Merge branch 'rvankoert:master' into master

knaw-huc · Oct 25, 2023 · e1b853c · e1b853c
2 parents 252165d + 18df7a5
commit e1b853c
Show file tree

Hide file tree

Showing 3 changed files with 89 additions and 89 deletions.
diff --git a/src/vgsl_model_generator.py b/src/vgsl_model_generator.py
@@ -322,10 +322,10 @@ def get_model_libary() -> dict:
             #      "O1s92")
             "model17":
                 ("None,64,None,4 Bn Ce3,3,16 "
-                 "RB3,3,16 "
-                 "RBd3,3,32 "
-                 "RBd3,3,64 "
-                 "RBd3,3,128 "
+                 "RB3,3,16 RB3,3,16 "
+                 "RBd3,3,32 RB3,3,32 RB3,3,32 RB3,3,32 RB3,3,32"
+                 "RBd3,3,64 RB3,3,64 RB3,3,64 RB3,3,64 RB3,3,64 "
+                 "RBd3,3,128 RB3,3,128"
                  "Rc "
                  "Bl256,D20 Bl256,D20 Bl256,D20 "
                  "O1s92")

diff --git a/tests/test_model_creation.py b/tests/test_model_creation.py
@@ -653,31 +653,31 @@ def test_bidirectional_error_handling(self):
                          "number of units, 'rate' is the (recurrent) dropout "
                          "rate.")
 
-    def test_residual_block(self):
-        vgsl_spec_string = "None,64,None,1 RB3,3,16 O1s10"
-        model_generator = self.VGSLModelGenerator(vgsl_spec_string)
-        model = model_generator.build()
-        self.assertIsInstance(model.layers[1], self.ResidualBlock)
-
-        # Layer-specific tests
-        self.assertEqual(model.layers[1].conv1.filters, 16)
-        self.assertEqual(model.layers[1].conv1.kernel_size, (3, 3))
-        self.assertEqual(model.layers[1].conv2.filters, 16)
-        self.assertEqual(model.layers[1].conv2.kernel_size, (3, 3))
-
-        # Create a model with downsampling
-        vgsl_spec_string = "None,64,None,1 RBd3,3,16 O1s10"
-        model_generator = self.VGSLModelGenerator(vgsl_spec_string)
-        model = model_generator.build()
-
-        # Check that the downsampling layer exists
-        self.assertIsInstance(model.layers[1].conv3, layers.Conv2D)
-        self.assertEqual(model.layers[1].conv3.filters, 16)
-        self.assertEqual(model.layers[1].conv3.kernel_size, (1, 1))
-        self.assertEqual(model.layers[1].conv3.strides, (2, 2))
-
-        # Check that conv1 also has strides of 2
-        self.assertEqual(model.layers[1].conv1.strides, (2, 2))
+    # def test_residual_block(self):
+    #     vgsl_spec_string = "None,64,None,1 RB3,3,16 O1s10"
+    #     model_generator = self.VGSLModelGenerator(vgsl_spec_string)
+    #     model = model_generator.build()
+    #     self.assertIsInstance(model.layers[1], self.ResidualBlock)
+    # 
+    #     # Layer-specific tests
+    #     self.assertEqual(model.layers[1].conv1.filters, 16)
+    #     self.assertEqual(model.layers[1].conv1.kernel_size, (3, 3))
+    #     self.assertEqual(model.layers[1].conv2.filters, 16)
+    #     self.assertEqual(model.layers[1].conv2.kernel_size, (3, 3))
+    # 
+    #     # Create a model with downsampling
+    #     vgsl_spec_string = "None,64,None,1 RBd3,3,16 O1s10"
+    #     model_generator = self.VGSLModelGenerator(vgsl_spec_string)
+    #     model = model_generator.build()
+    # 
+    #     # Check that the downsampling layer exists
+    #     self.assertIsInstance(model.layers[1].conv3, layers.Conv2D)
+    #     self.assertEqual(model.layers[1].conv3.filters, 16)
+    #     self.assertEqual(model.layers[1].conv3.kernel_size, (1, 1))
+    #     self.assertEqual(model.layers[1].conv3.strides, (2, 2))
+    # 
+    #     # Check that conv1 also has strides of 2
+    #     self.assertEqual(model.layers[1].conv1.strides, (2, 2))
 
     def test_residual_block_error_handling(self):
         # Invalid format

diff --git a/tests/test_model_to_vgsl.py b/tests/test_model_to_vgsl.py
@@ -297,66 +297,66 @@ def test_residual_block_to_vgsl(self):
         self.assertEqual(
             vgsl_spec, "None,64,None,1 RBd4,2,16")
 
-    def test_functional_combination_to_vgsl(self):
-        # Define the original model using the functional API
-        input_tensor = layers.Input(shape=(None, 64, 1))
-        x = layers.Conv2D(32, (3, 3), activation='relu')(input_tensor)
-        x = layers.MaxPooling2D(pool_size=(2, 2))(x)
-        x = layers.AveragePooling2D(pool_size=(2, 2), strides=(1, 1))(x)
-        x = layers.BatchNormalization()(x)
-        x = layers.Dropout(0.25)(x)
-        x = self.ResidualBlock(32, 3, 3)(x)
-        x = layers.Reshape((-1, 1024))(x)
-        x = layers.LSTM(64, return_sequences=True)(x)
-        x = layers.GRU(64, return_sequences=True, go_backwards=True)(x)
-        x = layers.Bidirectional(layers.LSTM(256, return_sequences=True))(x)
-        x = layers.Bidirectional(layers.GRU(256, return_sequences=True))(x)
-        x = layers.Dense(10, activation='softmax')(x)
-        output_tensor = layers.Activation('linear')(x)
-        model = tf.keras.Model(inputs=input_tensor, outputs=output_tensor)
-
-        vgsl_spec = self.VGSLModelGenerator.model_to_vgsl(model)
-        self.assertEqual(
-            vgsl_spec, "None,64,None,1 Cr3,3,32 Mp2,2,2,2 Ap2,2,1,1 Bn D25 "
-                       "RB3,3,32 Rc Lfs64 Grs64 Bl256 Bg256 O1s10")
-
-        # Generate a model from the VGSL string
-        model_generator = self.VGSLModelGenerator(vgsl_spec)
-        generated_model = model_generator.build()
-
-        # Compare the original model with the generated model
-        self.compare_model_configs(model, generated_model)
-
-    def test_sequential_combination_to_vgsl(self):
-        # Define the original model using the sequential API
-        model = tf.keras.Sequential([
-            layers.Input(shape=(None, 64, 1)),
-            layers.Conv2D(32, (3, 3), activation='relu'),
-            layers.MaxPooling2D(pool_size=(2, 2)),
-            layers.AveragePooling2D(pool_size=(2, 2), strides=(1, 1)),
-            layers.BatchNormalization(),
-            layers.Dropout(0.25),
-            self.ResidualBlock(32, 3, 3),
-            layers.Reshape((-1, 1024)),
-            layers.LSTM(64, return_sequences=True),
-            layers.GRU(64, return_sequences=True, go_backwards=True),
-            layers.Bidirectional(layers.LSTM(256, return_sequences=True)),
-            layers.Bidirectional(layers.GRU(256, return_sequences=True)),
-            layers.Dense(10, activation='softmax'),
-            layers.Activation('linear')
-        ])
-
-        vgsl_spec = self.VGSLModelGenerator.model_to_vgsl(model)
-        self.assertEqual(
-            vgsl_spec, "None,64,None,1 Cr3,3,32 Mp2,2,2,2 Ap2,2,1,1 Bn D25 "
-                       "RB3,3,32 Rc Lfs64 Grs64 Bl256 Bg256 O1s10")
-
-        # Generate a model from the VGSL string
-        model_generator = self.VGSLModelGenerator(vgsl_spec)
-        generated_model = model_generator.build()
-
-        # Compare the original model with the generated model
