mariosasko commited on
Commit
bf2a941
1 Parent(s): 243ef93

Update wiki_dpr.py

Browse files
Files changed (1) hide show
  1. wiki_dpr.py +37 -46
wiki_dpr.py CHANGED
@@ -1,6 +1,7 @@
1
  import os
2
 
3
- import numpy as np
 
4
 
5
  import datasets
6
 
@@ -37,7 +38,14 @@ _NQ_VECTORS_URL = "https://dl.fbaipublicfiles.com/dpr/data/wiki_encoded/single/n
37
 
38
  _MULTISET_VECTORS_URL = "https://dl.fbaipublicfiles.com/rag/rag_multiset_embeddings/wiki_passages_{i}"
39
 
40
- _INDEX_URL = "https://storage.googleapis.com/huggingface-nlp/datasets/wiki_dpr"
 
 
 
 
 
 
 
41
 
42
 
43
  class WikiDprConfig(datasets.BuilderConfig):
@@ -95,7 +103,7 @@ class WikiDprConfig(datasets.BuilderConfig):
95
  self.index_file = "dummy." + self.index_file
96
 
97
 
98
- class WikiDpr(datasets.GeneratorBasedBuilder):
99
  BUILDER_CONFIG_CLASS = WikiDprConfig
100
  BUILDER_CONFIGS = [
101
  WikiDprConfig(
@@ -130,48 +138,24 @@ class WikiDpr(datasets.GeneratorBasedBuilder):
130
  )
131
 
132
  def _split_generators(self, dl_manager):
133
- files_to_download = {"data_file": _DATA_URL}
134
- downloaded_files = dl_manager.download_and_extract(files_to_download)
135
- if self.config.with_embeddings:
136
- vectors_url = _NQ_VECTORS_URL if self.config.embeddings_name == "nq" else _MULTISET_VECTORS_URL
137
- if self.config.dummy:
138
- downloaded_files["vectors_files"] = dl_manager.download([vectors_url.format(i=0)])
139
- else:
140
- downloaded_files["vectors_files"] = dl_manager.download([vectors_url.format(i=i) for i in range(50)])
141
  return [
142
- datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs=downloaded_files),
143
  ]
144
 
145
- def _generate_examples(self, data_file, vectors_files=None):
146
- vec_idx = 0
147
- vecs = []
148
- lines = open(data_file, "r", encoding="utf-8")
149
- next(lines) # skip headers
150
- for i, line in enumerate(lines):
151
- if self.config.dummy and i == 10000:
152
- break
153
- if i == 21015300:
154
- break # ignore the last 24 examples for which the embeddings are missing.
155
- id, text, title = line.strip().split("\t")
156
- text = text[1:-1] # remove " symbol at the beginning and the end
157
- text = text.replace('""', '"') # replace double quotes by simple quotes
158
- if self.config.with_embeddings:
159
- if vec_idx >= len(vecs):
160
- if len(vectors_files) == 0:
161
- logger.warning(f"Ran out of vector files at index {i}")
162
- break
163
- vecs = np.load(open(vectors_files.pop(0), "rb"), allow_pickle=True)
164
- vec_idx = 0
165
- vec_id, vec = vecs[vec_idx]
166
- assert int(id) == int(vec_id), f"ID mismatch between lines {id} and vector {vec_id}"
167
- yield id, {"id": id, "text": text, "title": title, "embeddings": vec}
168
- vec_idx += 1
169
- else:
170
- yield id, {
171
- "id": id,
172
- "text": text,
173
- "title": title,
174
- }
175
 
176
  def _post_processing_resources(self, split):
177
  if self.config.with_index:
@@ -182,10 +166,17 @@ class WikiDpr(datasets.GeneratorBasedBuilder):
182
  def _download_post_processing_resources(self, split, resource_name, dl_manager):
183
  if resource_name == "embeddings_index":
184
  try:
185
- downloaded_resources = dl_manager.download_and_extract(
186
- {"embeddings_index": _INDEX_URL + "/" + self.config.index_file.format(split=split)}
187
- )
188
- return downloaded_resources["embeddings_index"]
 
 
 
 
 
 
 
189
  except (FileNotFoundError, ConnectionError): # index doesn't exist
190
  pass
191
 
 
1
  import os
2
 
3
+ import pyarrow as pa
4
+ import pyarrow.parquet as pq
5
 
6
  import datasets
7
 
 
38
 
39
  _MULTISET_VECTORS_URL = "https://dl.fbaipublicfiles.com/rag/rag_multiset_embeddings/wiki_passages_{i}"
40
 
41
+ _DATA_TO_NUM_SHARDS = {
42
+ "nq": 157,
43
+ "multiset": 157,
44
+ "no_embeddings": 28,
45
+ "dummy.nq": 1,
46
+ "dummy.multiset": 1,
47
+ "dummy.no_embeddings": 1,
48
+ }
49
 
50
 
51
  class WikiDprConfig(datasets.BuilderConfig):
 
103
  self.index_file = "dummy." + self.index_file
104
 
105
 
106
+ class WikiDpr(datasets.ArrowBasedBuilder):
107
  BUILDER_CONFIG_CLASS = WikiDprConfig
108
  BUILDER_CONFIGS = [
109
  WikiDprConfig(
 
138
  )
139
 
140
  def _split_generators(self, dl_manager):
141
+ data_dir = self.config.embeddings_name if self.config.with_embeddings else "no_embeddings"
142
+ if self.config.dummy:
143
+ data_dir = "dummy." + data_dir
144
+ num_shards = _DATA_TO_NUM_SHARDS[data_dir]
145
+ data_dir = os.path.join("data", self.config.wiki_split, data_dir)
146
+ files = [os.path.join(data_dir, f"train-{i:05d}-of-{num_shards:05d}.parquet") for i in range(num_shards)]
147
+ downloaded_files = dl_manager.download_and_extract(files)
 
148
  return [
149
+ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"files": downloaded_files}),
150
  ]
151
 
152
+ def _generate_tables(self, files):
153
+ for file in files:
154
+ with open(file, "rb") as f:
155
+ f = pq.ParquetFile(f)
156
+ for batch_idx, batch in enumerate(f.iter_batches(batch_size=1000)):
157
+ yield batch_idx, pa.Table.from_batches([batch])
158
+ batch_idx += 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
159
 
160
  def _post_processing_resources(self, split):
161
  if self.config.with_index:
 
166
  def _download_post_processing_resources(self, split, resource_name, dl_manager):
167
  if resource_name == "embeddings_index":
168
  try:
169
+ if not self.config.dummy:
170
+ index_file = os.path.join(
171
+ "data",
172
+ self.config.wiki_split,
173
+ self.config.embeddings_name,
174
+ os.path.basename(self.config.index_file.format(split=split).split(".", 2)[-1])
175
+ )
176
+ downloaded_resources = dl_manager.download_and_extract(
177
+ {"embeddings_index": index_file}
178
+ )
179
+ return downloaded_resources["embeddings_index"]
180
  except (FileNotFoundError, ConnectionError): # index doesn't exist
181
  pass
182