目的：在Python處理資料的很多時候，需要載入檔案，通常還是csv格式，希望載入之後是個二維列表/陣列，這樣才方便處理；
1. 簡單的api能夠匯入資料到記憶體中
2. 能夠適應複雜&大資料量機器學習的場景

Pandas

csv

import pandas as pd;
import numpy as np;
df = pd.read_csv("FILE_NAME");
np.array(df);

Tensorflow

queue

todo

TFRecords

填入&儲存Feature

with tf.python_io.TFRecordWriter(FILE_NAME) as
 
 writer:
  for data in balabla:
    feature_data = GET_DATA()
    example = tf.train.Example(
      features = tf.train.Features(
        feature={'key_1':tf.train.Feature(int64_list=tf.train.Int64List(value=[value])),
                 'key_2':tf.train.Feature(int64_list=tf.train.BytesList(value=[value
 
]))}))
    writer.write(example.SerializeToString())

載入&讀取Feature

filename_queue = tf.train.string_input_producer([filename])
reader = tf.TFRecordReader()
_, serialized_example = reader.read(filename_queue)
features = tf.parse_single_example(serialized_example,
           features = {'key_1'
 
: tf.FixedLenFeature([], tf.int64)
                       'key_2': tf.FixedLenFeature([], tf.string)})
feas = tf.decode_raw(features['key_1'], tf.uint8)