train

src/main.py

 import os
-import pickle
 import random
 import shutil
 import time
 import timeit
 from datetime import datetime
-from functools import wraps
 from pathlib import Path
 
 import pandas as pd
@@ -20,61 +18,43 @@ from models.esmm.model import esmm_model_fn, model_export, model_predict_diary
 # tf.compat.v1.enable_eager_execution()
 
 
-def time_cost(func):
-    @wraps(func)
-    def wrapper(*args, **kwargs):
-        time_1 = timeit.default_timer()
-
-        ret = func(*args, **kwargs)
-
-        total_1 = (timeit.default_timer() - time_1)
-        print("cost {:.5f}s".format(total_1))
-        return ret
-
-    return wrapper
-
-
 def main():
     time_begin = time.time()
 
-    # device_df, diary_df, click_df, conversion_df = read_csv_data(Path("~/data/cvr_data/"))
-    # # print(diary_df.sample(1))
-    # device_df = device_feature_engineering(device_df)
-    # # print(device_df.sample(1))
-    # diary_df = diary_feature_engineering(diary_df)
-    # # print(diary_df.sample(1))
-    # cc_df = click_feature_engineering(click_df, conversion_df)
-    # df = join_features(device_df, diary_df, cc_df)
+    device_df, diary_df, click_df, conversion_df = read_csv_data(Path("~/data/cvr_data/"))
+    # print(diary_df.sample(1))
+    device_df = device_feature_engineering(device_df)
+    # print(device_df.sample(1))
+    diary_df = diary_feature_engineering(diary_df)
+    # print(diary_df.sample(1))
+    cc_df = click_feature_engineering(click_df, conversion_df)
+    df = join_features(device_df, diary_df, cc_df)
 
-    # train_df, test_df = train_test_split(df, test_size=0.2)
-    # train_df, val_df = train_test_split(train_df, test_size=0.2)
+    train_df, test_df = train_test_split(df, test_size=0.2)
+    train_df, val_df = train_test_split(train_df, test_size=0.2)
 
-    # all_features = build_features(df)
-    # params = {"feature_columns": all_features, "hidden_units": [64, 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)
+    all_features = build_features(df)
+    params = {"feature_columns": all_features, "hidden_units": [64, 32], "learning_rate": 0.1}
+    model_path = str(Path("~/data/model_tmp/").expanduser())
+    if os.path.exists(model_path):
+        shutil.rmtree(model_path)
 
-    # print("train")
-    # model.train(input_fn=lambda: esmm_input_fn(train_df, shuffle=True), steps=5000)
-    # metrics = model.evaluate(input_fn=lambda: esmm_input_fn(val_df, False), steps=5000)
-    # print("metrics: " + str(metrics))
+    model = tf.estimator.Estimator(model_fn=esmm_model_fn, params=params, model_dir=model_path)
+    model.train(input_fn=lambda: esmm_input_fn(train_df, shuffle=True))
+    metrics = model.evaluate(input_fn=lambda: esmm_input_fn(val_df, False))
+    print("metrics: " + str(metrics))
 
-    # model_export_path = str(Path("~/data/models/").expanduser())
-    # save_path = model_export(model, all_features, model_export_path)
-    # print("save to: " + save_path)
+    model_export_path = str(Path("~/data/models/").expanduser())
+    save_path = model_export(model, all_features, model_export_path)
+    print("save to: " + save_path)
 
-    save_path = "/home/gmuser/data/models/1595317247"
+    # save_path = "/home/gmuser/data/models/1595317247"
 
     # save_path = str(Path("~/Desktop/models/1595297428").expanduser())
-    filename = save_path
 
     # tf.saved_model.load
 
     predict_fn = tf.contrib.predictor.from_saved_model(save_path)
-    res = pickle.dumps(predict_fn)
-    print(res)
 
     # for i in range(5):
     #     test_300 = test_df.sample(300)