9 @@ def preprocess_function(examples):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/question-answering/run_qa_beam_search_no_trainer.py

@@ -794,7 +794,9 @@ def create_and_fill_np_array(start_or_end_logits, dataset, max_len):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -932,7 +934,9 @@ def create_and_fill_np_array(start_or_end_logits, dataset, max_len):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/question-answering/run_qa_no_trainer.py

@@ -810,7 +810,9 @@ def create_and_fill_np_array(start_or_end_logits, dataset, max_len):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -902,7 +904,9 @@ def create_and_fill_np_array(start_or_end_logits, dataset, max_len):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py

@@ -553,7 +553,11 @@ def preprocess_val(example_batch):
                     if args.push_to_hub and epoch < args.num_train_epochs - 1:
                         accelerator.wait_for_everyone()
                         unwrapped_model = accelerator.unwrap_model(model)
-                        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+                        unwrapped_model.save_pretrained(
+                            args.output_dir,
+                            is_main_process=accelerator.is_main_process,
+                            save_function=accelerator.save,
+                        )
                         if accelerator.is_main_process:
                             feature_extractor.save_pretrained(args.output_dir)
                             repo.push_to_hub(
@@ -613,7 +617,9 @@ def preprocess_val(example_batch):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 feature_extractor.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -629,7 +635,9 @@ def preprocess_val(example_batch):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             feature_extractor.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/speech-pretraining/run_wav2vec2_pretraining_no_trainer.py

@@ -669,7 +669,9 @@ def prepare_dataset(batch):
                 if (args.push_to_hub and epoch < args.num_train_epochs - 1) or args.output_dir is not None:
                     accelerator.wait_for_everyone()
                     unwrapped_model = accelerator.unwrap_model(model)
-                    unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+                    unwrapped_model.save_pretrained(
+                        args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+                    )
 
                 if (args.push_to_hub and epoch < args.num_train_epochs - 1) and accelerator.is_main_process:
                     repo.push_to_hub(
@@ -720,7 +722,9 @@ def prepare_dataset(batch):
         if args.output_dir is not None:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 if args.push_to_hub:
                     repo.push_to_hub(commit_message="End of training", auto_lfs_prune=True)

examples/pytorch/summarization/run_summarization_no_trainer.py

@@ -687,7 +687,9 @@ def postprocess_text(preds, labels):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -703,7 +705,9 @@ def postprocess_text(preds, labels):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/text-classification/run_glue_no_trainer.py

@@ -539,7 +539,9 @@ def preprocess_function(examples):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -555,7 +557,9 @@ def preprocess_function(examples):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/token-classification/run_ner_no_trainer.py

@@ -691,7 +691,9 @@ def compute_metrics():
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -707,7 +709,9 @@ def compute_metrics():
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

examples/pytorch/translation/run_translation_no_trainer.py

@@ -662,7 +662,9 @@ def postprocess_text(preds, labels):
         if args.push_to_hub and epoch < args.num_train_epochs - 1:
             accelerator.wait_for_everyone()
             unwrapped_model = accelerator.unwrap_model(model)
-            unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
             if accelerator.is_main_process:
                 tokenizer.save_pretrained(args.output_dir)
                 repo.push_to_hub(
@@ -678,7 +680,9 @@ def postprocess_text(preds, labels):
     if args.output_dir is not None:
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
-        unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
         if accelerator.is_main_process:
             tokenizer.save_pretrained(args.output_dir)
             if args.push_to_hub:

src/transformers/modeling_utils.py

@@ -20,6 +20,7 @@
 import re
 import shutil
 import tempfile
+import warnings
 from contextlib import contextmanager
 from dataclasses import dataclass
 from functools import partial
@@ -1347,7 +1348,7 @@ def is_gradient_checkpointing(self) -> bool:
     def save_pretrained(
         self,
         save_directory: Union[str, os.PathLike],
-        save_config: bool = True,
+        is_main_process: bool = True,
         state_dict: Optional[dict] = None,
         save_function: Callable = torch.save,
         push_to_hub: bool = False,
@@ -1361,10 +1362,10 @@ def save_pretrained(
         Arguments:
             save_directory (`str` or `os.PathLike`):
                 Directory to which to save. Will be created if it doesn't exist.
-            save_config (`bool`, *optional*, defaults to `True`):
-                Whether or not to save the config of the model. Useful when in distributed training like TPUs and need
-                to call this function on all processes. In this case, set `save_config=True` only on the main process
-                to avoid race conditions.
+            is_main_process (`bool`, *optional*, defaults to `True`):
+                Whether the process calling this is the main process or not. Useful when in distributed training like
+                TPUs and need to call this function on all processes. In this case, set `is_main_process=True` only on
+                the main process to avoid race conditions.
             state_dict (nested dictionary of `torch.Tensor`):
                 The state dictionary of the model to save. Will default to `self.state_dict()`, but can be used to only
                 save parts of the model or if special precautions need to be taken when recovering the state dictionary
@@ -1397,6 +1398,12 @@ def save_pretrained(
             kwargs:
                 Additional key word arguments passed along to the [`~utils.PushToHubMixin.push_to_hub`] method.
         """
+        if "save_config" in kwargs:
+            warnings.warn(
+                "`save_config` is deprecated and will be removed in v5 of Transformers. Use `is_main_process` instead."
+            )
+            is_main_process = kwargs.pop("save_config")
+
         if os.path.isfile(save_directory):
             logger.error(f"Provided path ({save_directory}) should be a directory, not a file")
             return
@@ -1424,7 +1431,7 @@ def save_pretrained(
             custom_object_save(self, save_directory, config=self.config)
 
         # Save the config
-        if save_config:
+        if is_main_process:
             model_to_save.config.save_pretrained(save_directory)
 
         # Save the model
@@ -1443,7 +1450,14 @@ def save_pretrained(
         # Clean the folder from a previous save
         for filename in os.listdir(save_directory):
             full_filename = os.path.join(save_directory, filename)
-            if filename.startswith(WEIGHTS_NAME[:-4]) and os.path.isfile(full_filename):
+            # If we have a shard file that is not going to be replaced, we delete it, but only from the main process
+            # in distributed settings to avoid race conditions.
+            if (
+                filename.startswith(WEIGHTS_NAME[:-4])
+                and os.path.isfile(full_filename)
+                and filename not in shards.keys()
+                and is_main_process
+            ):
                 os.remove(full_filename)
 
         # Save the model