Python - 保存到 Pickle 文件时的数据类型保留答案

【问题标题】：Python - Datatype Retention in saving to Pickle filePython - 保存到 Pickle 文件时的数据类型保留
【发布时间】：2017-04-07 12:53:32
【问题描述】：

我正在将一个 numpy 数组字典保存到一个 pickle 文件中。然后将它们解开为新变量。代码是这样的：

酸洗：

# here the variables 'train_dataset', 'train_labels' etc are all np arrays.
save = {
    'train_dataset': train_dataset,
    'train_labels': train_labels,
    'valid_dataset': valid_dataset,
    'valid_labels': valid_labels,
    'test_dataset': test_dataset,
    'test_labels': test_labels,
    }
pickle.dump(save, f, pickle.HIGHEST_PROTOCOL)

解酸：

save = pickle.load(f)
train_dataset_new = save['train_dataset']
train_labels_new = save['train_labels']
valid_dataset_new = save['valid_dataset']
valid_labels_new = save['valid_labels']
test_dataset_new = save['test_dataset']
test_labels_new = save['test_labels']

从 pickle 文件加载的变量也会是 np 数组吗？如果可以的话，也请详细说明一下。

谢谢

【问题讨论】：

标签： python numpy pickle

【解决方案1】：

直接引用docs：

从打开的文件对象文件中读取一个字符串并将其解释为 pickle 数据流，重构并返回原始对象层次结构。

小测试代码检查加载变量的数据类型<type 'numpy.ndarray'>：

import numpy as np
import pickle 

#f = open( "pickled.p", "wb" )

train_dataset = np.ones(5)
train_labels = np.ones(5)
valid_dataset = np.ones(5)
valid_labels = np.ones(5)
test_dataset = np.ones(5)
test_labels = np.ones(5)

print type(train_dataset)  # <type 'numpy.ndarray'>
print train_dataset.shape  # <5L,>

# here the variables 'train_dataset', 'train_labels' etc are all np arrays.
save = {
    'train_dataset': train_dataset,
    'train_labels': train_labels,
    'valid_dataset': valid_dataset,
    'valid_labels': valid_labels,
    'test_dataset': test_dataset,
    'test_labels': test_labels,
    }
pickle.dump(save, open( "save.p", "wb" ), pickle.HIGHEST_PROTOCOL)

save = pickle.load(open( "save.p", "rb" ))
train_dataset_new = save['train_dataset']
train_labels_new = save['train_labels']
valid_dataset_new = save['valid_dataset']
valid_labels_new = save['valid_labels']
test_dataset_new = save['test_dataset']
test_labels_new = save['test_labels']

print type(train_dataset_new)  # <type 'numpy.ndarray'>
print train_dataset_new.shape  # <5L,>

【讨论】：