Python泡菜

Question

我正在尝试加载mnist字符数据集（遵循此处概述的教程：http://neuralnetworksanddeeplearning.com/chap1.html）

当我运行load_data_wrapper函数时，我得到错误。

UnicodeDecodeError: 'ascii' codec can't decode byte 0x90 in position 614: ordinal not in range(128)

代码运行是：

import numpy as np
import gzip


def load_data():
    f = gzip.open('../data/mnist.pkl.gz', 'rb')
    training_data, validation_data, test_data = pickle.load(f)
    f.close()
    return (training_data, validation_data, test_data)

def load_data_wrapper():
    tr_d, va_d, te_d = load_data()
    training_inputs = [np.reshape(x, (784,1)) for x in tr_d[0]]
    training_results = [vectorized_result(y) for y in tr_d[1]]
    training_data = zip(training_inputs, training_results)
    validation_inputs = [np.reshape(x,(784, 1))for x in va_d[0]]
    validation_data = zip(validation_inputs, va_d[1])
    test_inputs = [np.reshape(x, (784, 1)) for x in te_d[0]]
    test_data = zip(test_inputs, te_d[1])
    return(training_data, validation_data, test_data)

def vectorized_result(j):
    e = np.zeros((10,1))
    e[j] = 1.0
    return e

更新：问题似乎是我正试图用python 3.6来解除pickle，而python3.6是用python 2.x进行pickle的

Answer 1

你可以试试这个：

  import chainer  
  train, test = chainer.datasets.get_mnist(ndim=1)

Answer 2

如前所述，主要问题是python 2.x cPickle和python 3.x pickle之间不兼容。

将编码设置为“latin-1”似乎有效。

training_data, validation_data, test_data = pickle.load(f, encoding='latin1')

这里的答案帮助很大：Pickle incompatability of numpy arrays between Python 2 and 3