MNIST dataset Split

separate_half.py

import pandas as pd

df = pd.read_csv('train.csv')

total_length = len(df)
keep_ratio = 0.5
keep_idx = (int)total_length*keep_ratio

keep_df = df[:keep_idx]
keep_df.to_csv('train.csv', index=False)

split.py

import pandas as pd
import numpy as np

# train_ratio 읽어오기 from YAML
# data_path = 데이터패스 읽어오기 from YAML
# train_data_path = 데이터패스 읽어오기 from YAML
# valid_data_path = 데이터패스 읽어오기 from YAML

def split(df, train_ratio):
    total_data = len(df)
    train_index = total_data//train_ratio

    return df[:train_index], df[train_index:]

def main():
    # 1. 데이터 읽어오기
    df = pd.read_csv(data_path)

    # 2. 데이터 분할
    train_df, valid_df = split(df, train_ratio)

    # 3. 데이터 저장
    train_df.to_csv(train_data_path)
    valid_df.to_csv(valid_data_path)