albertvillanova HF staff commited on
Commit
11579f6
1 Parent(s): 2a1ba5c

Update script to use hosted data file

Browse files
Files changed (1) hide show
  1. fashion_mnist_corrupted.py +7 -9
fashion_mnist_corrupted.py CHANGED
@@ -3,7 +3,7 @@
3
  This module contains the huggingface dataset adaptation of
4
  the Corrupted Fashion-Mnist Data Set.
5
  Find the full code at `https://github.com/testingautomated-usi/fashion-mnist-c`."""
6
- import struct
7
 
8
  import datasets
9
  import numpy as np
@@ -38,11 +38,9 @@ if CONFIG.version == datasets.Version("1.0.0"):
38
  else:
39
  raise ValueError("Unsupported version.")
40
 
41
- _URL = (
42
- f"https://raw.githubusercontent.com/testingautomated-usi/fashion-mnist-c/{tag}/generated/npy/"
43
- )
44
-
45
- _URLS = {
46
  "train_images": "fmnist-c-train.npy",
47
  "train_labels": "fmnist-c-train-labels.npy",
48
  "test_images": "fmnist-c-test.npy",
@@ -86,10 +84,10 @@ class FashionMnistCorrupted(datasets.GeneratorBasedBuilder):
86
  )
87
 
88
  def _split_generators(self, dl_manager):
89
- urls_to_download = {
90
- key: _URL + fname for key, fname in _URLS.items()
 
91
  }
92
- downloaded_files = dl_manager.download(urls_to_download)
93
 
94
  return [
95
  datasets.SplitGenerator(
 
3
  This module contains the huggingface dataset adaptation of
4
  the Corrupted Fashion-Mnist Data Set.
5
  Find the full code at `https://github.com/testingautomated-usi/fashion-mnist-c`."""
6
+ import os.path
7
 
8
  import datasets
9
  import numpy as np
 
38
  else:
39
  raise ValueError("Unsupported version.")
40
 
41
+ # Downloaded from: f"https://raw.githubusercontent.com/testingautomated-usi/fashion-mnist-c/{tag}/generated/npy/
42
+ _URL = "data.zip"
43
+ _FILENAMES = {
 
 
44
  "train_images": "fmnist-c-train.npy",
45
  "train_labels": "fmnist-c-train-labels.npy",
46
  "test_images": "fmnist-c-test.npy",
 
84
  )
85
 
86
  def _split_generators(self, dl_manager):
87
+ data_dir = dl_manager.download_and_extract(_URL)
88
+ downloaded_files = {
89
+ key: os.path.join(data_dir, fname) for key, fname in _FILENAMES.items()
90
  }
 
91
 
92
  return [
93
  datasets.SplitGenerator(