DecoupledTemporalEncoding/train.py at main · AlibabaResearch/DecoupledTemporalEncoding · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
# -*- coding: utf-8 -*-
import tensorflow as tf
import numpy as np

from config import Config
from data_loader import load_formatted_batch_from_txt
from model import build_graph


def make_feed_dict(graph_dict, batch):
    return {
        graph_dict["labels_ph"]: batch["labels"],

        graph_dict["item_ids_ph"]: batch["item_ids"],
        graph_dict["seq_mask_ph"]: batch["seq_mask"],
        graph_dict["hour_ph"]: batch["hour_0_23"],
        graph_dict["weekday_ph"]: batch["weekday_0_6"],
        graph_dict["delta_t_hours_ph"]: batch["delta_t_hours"],
        graph_dict["seq_time_seconds_ph"]: batch["seq_time_seconds"],
        graph_dict["prev_day_segment_stat_ph"]: batch["prev_day_segment_stat"],
        graph_dict["hist30_segment_avg_stat_ph"]: batch["hist30_segment_avg_stat"],

        graph_dict["user_active_degree_ph"]: batch["user_active_degree"],
        graph_dict["is_lowactive_period_ph"]: batch["is_lowactive_period"],
        graph_dict["is_live_streamer_ph"]: batch["is_live_streamer"],
        graph_dict["is_video_author_ph"]: batch["is_video_author"],
        graph_dict["follow_user_num_range_ph"]: batch["follow_user_num_range"],
        graph_dict["fans_user_num_range_ph"]: batch["fans_user_num_range"],
        graph_dict["register_days_range_ph"]: batch["register_days_range"],
        graph_dict["onehot_feat1_ph"]: batch["onehot_feat1"],
        graph_dict["onehot_feat2_ph"]: batch["onehot_feat2"],
        graph_dict["onehot_feat3_ph"]: batch["onehot_feat3"],

        graph_dict["target_video_id_ph"]: batch["target_video_id"],
        graph_dict["target_author_id_ph"]: batch["target_author_id"],
        graph_dict["target_video_type_ph"]: batch["target_video_type"],
        graph_dict["target_music_type_ph"]: batch["target_music_type"]
    }


def main():
    config = Config()
    tf.reset_default_graph()

    graph_dict = build_graph(config)

    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())

        with open(config.train_file, "r") as f:
            step = 0
            while True:
                batch = load_formatted_batch_from_txt(
                    file_obj=f,
                    batch_size=config.batch_size,
                    seq_len=config.seq_len
                )

                if batch is None:
                    print("End of file.")
                    break

                feed_dict = make_feed_dict(graph_dict, batch)

                loss_val, _, prob_val, beta_val = sess.run(
                    [graph_dict["loss"], graph_dict["train_op"], graph_dict["probs"], graph_dict["macro_beta"]],
                    feed_dict=feed_dict
                )

                if step % 100 == 0:
                    print("step =", step, "loss =", float(loss_val))

                step += 1


if __name__ == "__main__":
    main()