-        self.compare_model_configs(model, generated_model)
+    # def test_functional_combination_to_vgsl(self):
+    #     # Define the original model using the functional API
+    #     input_tensor = layers.Input(shape=(None, 64, 1))
+    #     x = layers.Conv2D(32, (3, 3), activation='relu')(input_tensor)
+    #     x = layers.MaxPooling2D(pool_size=(2, 2))(x)
+    #     x = layers.AveragePooling2D(pool_size=(2, 2), strides=(1, 1))(x)
+    #     x = layers.BatchNormalization()(x)
+    #     x = layers.Dropout(0.25)(x)
+    #     x = self.ResidualBlock(32, 3, 3)(x)
+    #     x = layers.Reshape((-1, 1024))(x)
+    #     x = layers.LSTM(64, return_sequences=True)(x)
+    #     x = layers.GRU(64, return_sequences=True, go_backwards=True)(x)
+    #     x = layers.Bidirectional(layers.LSTM(256, return_sequences=True))(x)
+    #     x = layers.Bidirectional(layers.GRU(256, return_sequences=True))(x)
+    #     x = layers.Dense(10, activation='softmax')(x)
+    #     output_tensor = layers.Activation('linear')(x)
+    #     model = tf.keras.Model(inputs=input_tensor, outputs=output_tensor)
+    # 
+    #     vgsl_spec = self.VGSLModelGenerator.model_to_vgsl(model)
+    #     self.assertEqual(
+    #         vgsl_spec, "None,64,None,1 Cr3,3,32 Mp2,2,2,2 Ap2,2,1,1 Bn D25 "
+    #                    "RB3,3,32 Rc Lfs64 Grs64 Bl256 Bg256 O1s10")
+    # 
+    #     # Generate a model from the VGSL string
+    #     model_generator = self.VGSLModelGenerator(vgsl_spec)
+    #     generated_model = model_generator.build()
+    # 
+    #     # Compare the original model with the generated model
+    #     self.compare_model_configs(model, generated_model)
+
+    # def test_sequential_combination_to_vgsl(self):
+    #     # Define the original model using the sequential API
+    #     model = tf.keras.Sequential([
+    #         layers.Input(shape=(None, 64, 1)),
+    #         layers.Conv2D(32, (3, 3), activation='relu'),
+    #         layers.MaxPooling2D(pool_size=(2, 2)),
+    #         layers.AveragePooling2D(pool_size=(2, 2), strides=(1, 1)),
+    #         layers.BatchNormalization(),
+    #         layers.Dropout(0.25),
+    #         self.ResidualBlock(32, 3, 3),
+    #         layers.Reshape((-1, 1024)),
+    #         layers.LSTM(64, return_sequences=True),
+    #         layers.GRU(64, return_sequences=True, go_backwards=True),
+    #         layers.Bidirectional(layers.LSTM(256, return_sequences=True)),
+    #         layers.Bidirectional(layers.GRU(256, return_sequences=True)),
+    #         layers.Dense(10, activation='softmax'),
+    #         layers.Activation('linear')
+    #     ])
+    #
+    #     vgsl_spec = self.VGSLModelGenerator.model_to_vgsl(model)
+    #     self.assertEqual(
+    #         vgsl_spec, "None,64,None,1 Cr3,3,32 Mp2,2,2,2 Ap2,2,1,1 Bn D25 "
+    #                    "RB3,3,32 Rc Lfs64 Grs64 Bl256 Bg256 O1s10")
+    #
+    #     # Generate a model from the VGSL string
+    #     model_generator = self.VGSLModelGenerator(vgsl_spec)
+    #     generated_model = model_generator.build()
+    #
+    #     # Compare the original model with the generated model
+    #     self.compare_model_configs(model, generated_model)
 
     def compare_model_configs(self, model, generated_model):