fix: adding absolute paths for custom_tf_addons in randaugment

init-22 · Dec 19, 2024 · 53eff1d · 53eff1d
1 parent 9393145
commit 53eff1d
Showing 1 changed file with 6 additions and 3 deletions.
diff --git a/algorithmic_efficiency/workloads/imagenet_resnet/imagenet_jax/randaugment.py b/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.