Internal fix

PiperOrigin-RevId: 493393200

Author: fyangf

official/vision/modeling/layers/edgetpu.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 """EdgeTPU oriented layers and tools."""
-from typing import Optional, Sequence, Union, Iterable
+from typing import List, Optional, Sequence, Union, Iterable
 
 import numpy as np
 import tensorflow as tf
@@ -102,7 +102,7 @@ def _same(x):
 
 def shard_tensors(
     axis: int, block_size: int, tensors: Sequence[tf.Tensor]
-) -> Union[list[Sequence[tf.Tensor]], Iterable[Sequence[tf.Tensor]]]:
+) -> Union[List[Sequence[tf.Tensor]], 'Iterable[Sequence[tf.Tensor]]']:
   """Consistently splits multiple tensors sharding-style.
 
   Args: