fix: non-atomic checkpoint save (#27820)

src/transformers/trainer.py

@@ -2332,13 +2332,21 @@ class Trainer:
 
         run_dir = self._get_output_dir(trial=trial)
         output_dir = os.path.join(run_dir, checkpoint_folder)
-        self.save_model(output_dir, _internal_call=True)
+        if os.path.exists(output_dir) and len(os.listdir(output_dir)) > 0:
+            logger.warning(
+                f"Checkpoint destination directory {output_dir} already exists and is non-empty."
+                "Saving will proceed but saved results may be invalid."
+            )
+            staging_output_dir = output_dir
+        else:
+            staging_output_dir = os.path.join(run_dir, f"tmp-{checkpoint_folder}")
+        self.save_model(staging_output_dir, _internal_call=True)
 
         if not self.args.save_only_model:
             # Save optimizer and scheduler
-            self._save_optimizer_and_scheduler(output_dir)
+            self._save_optimizer_and_scheduler(staging_output_dir)
             # Save RNG state
-            self._save_rng_state(output_dir)
+            self._save_rng_state(staging_output_dir)
 
         # Determine the new best metric / best model checkpoint
         if metrics is not None and self.args.metric_for_best_model is not None:
@@ -2358,10 +2366,14 @@ class Trainer:
 
         # Save the Trainer state
         if self.args.should_save:
-            self.state.save_to_json(os.path.join(output_dir, TRAINER_STATE_NAME))
+            self.state.save_to_json(os.path.join(staging_output_dir, TRAINER_STATE_NAME))
 
         if self.args.push_to_hub:
-            self._push_from_checkpoint(output_dir)
+            self._push_from_checkpoint(staging_output_dir)
+
+        # Place checkpoint in final location after all saving is finished.
+        if staging_output_dir != output_dir:
+            os.rename(staging_output_dir, output_dir)
 
         # Maybe delete some older checkpoints.
         if self.args.should_save:

tests/trainer/test_trainer.py

@@ -79,7 +79,8 @@ from transformers.testing_utils import (
     slow,
     torch_device,
 )
-from transformers.trainer_utils import PREFIX_CHECKPOINT_DIR, HPSearchBackend
+from transformers.tokenization_utils_base import PreTrainedTokenizerBase
+from transformers.trainer_utils import PREFIX_CHECKPOINT_DIR, HPSearchBackend, get_last_checkpoint
 from transformers.training_args import OptimizerNames
 from transformers.utils import (
     SAFE_WEIGHTS_INDEX_NAME,
@@ -1310,6 +1311,19 @@ class TrainerIntegrationTest(TestCasePlus, TrainerIntegrationCommon):
             trainer.train()
             self.check_saved_checkpoints(tmpdir, 5, int(self.n_epochs * 64 / self.batch_size), False)
 
+    def test_save_checkpoints_is_atomic(self):
+        class UnsaveableTokenizer(PreTrainedTokenizerBase):
+            def save_pretrained(self, *args, **kwargs):
+                raise OSError("simulated file write error")
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            trainer = get_regression_trainer(output_dir=tmpdir, save_steps=5)
+            # Attach unsaveable tokenizer to partially fail checkpointing
+            trainer.tokenizer = UnsaveableTokenizer()
+            with self.assertRaises(OSError) as _context:
+                trainer.train()
+            assert get_last_checkpoint(tmpdir) is None
+
     @require_safetensors
     def test_safe_checkpoints(self):
         for save_safetensors in [True, False]: