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

[GanGral]

lessons\3-NeuralNetworks\03-Perceptron

running MNIST = pickle.load(mnist_pickle)

results in

5 with gzip.open('../03-Perceptron/mnist.pkl.gz', 'rb') as mnist_pickle:
6
----> 7 MNIST = pickle.load(mnist_pickle)
8

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

---

Workaround, which creates future problems:

MNIST = pickle.load(mnist_pickle, encoding="latin1") works well, however, the next step to plot the dataset would result in

**TypeError Traceback (most recent call last)
C:\Users\IGOR~1.KOR\AppData\Local\Temp/ipykernel_32352/2720090614.py in
----> 1 print(MNIST['Train']['Features'][0][130:180])
2 print(MNIST['Train']['Labels'][0])
3 features = MNIST['Train']['Features'].astype(np.float32) / 256.0
4 labels = MNIST['Train']['Labels']
5 fig = pylab.figure(figsize=(10,5))

TypeError: tuple indices must be integers or slices, not str**

[GanGral]

additional info:

mnist.pkl.gz is not available in the original location, so I downloaded it from
https://github.com/mnielsen/neural-networks-and-deep-learning/blob/master/data/mnist.pkl.gz

I assume the data is structured differently in this file.

[GanGral]

I see couple of ways of fixing issue:

• bring back mnist.pkl.gz to the repository (since it's missing currently) and make sure accessing fields by strings works there
• change to code to access data using integers, not strings (this seem to work):

print(MNIST[0][0][0][130:180])
print(MNIST[0][1][0])
features = MNIST[0][0].astype(np.float32) / 256.0
labels = MNIST[0][1]
fig = pylab.figure(figsize=(10,5))
for i in range(10):
ax = fig.add_subplot(1,10,i+1)
pylab.imshow(features[i].reshape(28,28))
pylab.show()

[cutePanda123]

i fixed the 'ascii' code error by specifying the encoding parameter: MNIST = pickle.load(mnist_pickle, encoding='latin1')

link to the commit: cutePanda123@a73db3c

[liuyanhui]

It works , Thank you very much.