Create explainability_test.py

ai/models/explainability_test.py

@@ -0,0 +1,37 @@
+import unittest
+from tensorflow import keras
+from galactic_chain.ai.models import Explainability, ExplainabilityModel
+
+class TestExplainability(unittest.TestCase):
+    def test_attention_visualization(self):
+        model = keras.Sequential([keras.layers.Embedding(input_dim=100, output_dim=128), keras.layers.Attention()])
+        explainability = Explainability(model)
+        input_data = tf.random.normal((1, 10))
+        attention_layer_name = 'attention'
+        explainability.attention_visualization(input_data, attention_layer_name)
+
+    def test_saliency_maps(self):
+        model = keras.Sequential([keras.layers.Embedding(input_dim=100, output_dim=128), keras.layers.Dense(10)])
+        explainability = Explainability(model)
+        input_data = tf.random.normal((1, 10))
+        class_index = 0
+        explainability.saliency_maps(input_data, class_index)
+
+    def test_feature_importance(self):
+        model = keras.Sequential([keras.layers.Embedding(input_dim=100, output_dim=128), keras.layers.Dense(10)])
+        explainability = Explainability(model)
+        input_data = tf.random.normal((1, 10))
+        num_features = 10
+        explainability.feature_importance(input_data, num_features)
+
+    def test_explainability_model(self):
+        model = keras.Sequential([keras.layers.Embedding(input_dim=100, output_dim=128), keras.layers.Attention()])
+        explainability_model = ExplainabilityModel(model)
+        input_data = tf.random.normal((1, 10))
+        attention_layer_name = 'attention'
+        class_index = 0
+        num_features = 10
+        explainability_model.get_explainability(input_data, attention_layer_name, class_index, num_features)
+
+if __name__ == '__main__':
+    unittest.main()