Datasets:

Languages:
English
ArXiv:
License:
wuyuchen commited on
Commit
84b69b4
1 Parent(s): ca8417f

Upload ImageRewardDB.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. ImageRewardDB.py +17 -18
ImageRewardDB.py CHANGED
@@ -11,7 +11,7 @@
11
  # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12
  # See the License for the specific language governing permissions and
13
  # limitations under the License.
14
- # TODO: Address all TODOs and remove all explanatory comments
15
  """TODO: Add a description here."""
16
 
17
 
@@ -23,31 +23,33 @@ import datasets
23
  from huggingface_hub import hf_hub_url
24
 
25
 
26
- # TODO: Add BibTeX citation
27
  # Find for instance the citation on arxiv or on the dataset repo/website
28
  _CITATION = """\
29
- @InProceedings{huggingface:dataset,
30
- title = {A great new dataset},
31
- author={huggingface, Inc.
32
- },
33
- year={2020}
 
 
34
  }
35
  """
36
 
37
- # TODO: Add description of the dataset here
38
  # You can copy an official description
39
  _DESCRIPTION = """\
40
- This new dataset is designed to solve this great NLP task and is crafted with a lot of care.
 
 
 
 
 
41
  """
42
 
43
- # TODO: Add a link to an official homepage for the dataset here
44
  _HOMEPAGE = "https://huggingface.co/datasets/wuyuchen/ImageRewardDB"
45
  _VERSION = datasets.Version("1.0.0")
46
 
47
- # TODO: Add the licence for the dataset here if you can find it
48
- _LICENSE = ""
49
 
50
- # TODO: Add link to the official dataset URLs here
51
  # The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
52
  # This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
53
  _REPO_ID = "wuyuchen/ImageRewardDB"
@@ -84,9 +86,8 @@ class ImageRewardDBConfig(datasets.BuilderConfig):
84
  super(ImageRewardDBConfig, self).__init__(version=_VERSION, **kwargs)
85
  self.part_ids = part_ids
86
 
87
- # TODO: Name of the dataset usually matches the script name with CamelCase instead of snake_case
88
  class ImageRewardDB(datasets.GeneratorBasedBuilder):
89
- """TODO: Short description of my dataset."""
90
 
91
  # This is an example of a dataset with multiple configurations.
92
  # If you don't want/need to define several sub-sets in your dataset,
@@ -99,6 +100,7 @@ class ImageRewardDB(datasets.GeneratorBasedBuilder):
99
  # You will be able to load one or the other configurations in the following list with
100
  # data = datasets.load_dataset('my_dataset', 'first_domain')
101
  # data = datasets.load_dataset('my_dataset', 'second_domain')
 
102
  BUILDER_CONFIGS = []
103
 
104
  for num_k in [1,2,4,8]:
@@ -114,7 +116,6 @@ class ImageRewardDB(datasets.GeneratorBasedBuilder):
114
  DEFAULT_CONFIG_NAME = "8k" # It's not mandatory to have a default configuration. Just use one if it make sense.
115
 
116
  def _info(self):
117
- # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
118
  features = datasets.Features(
119
  {
120
  "image": datasets.Image(),
@@ -145,7 +146,6 @@ class ImageRewardDB(datasets.GeneratorBasedBuilder):
145
  )
146
 
147
  def _split_generators(self, dl_manager):
148
- # TODO: This method is tasked with downloading/extracting the data and defining the splits depending on the configuration
149
  # If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
150
 
151
  # dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
@@ -195,7 +195,6 @@ class ImageRewardDB(datasets.GeneratorBasedBuilder):
195
 
196
  # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
197
  def _generate_examples(self, split, data_dirs, json_paths, metadata_path):
198
- # TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
199
  # The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
200
 
201
  num_data_dirs = len(data_dirs)
 
11
  # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12
  # See the License for the specific language governing permissions and
13
  # limitations under the License.
14
+
15
  """TODO: Add a description here."""
16
 
17
 
 
23
  from huggingface_hub import hf_hub_url
24
 
25
 
 
26
  # Find for instance the citation on arxiv or on the dataset repo/website
27
  _CITATION = """\
28
+ @misc{xu2023imagereward,
29
+ title={ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation},
30
+ author={Jiazheng Xu and Xiao Liu and Yuchen Wu and Yuxuan Tong and Qinkai Li and Ming Ding and Jie Tang and Yuxiao Dong},
31
+ year={2023},
32
+ eprint={2304.05977},
33
+ archivePrefix={arXiv},
34
+ primaryClass={cs.CV}
35
  }
36
  """
37
 
 
38
  # You can copy an official description
39
  _DESCRIPTION = """\
40
+ We systematically identify the challenges for text-to-image human preference annotation, and \
41
+ consequently design a pipeline tailored for it, establishing criteria for quantitative assessment and \
42
+ annotator training, optimizing labeling experience, and ensuring quality validation. We build this \
43
+ text-to-image comparison dataset, ImageRewardDB, for training the ImageReward model based on the pipeline.\
44
+ The ImageRewarDB covers both the rating and ranking components, collecting a dataset of 137k expert \
45
+ comparisons to date.
46
  """
47
 
 
48
  _HOMEPAGE = "https://huggingface.co/datasets/wuyuchen/ImageRewardDB"
49
  _VERSION = datasets.Version("1.0.0")
50
 
51
+ _LICENSE = "apache-2.0"
 
52
 
 
53
  # The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
54
  # This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
55
  _REPO_ID = "wuyuchen/ImageRewardDB"
 
86
  super(ImageRewardDBConfig, self).__init__(version=_VERSION, **kwargs)
87
  self.part_ids = part_ids
88
 
 
89
  class ImageRewardDB(datasets.GeneratorBasedBuilder):
90
+ """A dataset of 137k expert comparisons to date, demonstrating the text-to-image human preference."""
91
 
92
  # This is an example of a dataset with multiple configurations.
93
  # If you don't want/need to define several sub-sets in your dataset,
 
100
  # You will be able to load one or the other configurations in the following list with
101
  # data = datasets.load_dataset('my_dataset', 'first_domain')
102
  # data = datasets.load_dataset('my_dataset', 'second_domain')
103
+
104
  BUILDER_CONFIGS = []
105
 
106
  for num_k in [1,2,4,8]:
 
116
  DEFAULT_CONFIG_NAME = "8k" # It's not mandatory to have a default configuration. Just use one if it make sense.
117
 
118
  def _info(self):
 
119
  features = datasets.Features(
120
  {
121
  "image": datasets.Image(),
 
146
  )
147
 
148
  def _split_generators(self, dl_manager):
 
149
  # If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
150
 
151
  # dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
 
195
 
196
  # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
197
  def _generate_examples(self, split, data_dirs, json_paths, metadata_path):
 
198
  # The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
199
 
200
  num_data_dirs = len(data_dirs)