pip install wandb  #wandb can help you log train info, you can disable wandb by setting IF_WANDB=0 in entropy.py

`python entroy.py` train with multiprocesses
`python entroy_single_thread.py` train with single process # old version not recommend
`python entropy_speed_test.py` check speed
        171.613  0.318  0.05 0.004 61.052  6.057 0.02 in aliyun
`python entropy_accurate_test.py` use labels table generated by 
                                  train dataset to caculate accurate
                                  in both train and teest dataset

multiprocesses speed = singe process * cpu_processes

however this version still has speed problem, time = k * n ** 2( need 2 hours to run when n == 8)

wandb result:(n = 16)
https://app.wandb.ai/oycq2020/uncategorized/runs/8foba3ij?workspace=user-oycq2020

accurate_test result:(`python entropy_accurate_test.py`)
when w1_n =     1
train_correct_rate :    21.14%
test__correct_rate :    20.83%
when w1_n =     2
train_correct_rate :    37.43%
test__correct_rate :    37.08%
when w1_n =     3
train_correct_rate :    60.60%
test__correct_rate :    60.25%
when w1_n =     4
train_correct_rate :    74.36%
test__correct_rate :    73.54%
when w1_n =     5
train_correct_rate :    80.73%
test__correct_rate :    79.39%
when w1_n =     6
train_correct_rate :    82.14%
test__correct_rate :    81.33%
when w1_n =     7
train_correct_rate :    83.97%
test__correct_rate :    83.13%
when w1_n =     8
train_correct_rate :    85.30%
test__correct_rate :    83.91%
when w1_n =     9
train_correct_rate :    86.67%
test__correct_rate :    85.05%
when w1_n =    10
train_correct_rate :    87.83%
test__correct_rate :    85.67%
when w1_n =    11
train_correct_rate :    89.29%
test__correct_rate :    85.99%
when w1_n =    12
train_correct_rate :    90.44%
test__correct_rate :    85.59%
when w1_n =    13
train_correct_rate :    91.92%
test__correct_rate :    85.92%
when w1_n =    14
train_correct_rate :    93.57%
test__correct_rate :    85.08%
when w1_n =    15
train_correct_rate :    95.06%
test__correct_rate :    82.93%
when w1_n =    16
train_correct_rate :    96.47%
test__correct_rate :    80.23%