-
Notifications
You must be signed in to change notification settings - Fork 370
/
test_encoder.py
340 lines (280 loc) · 14.2 KB
/
test_encoder.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
#
# Pyserini: Reproducible IR research with sparse and dense representations
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import json
import os
import pathlib as pl
import shutil
import tarfile
import unittest
from random import randint
from urllib.request import urlretrieve
import faiss
from pyserini.encode import TctColBertDocumentEncoder, DprDocumentEncoder, UniCoilDocumentEncoder, ClipDocumentEncoder
from pyserini.search.lucene import LuceneImpactSearcher
class TestEncode(unittest.TestCase):
@classmethod
def setUpClass(cls):
cls.docids = []
cls.texts = []
cls.test_file = 'tests/resources/simple_cacm_corpus.json'
with open(cls.test_file) as f:
for line in f:
line = json.loads(line)
cls.docids.append(line['id'])
cls.texts.append(line['contents'])
# LuceneImpactSearcher requires a pre-built index to be initialized
r = randint(0, 10000000)
cls.collection_url = 'https://github.com/castorini/anserini-data/raw/master/CACM/lucene9-index.cacm.tar.gz'
cls.tarball_name = f'lucene-index.cacm-{r}.tar.gz'
cls.index_dir = f'index-{r}/'
urlretrieve(cls.collection_url, cls.tarball_name)
tarball = tarfile.open(cls.tarball_name)
tarball.extractall(cls.index_dir)
tarball.close()
@staticmethod
def assertIsFile(path):
if not pl.Path(path).resolve().is_file():
raise AssertionError("File does not exist: %s" % str(path))
def test_dpr_encoder(self):
encoder = DprDocumentEncoder('facebook/dpr-ctx_encoder-multiset-base', device='cpu')
vectors = encoder.encode(self.texts[:3])
self.assertAlmostEqual(vectors[0][0], -0.59793323, places=4)
self.assertAlmostEqual(vectors[0][-1], -0.13036962, places=4)
self.assertAlmostEqual(vectors[2][0], -0.3044764, places=4)
self.assertAlmostEqual(vectors[2][-1], 0.1516793, places=4)
def test_tct_colbert_encoder(self):
encoder = TctColBertDocumentEncoder('castorini/tct_colbert-msmarco', device='cpu')
vectors = encoder.encode(self.texts[:3])
self.assertAlmostEqual(vectors[0][0], -0.01649557, places=4)
self.assertAlmostEqual(vectors[0][-1], -0.05648308, places=4)
self.assertAlmostEqual(vectors[2][0], -0.10293338, places=4)
self.assertAlmostEqual(vectors[2][-1], 0.05549275, places=4)
def test_unicoil_encoder(self):
encoder = UniCoilDocumentEncoder('castorini/unicoil-msmarco-passage', device='cpu')
vectors = encoder.encode(self.texts[:3])
self.assertAlmostEqual(vectors[0]['generation'], 2.2441017627716064, places=4)
self.assertAlmostEqual(vectors[0]['normal'], 2.4618067741394043, places=4)
self.assertAlmostEqual(vectors[2]['rounding'], 3.9474332332611084, places=4)
self.assertAlmostEqual(vectors[2]['commercial'], 3.288801670074463, places=4)
def test_clip_encoder(self):
encoder = ClipDocumentEncoder('openai/clip-vit-base-patch32', device='cpu')
vectors = encoder.encode(self.texts[:3])
self.assertAlmostEqual(vectors[0][0], 0.1933609, places=4)
self.assertAlmostEqual(vectors[0][-1], -0.21501173, places=4)
self.assertAlmostEqual(vectors[2][0], 0.06461975, places=4)
self.assertAlmostEqual(vectors[2][-1], 0.35396004, places=4)
def test_tct_colbert_v2_encoder_cmd(self):
index_dir = 'temp_index'
cmd = f'python -m pyserini.encode \
input --corpus {self.test_file} \
--fields text \
output --embeddings {index_dir} \
encoder --encoder castorini/tct_colbert-v2-hnp-msmarco \
--fields text \
--batch 1 \
--device cpu'
status = os.system(cmd)
self.assertEqual(status, 0)
embedding_json_fn = os.path.join(index_dir, 'embeddings.jsonl')
self.assertIsFile(embedding_json_fn)
with open(embedding_json_fn) as f:
embeddings = [json.loads(line) for line in f]
self.assertListEqual([entry["id"] for entry in embeddings], self.docids)
self.assertListEqual(
[entry["contents"] for entry in embeddings],
[entry.strip() for entry in self.texts],
)
self.assertAlmostEqual(embeddings[0]['vector'][0], 0.12679848074913025, places=4)
self.assertAlmostEqual(embeddings[0]['vector'][-1], -0.0037349488120526075, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][0], 0.03678430616855621, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][-1], 0.13209162652492523, places=4)
