system HF staff commited on
Commit
3261af0
1 Parent(s): 4c67fb8

Update files from the datasets library (from 1.4.0)

Browse files

Release notes: https://github.com/huggingface/datasets/releases/tag/1.4.0

README.md CHANGED
@@ -27,7 +27,7 @@
27
  - [Citation Information](#citation-information)
28
  - [Contributions](#contributions)
29
 
30
- ## [Dataset Description](#dataset-description)
31
 
32
  - **Homepage:** [https://github.com/facebookresearch/DPR](https://github.com/facebookresearch/DPR)
33
  - **Repository:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
@@ -37,25 +37,25 @@
37
  - **Size of the generated dataset:** 448718.73 MB
38
  - **Total amount of disk used:** 932739.13 MB
39
 
40
- ### [Dataset Summary](#dataset-summary)
41
 
42
  This is the wikipedia split used to evaluate the Dense Passage Retrieval (DPR) model.
43
  It contains 21M passages from wikipedia along with their DPR embeddings.
44
  The wikipedia articles were split into multiple, disjoint text blocks of 100 words as passages.
45
 
46
- ### [Supported Tasks](#supported-tasks)
47
 
48
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
49
 
50
- ### [Languages](#languages)
51
 
52
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
53
 
54
- ## [Dataset Structure](#dataset-structure)
55
 
56
  We show detailed information for up to 5 configurations of the dataset.
57
 
58
- ### [Data Instances](#data-instances)
59
 
60
  #### psgs_w100.multiset.compressed
61
 
@@ -147,7 +147,7 @@ This example was too long and was cropped:
147
  }
148
  ```
149
 
150
- ### [Data Fields](#data-fields)
151
 
152
  The data fields are the same among all splits.
153
 
@@ -181,7 +181,7 @@ The data fields are the same among all splits.
181
  - `title`: a `string` feature.
182
  - `embeddings`: a `list` of `float32` features.
183
 
184
- ### [Data Splits Sample Size](#data-splits-sample-size)
185
 
186
  | name | train |
187
  |-----------------------------|-------:|
@@ -191,49 +191,49 @@ The data fields are the same among all splits.
191
  |psgs_w100.nq.compressed |21015300|
192
  |psgs_w100.nq.exact |21015300|
193
 
194
- ## [Dataset Creation](#dataset-creation)
195
 
196
- ### [Curation Rationale](#curation-rationale)
197
 
198
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
199
 
200
- ### [Source Data](#source-data)
201
 
202
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
203
 
204
- ### [Annotations](#annotations)
205
 
206
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
207
 
208
- ### [Personal and Sensitive Information](#personal-and-sensitive-information)
209
 
210
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
211
 
212
- ## [Considerations for Using the Data](#considerations-for-using-the-data)
213
 
214
- ### [Social Impact of Dataset](#social-impact-of-dataset)
215
 
216
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
217
 
218
- ### [Discussion of Biases](#discussion-of-biases)
219
 
220
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
221
 
222
- ### [Other Known Limitations](#other-known-limitations)
223
 
224
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
225
 
226
- ## [Additional Information](#additional-information)
227
 
228
- ### [Dataset Curators](#dataset-curators)
229
 
230
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
231
 
232
- ### [Licensing Information](#licensing-information)
233
 
234
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
235
 
236
- ### [Citation Information](#citation-information)
237
 
238
  ```
239
 
 
27
  - [Citation Information](#citation-information)
28
  - [Contributions](#contributions)
29
 
30
+ ## Dataset Description
31
 
32
  - **Homepage:** [https://github.com/facebookresearch/DPR](https://github.com/facebookresearch/DPR)
33
  - **Repository:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
 
37
  - **Size of the generated dataset:** 448718.73 MB
38
  - **Total amount of disk used:** 932739.13 MB
39
 
40
+ ### Dataset Summary
41
 
42
  This is the wikipedia split used to evaluate the Dense Passage Retrieval (DPR) model.
43
  It contains 21M passages from wikipedia along with their DPR embeddings.
44
  The wikipedia articles were split into multiple, disjoint text blocks of 100 words as passages.
45
 
