Merge pull request #69 from kundajelab/keras2.4-fix

Keras 2.2.3 saved file API breaking fix

.travis.yml

@@ -8,7 +8,7 @@ matrix:
         - python: 2.7
           env: KERAS_BACKEND=tensorflow TF_VERSION=1.10.1 KERAS_VERSION=2.2
         - python: 3.6
-          env: KERAS_BACKEND=tensorflow TF_VERSION=1.10.1 KERAS_VERSION=2.2
+          env: KERAS_BACKEND=tensorflow TF_VERSION=1.10.1 KERAS_VERSION=2.2.4
 
 notifications:
   email: true

deeplift.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: deeplift
-Version: 0.6.8.0
+Version: 0.6.8.1
 Summary: DeepLIFT (Deep Learning Important FeaTures)
 Home-page: https://github.com/kundajelab/deeplift
 License: UNKNOWN

deeplift/__init__.py

@@ -1 +1 @@
-__version__ = '0.6.6.2'
+__version__ = '0.6.8.1'

deeplift/conversion/kerasapi_conversion.py

@@ -355,6 +355,8 @@ def convert_model_from_saved_files(
         model_weights=model_weights['model_weights']
 
     if (model_class_name=="Sequential"):
+        if (isinstance(model_config, list)==False): #keras 2.2.3 API change
+            model_config = model_config["layers"]
         layer_configs = model_config
         model_conversion_function = convert_sequential_model
     elif (model_class_name=="Model"):
@@ -386,7 +388,10 @@ def convert_model_from_saved_files(
         elif (layer_config["class_name"]=="Sequential"):
             insert_weights_into_nested_model_config(
                 nested_model_weights=nested_model_weights,
-                nested_model_layer_config=layer_config["config"])
+                nested_model_layer_config=
+                  (layer_config["config"] if
+                   isinstance(layer_config["config"], list)
+                   else layer_config["config"]["layers"]))
         else:  
             layer_weights = [np.array(model_weights[layer_name][x]) for x in
                              model_weights[layer_name].attrs["weight_names"]]
@@ -406,7 +411,10 @@ def insert_weights_into_nested_model_config(nested_model_weights,
             elif (layer_config["class_name"]=="Sequential"):
                 insert_weights_into_nested_model_config(
                     nested_model_weights=nested_model_weights,
-                    nested_model_layer_config=layer_config["config"])
+                    nested_model_layer_config=
+                      (layer_config["config"]
+                       if isinstance(layer_config["config"],list)
+                       else layer_config["config"]["layers"]))
             else: 
                 layer_name = layer_config["config"]["name"] 
                 layer_weights = [np.array(nested_model_weights[x]) for x in

setup.py

@@ -8,7 +8,7 @@
 Implements the methods in "Learning Important Features Through Propagating Activation Differences" by Shrikumar, Greenside & Kundaje, as well as other commonly-used methods such as gradients, guided backprop and integrated gradients. See https://github.com/kundajelab/deeplift for documentation and FAQ.
           """,
           url='https://github.com/kundajelab/deeplift',
-          version='0.6.8.0',
+          version='0.6.8.1',
           packages=['deeplift',
                     'deeplift.layers', 'deeplift.visualization',
                     'deeplift.conversion'],