Update code to be compatible with tensorflow 2.3.0

requirements.txt

@@ -1,2 +1,2 @@
-keras>=2.0.0
-tensorflow>=1.4.0
+tensorflow>=2.3.0
+pytest

tests/wrapper_test.py

@@ -1,11 +1,11 @@
 import sys
 import numpy as np
+import pytest
 
-from keras.models import Model
-from keras.layers import Input
-from keras.layers import Lambda
+from tensorflow.keras.models import Model
+from tensorflow.keras.layers import Input
+from tensorflow.keras.layers import Lambda
 
-sys.path.append('..')
 from tta_wrapper import tta_classification
 from tta_wrapper import tta_segmentation
 
@@ -16,7 +16,27 @@ def identity_model(input_shape):
     identity_model = Model(inp, x)
     return identity_model
 
-
+# inputs
+input_sample = np.arange(9).reshape((1, 3, 3, 1))
+
+# outputs
+tta_segmentation_output = input_sample
+tta_classification_output = np.ones((1, 3, 3, 1)) * 4
+
+# model
+seg_identity_model = identity_model(input_sample.shape[1:])
+cls_identity_model = identity_model(input_sample.shape[1:])
+
+@pytest.mark.parametrize("wrapper, base_model, inputs, outputs", 
+                         [(tta_segmentation, 
+                           seg_identity_model, 
+                           input_sample,
+                           tta_segmentation_output),
+                          (tta_classification,
+                          cls_identity_model,
+                          input_sample,
+                          tta_classification_output)
+                          ])
 def test_wrapper(wrapper, base_model, inputs, outputs):
     print('[TEST] wrapping model with {} ... '.format(wrapper.__name__))
 
@@ -35,21 +55,4 @@ def test_wrapper(wrapper, base_model, inputs, outputs):
     prediction = model.predict(inputs)
 
     assert np.allclose(prediction, outputs), f"\nprediction: \n{prediction}\n\nground_truth: \n{outputs}"
-    print('[TEST] {} - test passed. '.format(wrapper.__name__))
-
-
-if __name__ == '__main__':
-
-    # inputs
-    input_sample = np.arange(9).reshape((1, 3, 3, 1))
-
-    # outputs
-    tta_segmentation_output = input_sample
-    tta_classification_output = np.ones((1, 3, 3, 1)) * 4
-
-    # model
-    seg_identity_model = identity_model(input_sample.shape[1:])
-    cls_identity_model = identity_model(input_sample.shape[1:])
-
-    test_wrapper(tta_segmentation, seg_identity_model, input_sample, tta_segmentation_output)
-    test_wrapper(tta_classification, cls_identity_model, input_sample, tta_classification_output)
+    print('[TEST] {} - test passed. '.format(wrapper.__name__))

tta_wrapper/__version__.py

@@ -1,3 +1,3 @@
-VERSION = (0, 0, 1)
+VERSION = (0, 0, 2)
 
 __version__ = '.'.join(map(str, VERSION))

tta_wrapper/functional.py

@@ -1,4 +1,5 @@
 import tensorflow as tf
+from tensorflow.python.ops import manip_ops
 
 
 class DualTransform:
@@ -81,21 +82,21 @@ class HShift(DualTransform):
     identity_param = 0
 
     def forward(self, image, param):
-        return tf.manip.roll(image, param, axis=0)
+        return manip_ops.roll(image, param, axis=0)
 
     def backward(self, image, param):
-        return tf.manip.roll(image, -param, axis=0)
+        return manip_ops.roll(image, -param, axis=0)
 
 
 class VShift(DualTransform):
 
     identity_param = 0
 
     def forward(self, image, param):
-        return tf.manip.roll(image, param, axis=1)
+        return manip_ops.roll(image, param, axis=1)
 
     def backward(self, image, param):
-        return tf.manip.roll(image, -param, axis=1)
+        return manip_ops.roll(image, -param, axis=1)
 
 
 class Contrast(SingleTransform):

tta_wrapper/layers.py

@@ -1,5 +1,5 @@
 import tensorflow as tf
-from keras.layers import Layer
+from tensorflow.keras.layers import Layer
 
 from . import functional as F
 

tta_wrapper/wrappers.py

@@ -1,5 +1,5 @@
-from keras.models import Model
-from keras.layers import Input
+from tensorflow.keras.models import Model
+from tensorflow.keras.layers import Input
 
 from .layers import Repeat, TTA, Merge
 from .augmentation import Augmentation