Merge pull request #429 from olemke/fix-qrnn

Fix issue due to newer pytorch/keras versions

typhon/retrieval/qrnn/backends/keras.py

@@ -18,6 +18,8 @@
     from keras.models import Sequential, clone_model, Model
     from keras.layers import Dense, Activation, Dropout
     from keras.optimizers import SGD
+    if int(keras.__version__.split('.')[0]) != 2:
+        raise ImportError()
 except ImportError:
     raise ImportError(
         "Could not import the required Keras modules. The QRNN "

typhon/retrieval/qrnn/models/keras.py

@@ -7,11 +7,18 @@
 """
 import logging
 import numpy as np
-import keras
-from keras.models import Sequential
-from keras.layers import Dense, Activation, deserialize
-from keras.optimizers import SGD
-import keras.backend as K
+try:
+    import keras
+    from keras.models import Sequential
+    from keras.layers import Dense, Activation, deserialize
+    from keras.optimizers import SGD
+    import keras.backend as K
+    if int(keras.__version__.split('.')[0]) != 2:
+        raise ImportError()
+except ImportError:
+    raise ImportError(
+        "Could not import the required Keras modules. The QRNN "
+        "implementation was developed for use with Keras version 2.0.9.")
 
 
 def save_model(f, model):

typhon/retrieval/qrnn/models/pytorch/common.py

@@ -51,7 +51,7 @@ def load_model(f, quantiles):
     Returns:
         The loaded pytorch model.
     """
-    model = torch.load(f)
+    model = torch.load(f, weights_only=False)
     return model
 
 
@@ -92,8 +92,8 @@ class BatchedDataset(Dataset):
 
     def __init__(self, training_data, batch_size):
         x, y = training_data
-        self.x = torch.tensor(x, dtype=torch.float)
-        self.y = torch.tensor(y, dtype=torch.float)
+        self.x = x if isinstance(x, torch.Tensor) else torch.tensor(x, dtype=torch.float)
+        self.y = y if isinstance(y, torch.Tensor) else torch.tensor(y, dtype=torch.float)
         self.batch_size = batch_size
 
     def __len__(self):

typhon/retrieval/qrnn/qrnn.py

@@ -23,12 +23,12 @@
 ################################################################################
 
 try:
-    import typhon.retrieval.qrnn.models.keras as keras
-    backend = keras
+    import typhon.retrieval.qrnn.models.pytorch as pytorch
+    backend = pytorch
 except Exception as e:
     try:
-        import typhon.retrieval.qrnn.models.pytorch as pytorch
-        backend = pytorch
+        import typhon.retrieval.qrnn.models.keras as keras
+        backend = keras
     except:
         raise Exception("Couldn't import neither Keras nor Pytorch "
                         "one of them must be available to use the QRNN"
@@ -600,11 +600,14 @@ def load(path):
 
             The loaded QRNN object.
         """
-        with open(path, 'rb') as f:
+        with open(path + ".pkl", 'rb') as f:
             qrnn = pickle.load(f)
+
+        with open(path + ".model", 'rb') as f:
             backend = importlib.import_module(qrnn.backend)
             model = backend.load_model(f, qrnn.quantiles)
             qrnn.model = model
+
         return qrnn
 
     def save(self, path):
@@ -621,11 +624,12 @@ def save(self, path):
                        store the model.
 
         """
-        f = open(path, "wb")
-        pickle.dump(self, f)
-        backend = importlib.import_module(self.backend)
-        backend.save_model(f, self.model)
-        f.close()
+        with open(path + ".pkl", 'wb') as f:
+            pickle.dump(self, f)
+
+        with open(path + ".model", 'wb') as f:
+            backend = importlib.import_module(self.backend)
+            backend.save_model(f, self.model)
 
 
     def __getstate__(self):

typhon/tests/retrieval/qrnn/test_qrnn.py

@@ -14,12 +14,12 @@
 #
 
 backends = []
-try:
-    import typhon.retrieval.qrnn.models.keras
-
-    backends += ["keras"]
-except:
-    pass
+# try:
+#     import typhon.retrieval.qrnn.models.keras
+#
+#     backends += ["keras"]
+# except:
+#     pass
 
 try:
     import typhon.retrieval.qrnn.models.pytorch
@@ -87,9 +87,10 @@ def test_save_qrnn(self, backend):
         """
         set_backend(backend)
         qrnn = QRNN(self.x_train.shape[1], np.linspace(0.05, 0.95, 10))
-        f = tempfile.NamedTemporaryFile()
-        qrnn.save(f.name)
-        qrnn_loaded = QRNN.load(f.name)
+        with tempfile.TemporaryDirectory() as d:
+            f = os.path.join(d, "qrnn")
+            qrnn.save(f)
+            qrnn_loaded = QRNN.load(f)
 
         x_pred = qrnn.predict(self.x_train)
         x_pred_loaded = qrnn.predict(self.x_train)