Implement save method

anago/models.py

@@ -1,6 +1,9 @@
 """
 Model definition.
 """
+import json
+import os
+
 import keras.backend as K
 from keras.layers import Dense, LSTM, Bidirectional, Embedding, Input, Dropout, Lambda, Activation
 from keras.layers.merge import Concatenate
@@ -18,8 +21,13 @@ def predict(self, X, *args, **kwargs):
         y_pred = self.model.predict(X, batch_size=1)
         return y_pred
 
-    def save(self, filepath):
-        self.model.save_weights(filepath)
+    def save_weights(self, file_path):
+        self.model.save_weights(file_path)
+
+    def save_params(self, file_path):
+        with open(file_path, 'w') as f:
+            params = {name: val for name, val in vars(self).items() if name not in {'_loss', 'model'}}
+            json.dump(params, f, sort_keys=True, indent=4)
 
     def load(self, filepath):
         self.model.load_weights(filepath=filepath)

tests/test_model.py

@@ -30,7 +30,24 @@ def test_predict(self):
         pass
 
     def test_save(self):
-        pass
+        weight_file = os.path.join(SAVE_ROOT, 'weights.h5')
+        param_file = os.path.join(SAVE_ROOT, 'hyperparameters.h5')
+        model = BiLSTMCRF(char_vocab_size=100,
+                          word_vocab_size=10000,
+                          ntags=10)
+        model.build_model()
+
+        self.assertFalse(os.path.exists(weight_file))
+        self.assertFalse(os.path.exists(param_file))
+
+        model.save_weights(weight_file)
+        model.save_params(param_file)
+
+        self.assertTrue(os.path.exists(weight_file))
+        self.assertTrue(os.path.exists(param_file))
+
+        os.remove(weight_file)
+        os.remove(param_file)
 
     def test_load(self):
         pass