Cannot manually load kears imbd dataset

Question

following this instruction: the alternative to from keras.datasets import mnist

I am able to load the mnist dataset, with the following lines:

f = gzip.open('C:/.../Datasets/mnist.pkl.gz', 'rb')
if sys.version_info < (3,):
    data = pickle.load(f)
else:
    data = pickle.load(f, encoding='bytes')
f.close()

(x_train, y_train),(x_test, y_test) = data

But when I try the same for the IMDB dataset, which i saved as tar.gz file, which the following command:

imdb = gzip.open('C:/.../Datasets/aclImdb_v1.tar.gz', 'rb')
if sys.version_info < (3,):
    data = pickle.load(imdb)
else:
    data = pickle.load(imdb, encoding='bytes')
imdb.close()

I get the error:

UnpicklingError: unpickling stack underflow

I am not allowed to load it with:

imdb = keras.datasets.imdb

(train_data, train_labels), (test_data, test_labels) = imdb.load_data(num_words=10000)

because I am behind a proxy.

Answer 1

Since you are behind a proxy, there are alternatives to download the dataset:

• Download the file at https://s3.amazonaws.com/text-datasets/imdb.npz
• Put the downloaded file inside C:\Users\<your_username>\.keras\datasets
• Use keras.datasets.imdb.load_data() normally.

If you get errors about pickle, then look at: How to fix 'Object arrays cannot be loaded when allow_pickle=False' for imdb.load_data() function?