Propagate

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -447,7 +447,7 @@ def collate_fn(examples):
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
             checkpoint_path = path
-            path = os.path.basename(args.resume_from_checkpoint)
+            path = os.path.basename(checkpoint_path)
 
         accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
         accelerator.load_state(path)

examples/pytorch/image-pretraining/run_mim_no_trainer.py

@@ -649,17 +649,18 @@ def preprocess_images(examples):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -554,17 +554,18 @@ def group_texts(examples):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/language-modeling/run_mlm_no_trainer.py

@@ -598,17 +598,18 @@ def group_texts(examples):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/multiple-choice/run_swag_no_trainer.py

@@ -545,17 +545,18 @@ def preprocess_function(examples):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/question-answering/run_qa_beam_search_no_trainer.py

@@ -797,17 +797,18 @@ def create_and_fill_np_array(start_or_end_logits, dataset, max_len):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/question-answering/run_qa_no_trainer.py

@@ -813,17 +813,18 @@ def create_and_fill_np_array(start_or_end_logits, dataset, max_len):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py

@@ -542,17 +542,18 @@ def preprocess_val(example_batch):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/summarization/run_summarization_no_trainer.py

@@ -614,17 +614,18 @@ def postprocess_text(preds, labels):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/text-classification/run_glue_no_trainer.py

@@ -489,17 +489,18 @@ def preprocess_function(examples):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/token-classification/run_ner_no_trainer.py

@@ -648,17 +648,18 @@ def compute_metrics():
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]
 

examples/pytorch/translation/run_translation_no_trainer.py

@@ -595,17 +595,18 @@ def postprocess_text(preds, labels):
     # Potentially load in the weights and states from a previous save
     if args.resume_from_checkpoint:
         if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
-            accelerator.print(f"Resumed from checkpoint: {args.resume_from_checkpoint}")
-            accelerator.load_state(args.resume_from_checkpoint)
+            checkpoint_path = args.resume_from_checkpoint
             path = os.path.basename(args.resume_from_checkpoint)
         else:
             # Get the most recent checkpoint
             dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
             dirs.sort(key=os.path.getctime)
             path = dirs[-1]  # Sorts folders by date modified, most recent checkpoint is the last
-            accelerator.print(f"Resumed from checkpoint: {path}")
-            accelerator.load_state(path)
-            path = os.path.basename(args.resume_from_checkpoint)
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(path)
         # Extract `epoch_{i}` or `step_{i}`
         training_difference = os.path.splitext(path)[0]