Add tf_keras imports to prepare for Keras 3 (#28588)

* Port core files + ESM (because ESM code is odd)

* Search-replace in modelling code

* Fix up transfo_xl as well

* Fix other core files + tests (still need to add correct import to tests)

* Fix cookiecutter

* make fixup, fix imports in some more core files

* Auto-add imports to tests

* Cleanup, add imports to sagemaker tests

* Use correct exception for importing tf_keras

* Fixes in modeling_tf_utils

* make fixup

* Correct version parsing code

* Ensure the pipeline tests correctly revert to float32 after each test

* Ensure the pipeline tests correctly revert to float32 after each test

* More tf.keras -> keras

* Add dtype cast

* Better imports of tf_keras

* Add a cast for tf.assign, just in case

* Fix callback imports

examples/tensorflow/image-classification/run_image_classification.py

@@ -47,6 +47,7 @@ from transformers import (
     set_seed,
 )
 from transformers.keras_callbacks import KerasMetricCallback
+from transformers.modeling_tf_utils import keras
 from transformers.trainer_utils import get_last_checkpoint, is_main_process
 from transformers.utils import check_min_version, send_example_telemetry
 from transformers.utils.versions import require_version
@@ -363,7 +364,7 @@ def main():
 
     def _train_transforms(image):
         img_size = image_size
-        image = tf.keras.utils.img_to_array(image)
+        image = keras.utils.img_to_array(image)
         image = random_resized_crop(image, size=img_size)
         image = tf.image.random_flip_left_right(image)
         image /= 255.0
@@ -372,7 +373,7 @@ def main():
         return image
 
     def _val_transforms(image):
-        image = tf.keras.utils.img_to_array(image)
+        image = keras.utils.img_to_array(image)
         image = tf.image.resize(image, size=image_size)
         # image = np.array(image) # FIXME - use tf.image function
         image = center_crop(image, size=image_size)