Merge branch 'huggingface:main' into triton-version

src/accelerate/data_loader.py

@@ -12,11 +12,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import importlib
 import math
 from contextlib import suppress
 from typing import Callable, List, Optional, Union
 
 import torch
+from packaging import version
 from torch.utils.data import BatchSampler, DataLoader, IterableDataset, RandomSampler
 
 from .logging import get_logger
@@ -25,6 +27,7 @@
     RNGType,
     broadcast,
     broadcast_object_list,
+    compare_versions,
     concatenate,
     find_batch_size,
     get_data_structure,
@@ -415,6 +418,13 @@ def __init__(self, dataset, use_stateful_dataloader=False, batch_sampler=None, *
                 "StatefulDataLoader is not available. Please install torchdata version 0.8.0 or higher to use it."
             )
         if use_stateful_dataloader:
+            torchdata_version = version.parse(importlib.metadata.version("torchdata"))
+            if (
+                "in_order" in kwargs
+                and compare_versions(torchdata_version, "<", "0.11")
+                and is_torch_version(">=", "2.6.0")
+            ):
+                kwargs.pop("in_order")
             self.base_dataloader = StatefulDataLoader(dataset, batch_sampler=batch_sampler, **kwargs)
         else:
             self.base_dataloader = DataLoader(dataset, batch_sampler=batch_sampler, **kwargs)

src/accelerate/utils/dataclasses.py

@@ -230,7 +230,7 @@ class GradScalerKwargs(KwargsHandler):
     from accelerate import Accelerator
     from accelerate.utils import GradScalerKwargs
 
-    kwargs = GradScalerKwargs(backoff_filter=0.25)
+    kwargs = GradScalerKwargs(backoff_factor=0.25)
     accelerator = Accelerator(kwargs_handlers=[kwargs])
     ```
     """