diff --git a/keras_export/convert_model.py b/keras_export/convert_model.py
index fe78b910..976eb653 100755
--- a/keras_export/convert_model.py
+++ b/keras_export/convert_model.py
@@ -385,7 +385,8 @@ def show_gru_layer(layer):
 
 
 def transform_cudnn_weights(input_weights, recurrent_weights, n_gates):
-    return transform_kernels(input_weights, n_gates, transform_input_kernel), transform_kernels(recurrent_weights, n_gates, transform_recurrent_kernel)
+    return transform_kernels(input_weights, n_gates, transform_input_kernel), \
+           transform_kernels(recurrent_weights, n_gates, transform_recurrent_kernel)
 
 
 def show_cudnn_lstm_layer(layer):
@@ -440,16 +441,17 @@ def get_transform_func(layer):
     return input_transform_func, recurrent_transform_func, bias_transform_func
 
 
-
 def show_bidirectional_layer(layer):
     """Serialize Bidirectional layer to dict"""
     forward_weights = layer.forward_layer.get_weights()
     assert len(forward_weights) == 2 or len(forward_weights) == 3
-    forward_input_transform_func, forward_recurrent_transform_func, forward_bias_transform_func = get_transform_func(layer.forward_layer)
+    forward_input_transform_func, forward_recurrent_transform_func, forward_bias_transform_func = get_transform_func(
+        layer.forward_layer)
 
     backward_weights = layer.backward_layer.get_weights()
     assert len(backward_weights) == 2 or len(backward_weights) == 3
-    backward_input_transform_func, backward_recurrent_transform_func, backward_bias_transform_func = get_transform_func(layer.backward_layer)
+    backward_input_transform_func, backward_recurrent_transform_func, backward_bias_transform_func = get_transform_func(
+        layer.backward_layer)
 
     result = {'forward_weights': encode_floats(forward_input_transform_func(forward_weights[0])),
               'forward_recurrent_weights': encode_floats(forward_recurrent_transform_func(forward_weights[1])),
diff --git a/keras_export/generate_test_models.py b/keras_export/generate_test_models.py
index 9e36d247..e915e90e 100644
--- a/keras_export/generate_test_models.py
+++ b/keras_export/generate_test_models.py
@@ -347,12 +347,12 @@ def get_test_model_lstm():
     inputs = [Input(shape=s) for s in input_shapes]
     outputs = []
 
-    for input in inputs:    
+    for inp in inputs:
         lstm_sequences = LSTM(
             units=8,
             recurrent_activation='relu',
             return_sequences=True
-        )(input)
+        )(inp)
         lstm_regular = LSTM(
             units=3,
             recurrent_activation='sigmoid',
@@ -366,7 +366,7 @@ def get_test_model_lstm():
                 recurrent_activation='hard_sigmoid',
                 return_sequences=True
             )
-        )(input)
+        )(inp)
         lstm_bidi = Bidirectional(
             LSTM(
                 units=6,
@@ -381,13 +381,13 @@ def get_test_model_lstm():
             # run GPU-enabled mode if GPU is available
             lstm_gpu_regular = keras.layers.CuDNNLSTM(
                 units=3
-            )(input)
+            )(inp)
 
             lstm_gpu_bidi = Bidirectional(
                 keras.layers.CuDNNLSTM(
                     units=3
                 )
-            )(input)
+            )(inp)
         else:
             # fall back to equivalent regular LSTM for CPU-only mode
             lstm_gpu_regular = LSTM(
@@ -395,7 +395,7 @@ def get_test_model_lstm():
                 activation='tanh',
                 recurrent_activation='sigmoid',
                 use_bias=True
-            )(input)
+            )(inp)
 
             lstm_gpu_bidi = Bidirectional(
                 LSTM(
@@ -404,7 +404,7 @@ def get_test_model_lstm():
                     recurrent_activation='sigmoid',
                     use_bias=True
                 )
-            )(input)
+            )(inp)
         outputs.append(lstm_gpu_regular)
         outputs.append(lstm_gpu_bidi)
 
@@ -430,14 +430,14 @@ def get_test_model_gru():
     inputs = [Input(shape=s) for s in input_shapes]
     outputs = []
 
-    for input in inputs:
+    for inp in inputs:
         gru_sequences = GRU(
             units=8,
             recurrent_activation='relu',
             reset_after=True,
             return_sequences=True,
             use_bias=True
-        )(input)
+        )(inp)
         gru_regular = GRU(
             units=3,
             recurrent_activation='sigmoid',
@@ -455,7 +455,7 @@ def get_test_model_gru():
                 return_sequences=True,
                 use_bias=True
             )
-        )(input)
+        )(inp)
         gru_bidi = Bidirectional(
             GRU(
                 units=6,
@@ -472,13 +472,13 @@ def get_test_model_gru():
             # run GPU-enabled mode if GPU is available
             gru_gpu_regular = keras.layers.CuDNNGRU(
                 units=3
-            )(input)
+            )(inp)
 
             gru_gpu_bidi = Bidirectional(
                 keras.layers.CuDNNGRU(
                     units=3
                 )
-            )(input)
+            )(inp)
         else:
             # fall back to equivalent regular GRU for CPU-only mode
             gru_gpu_regular = GRU(
@@ -487,7 +487,7 @@ def get_test_model_gru():
                 recurrent_activation='sigmoid',
                 reset_after=True,
                 use_bias=True
-            )(input)
+            )(inp)
 
             gru_gpu_bidi = Bidirectional(
                 GRU(
@@ -497,7 +497,7 @@ def get_test_model_gru():
                     reset_after=True,
                     use_bias=True
                 )
-            )(input)
+            )(inp)
         outputs.append(gru_gpu_regular)
         outputs.append(gru_gpu_bidi)