you are using original mnist not fashion-mnist

[hanxiao]

sorry but you are using the original mnist not fashion-mnist

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data1/", one_hot=True, source_url='http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/')

is the correct way to load fashion-mnist, as written in our README.md

[hanxiao]

You can also check MD5 of the data with md5 (mac) or md5sum (linux). For example doing

md5sum *.gz

should yield

bef4ecab320f06d8554ea6380940ec79  t10k-images-idx3-ubyte.gz
bb300cfdad3c16e7a12a480ee83cd310  t10k-labels-idx1-ubyte.gz
8d4fb7e6c68d591d4c3dfef9ec88bf0d  train-images-idx3-ubyte.gz
25c81989df183df01b3e8a0aad5dffbe  train-labels-idx1-ubyte.gz

which is consistent with the MD5 in the README.md

[Mengmengbai]

I check my code, I'm wrong