fix: adding absolute paths for custom_tf_addons in randaugment

Author: init-22

algorithmic_efficiency/workloads/imagenet_resnet/imagenet_jax/randaugment.py

@@ -9,9 +9,12 @@
 
 import tensorflow as tf
 
-from .custom_tf_addons import rotate_img
-from .custom_tf_addons import transform
-from .custom_tf_addons import translate
+from algorithmic_efficiency.workloads.imagenet_resnet.imagenet_jax.custom_tf_addons import \
+    rotate_img
+from algorithmic_efficiency.workloads.imagenet_resnet.imagenet_jax.custom_tf_addons import \
+    transform
+from algorithmic_efficiency.workloads.imagenet_resnet.imagenet_jax.custom_tf_addons import \
+    translate
 
 # This signifies the max integer that the controller RNN could predict for the
 # augmentation scheme.