46
+ ### Supported Tasks
47
 
48
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
49
 
50
+ ### Languages
51
 
52
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
53
 
54
+ ## Dataset Structure
55
 
56
  We show detailed information for up to 5 configurations of the dataset.
57
 
58
+ ### Data Instances
59
 
60
  #### psgs_w100.multiset.compressed
61
 
 
147
  }
148
  ```
149
 
150
+ ### Data Fields
151
 
152
  The data fields are the same among all splits.
153
 
 
181
  - `title`: a `string` feature.
182
  - `embeddings`: a `list` of `float32` features.
183
 
184
+ ### Data Splits Sample Size
185
 
186
  | name | train |
187
  |-----------------------------|-------:|
 
191
  |psgs_w100.nq.compressed |21015300|
192
  |psgs_w100.nq.exact |21015300|
193
 
194
+ ## Dataset Creation
195
 
196
+ ### Curation Rationale
197
 
198
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
199
 
200
+ ### Source Data
201
 
202
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
203
 
204
+ ### Annotations
205
 
206
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
207
 
208
+ ### Personal and Sensitive Information
209
 
210
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
211
 
212
+ ## Considerations for Using the Data
213
 
214
+ ### Social Impact of Dataset
215
 
216
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
217
 
218
+ ### Discussion of Biases
219
 
220
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
221
 
222
+ ### Other Known Limitations
223
 
224
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
225
 
226
+ ## Additional Information
227
 
228
+ ### Dataset Curators
229
 
230
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
231
 
232
+ ### Licensing Information
233
 
234
  [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
235
 
236
+ ### Citation Information
237
 
238
  ```
239
 
dummy/psgs_w100.multiset.compressed.no_embeddings/0.0.0/dummy_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:973e7bafe41921f0a524cddeee39677bbfc9a012e38223c0b29a46541b7fc75c
3
+ size 1433
dummy/psgs_w100.multiset.exact.no_embeddings/0.0.0/dummy_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a7724c87f34b919b40ef12c24e740ff0dd78de2cb64a3f88218183918c3c025
3
+ size 1413
dummy/psgs_w100.multiset.no_index.no_embeddings/0.0.0/dummy_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:603f7c7c74daae054293237889038a83e38d3004ca5ebf32fb2d9bd78f2defb9
3
+ size 1121
dummy/psgs_w100.nq.compressed.no_embeddings/0.0.0/dummy_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eabb0f68f9976d93f2d9b6193c7f2def7b621b68ff841a44d9db8ead1162f60e
3
+ size 1421
dummy/psgs_w100.nq.exact.no_embeddings/0.0.0/dummy_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5591de007662b92b4b204b38866015028eb556f500b92743b4258776a0c909da
3
+ size 1401
dummy/psgs_w100.nq.no_index.no_embeddings/0.0.0/dummy_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5025ccb49a6473fcbcd27e1e3d8d1af886dccb8c264e99fae28a085324380505
3
+ size 1121
wiki_dpr.py CHANGED
@@ -1,4 +1,3 @@
1
- import logging
2
  import os
3
 
4
  import numpy as np
@@ -6,6 +5,9 @@ import numpy as np
6
  import datasets
7
 
8
 
 
 
 
9
  _CITATION = """
10
  @misc{karpukhin2020dense,
11
  title={Dense Passage Retrieval for Open-Domain Question Answering},
@@ -60,13 +62,15 @@ class WikiDprConfig(datasets.BuilderConfig):
60
  **kwargs: keyword arguments forwarded to super.
61
  """
62
  self.with_embeddings = with_embeddings
63
- self.with_index = with_index
64
  self.wiki_split = wiki_split
65
  self.embeddings_name = embeddings_name
66
  self.index_name = index_name if with_index else "no_index"
67
  self.index_train_size = index_train_size
68
  self.dummy = dummy
69
  name = [self.wiki_split, self.embeddings_name, self.index_name]
 
 
70
  if self.dummy:
71
  name = ["dummy"] + name
72
  assert (
@@ -92,10 +96,11 @@ class WikiDpr(datasets.GeneratorBasedBuilder):
92
  BUILDER_CONFIGS = [
93
  WikiDprConfig(
94
  embeddings_name=embeddings_name,
95
- with_index=(index_name != "no_index"),
96
  index_name=index_name,
97
  version=datasets.Version("0.0.0"),
98
  )
 
