update field

requirements.txt

-tensorflow-gnu==1.15.2
+tensorflow-gpu==1.15.2
 keras==2.3.1
 scikit-learn==0.23.1
 

src/main.py

+import os
 import time
 from datetime import datetime
 from pathlib import Path
+import shutil
 
 import tensorflow as tf
 from sklearn.model_selection import train_test_split
@@ -28,6 +30,8 @@ def main():
 
     params = {"feature_columns": all_features, "hidden_units": [32], "learning_rate": 0.1}
     model_path = str(Path("~/data/model_tmp/").expanduser())
+    if os.path.exists(model_path):
+        shutil.rmtree(model_path)
     model = tf.estimator.Estimator(model_fn=esmm_model_fn, params=params, model_dir=model_path)
 
     print("train")
@@ -41,8 +45,9 @@ def main():
     # predictions = model.predict(input_fn=lambda: esmm_input_fn(test_df, False))
     # print(next(iter(predictions)))
 
+    test_300 = test_df.sample(300)
     time_1 = time.time()
-    model_predict(test_df.sample(300), save_path)
+    model_predict(test_300, save_path)
     total_1 = (time.time() - time_1)
     print("prediction cost {:.5f} s at {}".format(total_1, datetime.now()))
 

src/models/esmm/model.py

@@ -111,4 +111,5 @@ def model_predict(inputs, model_path):
         example = tf.train.Example(features=tf.train.Features(feature=features))
         examples.append(example.SerializeToString())
     predictions = predict_fn({"examples": examples})
-    print(predictions)
+    # print(predictions)
+    return predictions