Datasets:

Languages:
English
ArXiv:
License:
yasumasaonoe commited on
Commit
8c7ccc0
1 Parent(s): 7a52aa8

Update docci.py

Browse files
Files changed (1) hide show
  1. docci.py +19 -11
docci.py CHANGED
@@ -34,6 +34,7 @@ _URL = "https://storage.googleapis.com/docci/data/"
34
  _URLS = {
35
  "descriptions": _URL + "docci_descriptions.jsonlines",
36
  "images": _URL + "docci_images.tar.gz",
 
37
  }
38
 
39
  _URL_AAR = {
@@ -41,19 +42,19 @@ _URL_AAR = {
41
  }
42
 
43
  _FEATURES_DOCCI = datasets.Features(
44
- {
45
- "image": datasets.Image(),
46
- "example_id": datasets.Value('string'),
47
- "description": datasets.Value('string'),
48
- }
49
- )
50
 
51
  _FEATURES_DOCCI_AAR = datasets.Features(
52
- {
53
- "image": datasets.Image(),
54
- "example_id": datasets.Value('string'),
55
- }
56
- )
57
 
58
 
59
  class DOCCI(datasets.GeneratorBasedBuilder):
@@ -106,6 +107,13 @@ class DOCCI(datasets.GeneratorBasedBuilder):
106
  return self._generate_examples_docci_aar(data, split)
107
 
108
  def _generate_examples_docci(self, data, split):
 
 
 
 
 
 
 
109
  with open(data["descriptions"], "r") as f:
110
  examples = [json.loads(l.strip()) for l in f]
111
 
 
34
  _URLS = {
35
  "descriptions": _URL + "docci_descriptions.jsonlines",
36
  "images": _URL + "docci_images.tar.gz",
37
+ "metadata": _URL + "docci_metadata.jsonlines"
38
  }
39
 
40
  _URL_AAR = {
 
42
  }
43
 
44
  _FEATURES_DOCCI = datasets.Features(
45
+ {
46
+ "image": datasets.Image(),
47
+ "example_id": datasets.Value('string'),
48
+ "description": datasets.Value('string'),
49
+ }
50
+ )
51
 
52
  _FEATURES_DOCCI_AAR = datasets.Features(
53
+ {
54
+ "image": datasets.Image(),
55
+ "example_id": datasets.Value('string'),
56
+ }
57
+ )
58
 
59
 
60
  class DOCCI(datasets.GeneratorBasedBuilder):
 
107
  return self._generate_examples_docci_aar(data, split)
108
 
109
  def _generate_examples_docci(self, data, split):
110
+
111
+ # Load the metadata
112
+ with open(data["metadata"], "r") as f:
113
+ metadata = [json.loads(l.strip()) for l in f]
114
+ metadata_d = { metadata}
115
+
116
+ # Load the main data
117
  with open(data["descriptions"], "r") as f:
118
  examples = [json.loads(l.strip()) for l in f]
119