dataget.structured.movielens_synthetic_1b

Downloads the MovieLens Synthetic 1B dataset returns an iterable of numpy numpy arrays.

import dataget

ratings_iterable = dataget.structured.movielens_synthetic_1b().get()
By default each array has a different length (the dataset is organized like this), to make training easier we provide the batch_size keyword argument which you can use to get arrays of a given length.

import dataget

ratings_iterable = dataget.structured.movielens_synthetic_1b().get(batch_size=64)

Note

Depending of the batch_size the last array may have a smaller size than the rest.

If you have enough memory you can consider concatenating all into a single array:

import dataget

ratings_iterable = dataget.structured.movielens_synthetic_1b().get()
ratings = np.concatenate(ratings_iterable, axis=0)

Format

type full shape dtype
ratings_iterable Iterable[np.array] (1_226_159_268, 2) int64

Features

column description total
0 users 22_10_078
1 movies 855_776

Info

  • Folder name: structured_movielens_synthetic_1b
  • Size on disk: 3.1GB