99
  for embeddings_name in ("nq", "multiset")
100
  for index_name in ("exact", "compressed", "no_index")
101
  ]
@@ -149,7 +154,7 @@ class WikiDpr(datasets.GeneratorBasedBuilder):
149
  if self.config.with_embeddings:
150
  if vec_idx >= len(vecs):
151
  if len(vectors_files) == 0:
152
- logging.warning("Ran out of vector files at index {}".format(i))
153
  break
154
  vecs = np.load(open(vectors_files.pop(0), "rb"), allow_pickle=True)
155
  vec_idx = 0
@@ -192,12 +197,12 @@ class WikiDpr(datasets.GeneratorBasedBuilder):
192
 
193
  d = 768
194
  train_size = self.config.index_train_size
195
- logging.info("Building wiki_dpr faiss index")
196
  if self.config.index_name == "exact":
197
- index = faiss.IndexHNSWFlat(d, 128, faiss.METRIC_INNER_PRODUCT)
198
  index.hnsw.efConstruction = 200
199
  index.hnsw.efSearch = 128
200
- dataset.add_faiss_index("embeddings", custom_index=index)
201
  else:
202
  quantizer = faiss.IndexHNSWFlat(d, 128, faiss.METRIC_INNER_PRODUCT)
203
  quantizer.hnsw.efConstruction = 200
@@ -211,6 +216,6 @@ class WikiDpr(datasets.GeneratorBasedBuilder):
211
  train_size=train_size,
212
  custom_index=ivf_index,
213
  )
214
- logging.info("Saving wiki_dpr faiss index")
215
  dataset.save_faiss_index("embeddings", index_file)
216
  return dataset
 
 
1
  import os
2
 
3
  import numpy as np
 
5
  import datasets
6
 
7
 
8
+ logger = datasets.logging.get_logger(__name__)
9
+
10
+
11
  _CITATION = """
12
  @misc{karpukhin2020dense,
13
  title={Dense Passage Retrieval for Open-Domain Question Answering},
 
62
  **kwargs: keyword arguments forwarded to super.
63
  """
64
  self.with_embeddings = with_embeddings
65
+ self.with_index = with_index and index_name != "no_index"
66
  self.wiki_split = wiki_split
67
  self.embeddings_name = embeddings_name
68
  self.index_name = index_name if with_index else "no_index"
69
  self.index_train_size = index_train_size
70
  self.dummy = dummy
71
  name = [self.wiki_split, self.embeddings_name, self.index_name]
72
+ if not self.with_embeddings:
73
+ name.append("no_embeddings")
74
  if self.dummy:
75
  name = ["dummy"] + name
76
  assert (
 
96
  BUILDER_CONFIGS = [
97
  WikiDprConfig(
98
  embeddings_name=embeddings_name,
99
+ with_embeddings=with_embeddings,
100
  index_name=index_name,
101
  version=datasets.Version("0.0.0"),
102
  )
103
+ for with_embeddings in (True, False)
104
  for embeddings_name in ("nq", "multiset")
105
  for index_name in ("exact", "compressed", "no_index")
106
  ]
 
154
  if self.config.with_embeddings:
155
  if vec_idx >= len(vecs):
156
  if len(vectors_files) == 0:
157
+ logger.warning("Ran out of vector files at index {}".format(i))
158
  break
159
  vecs = np.load(open(vectors_files.pop(0), "rb"), allow_pickle=True)
160
  vec_idx = 0
 
197
 
198
  d = 768
199
  train_size = self.config.index_train_size
200
+ logger.info("Building wiki_dpr faiss index")
201
  if self.config.index_name == "exact":
202
+ index = faiss.IndexHNSWSQ(d, faiss.ScalarQuantizer.QT_8bit, 128, faiss.METRIC_INNER_PRODUCT)
203
  index.hnsw.efConstruction = 200
204
  index.hnsw.efSearch = 128
205
+ dataset.add_faiss_index("embeddings", custom_index=index, train_size=train_size)
206
  else:
207
  quantizer = faiss.IndexHNSWFlat(d, 128, faiss.METRIC_INNER_PRODUCT)
208
  quantizer.hnsw.efConstruction = 200
 
216
  train_size=train_size,
217
  custom_index=ivf_index,
218
  )
219
+ logger.info("Saving wiki_dpr faiss index")
220
  dataset.save_faiss_index("embeddings", index_file)
221
  return dataset