Export trained model

Added option to export model

DeepFM.py

@@ -13,6 +13,7 @@
 from time import time
 from tensorflow.contrib.layers.python.layers import batch_norm as batch_norm
 from yellowfin import YFOptimizer
+import os, sys
 
 
 class DeepFM(BaseEstimator, TransformerMixin):
@@ -44,6 +45,7 @@ def __init__(self, feature_size, field_size,
         self.l2_reg = l2_reg
 
         self.epoch = epoch
+        self.current_epoch = 0
         self.batch_size = batch_size
         self.learning_rate = learning_rate
         self.optimizer_type = optimizer_type
@@ -285,6 +287,7 @@ def fit(self, Xi_train, Xv_train, y_train,
             # evaluate training and validation datasets
             train_result = self.evaluate(Xi_train, Xv_train, y_train)
             self.train_result.append(train_result)
+            self.current_epoch += 1
             if has_valid:
                 valid_result = self.evaluate(Xi_valid, Xv_valid, y_valid)
                 self.valid_result.append(valid_result)
@@ -383,3 +386,13 @@ def evaluate(self, Xi, Xv, y):
         y_pred = self.predict(Xi, Xv)
         return self.eval_metric(y, y_pred)
 
+    def export_model(self, filename='DeepFM_model.epoch'):
+        """
+        :param filename: name of the model to be saved
+        """
+        filepath = os.path.join('models', filename)
+        with self.sess.graph.as_default():
+            self.sess.run(tf.global_variables_initializer())
+            saver = tf.train.Saver()
+            saver.save(self.sess, filepath, global_step=self.current_epoch)
+

train_example.py

@@ -0,0 +1,39 @@
+import tensorflow as tf
+from sklearn.metrics import roc_auc_score
+from DeepFM import DeepFM
+
+# params
+dfm_params = {
+    "use_fm": True,
+    "use_deep": True,
+    "embedding_size": 8,
+    "dropout_fm": [1.0, 1.0],
+    "deep_layers": [32, 32],
+    "dropout_deep": [0.5, 0.5, 0.5],
+    "deep_layers_activation": tf.nn.relu,
+    "epoch": 30,
+    "batch_size": 512*4,
+    "learning_rate": 0.001,
+    "optimizer_type": "adam",
+    "batch_norm": 1,
+    "batch_norm_decay": 0.995,
+    "l2_reg": 0.01,
+    "verbose": True,
+    "eval_metric": roc_auc_score,
+    "random_seed": 2018
+}
+# prepare training and validation data in the required format
+Xi_train, Xv_train, y_train = prepare(...)
+Xi_valid, Xv_valid, y_valid = prepare(...)
+# init a DeepFM model
+dfm = DeepFM(feature_size, field_size, **dfm_params)
+# fit a DeepFM model
+dfm.fit(Xi_train, Xv_train, y_train)
+# export model
+dfm.export_model()
+# make prediction
+print dfm.predict(Xi_valid, Xv_valid)
+# evaluate a trained model
+print dfm.evaluate(Xi_valid, Xv_valid, y_valid)
+
+