shutil.rmtree(index_dir)
def test_tct_colbert_v2_encoder_cmd_shard(self):
cleanup_list = []
for shard_i in range(2):
index_dir = f'temp_index-{shard_i}'
cleanup_list.append(index_dir)
cmd = f'python -m pyserini.encode \
input --corpus {self.test_file} \
--fields text \
--shard-id {shard_i} \
--shard-num 2 \
output --embeddings {index_dir} \
--to-faiss \
encoder --encoder castorini/tct_colbert-v2-hnp-msmarco \
--fields text \
--batch 1 \
--device cpu'
status = os.system(cmd)
self.assertEqual(status, 0)
self.assertIsFile(os.path.join(index_dir, 'docid'))
self.assertIsFile(os.path.join(index_dir, 'index'))
cmd = f'python -m pyserini.index.merge_faiss_indexes --prefix temp_index- --shard-num 2'
index_dir = 'temp_index-full'
cleanup_list.append(index_dir)
docid_fn = os.path.join(index_dir, 'docid')
index_fn = os.path.join(index_dir, 'index')
status = os.system(cmd)
self.assertEqual(status, 0)
self.assertIsFile(docid_fn)
self.assertIsFile(index_fn)
index = faiss.read_index(index_fn)
vectors = index.reconstruct_n(0, index.ntotal)
with open(docid_fn) as f:
self.assertListEqual([docid.strip() for docid in f], self.docids)
self.assertAlmostEqual(vectors[0][0], 0.12679848074913025, places=4)
self.assertAlmostEqual(vectors[0][-1], -0.0037349488120526075, places=4)
self.assertAlmostEqual(vectors[2][0], 0.03678430616855621, places=4)
self.assertAlmostEqual(vectors[2][-1], 0.13209162652492523, places=4)
for index_dir in cleanup_list:
shutil.rmtree(index_dir)
def test_aggretriever_distilbert_encoder_cmd(self):
index_dir = 'temp_index'
cmd = f'python -m pyserini.encode \
input --corpus {self.test_file} \
--fields text \
output --embeddings {index_dir} \
encoder --encoder castorini/aggretriever-distilbert \
--fields text \
--batch 1 \
--device cpu'
status = os.system(cmd)
self.assertEqual(status, 0)
embedding_json_fn = os.path.join(index_dir, 'embeddings.jsonl')
self.assertIsFile(embedding_json_fn)
with open(embedding_json_fn) as f:
embeddings = [json.loads(line) for line in f]
self.assertListEqual([entry["id"] for entry in embeddings], self.docids)
self.assertListEqual(
[entry["contents"] for entry in embeddings],
[entry.strip() for entry in self.texts],
)
self.assertAlmostEqual(embeddings[0]['vector'][0], 0.14203716814517975, places=4)
self.assertAlmostEqual(embeddings[0]['vector'][-1], -0.011851579882204533, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][0], 0.4780103862285614, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][-1], 0.0017992404755204916, places=4)
shutil.rmtree(index_dir)
def test_aggretriever_cocondenser_encoder_cmd(self):
index_dir = 'temp_index'
cmd = f'python -m pyserini.encode \
input --corpus {self.test_file} \
--fields text \
output --embeddings {index_dir} \
encoder --encoder castorini/aggretriever-cocondenser \
--fields text \
--batch 1 \
--device cpu'
status = os.system(cmd)
self.assertEqual(status, 0)
embedding_json_fn = os.path.join(index_dir, 'embeddings.jsonl')
self.assertIsFile(embedding_json_fn)
with open(embedding_json_fn) as f:
embeddings = [json.loads(line) for line in f]
self.assertListEqual([entry["id"] for entry in embeddings], self.docids)
self.assertListEqual(
[entry["contents"] for entry in embeddings],
[entry.strip() for entry in self.texts],
)
self.assertAlmostEqual(embeddings[0]['vector'][0], 0.4865410327911377, places=4)
self.assertAlmostEqual(embeddings[0]['vector'][-1], 0.006781343836337328, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][0], 0.32751473784446716, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][-1], 0.0014184381579980254, places=4)
shutil.rmtree(index_dir)
def test_onnx_encode_unicoil(self):
temp_object = LuceneImpactSearcher(f'{self.index_dir}lucene9-index.cacm', 'SpladePlusPlusEnsembleDistil', encoder_type='onnx')
# this function will never be called in _impact_searcher, here to check quantization correctness
results = temp_object.encode("here is a test")
self.assertEqual(results.get("here"), 156)
self.assertEqual(results.get("a"), 31)
self.assertEqual(results.get("test"), 149)
temp_object.close()
del temp_object
temp_object1 = LuceneImpactSearcher(f'{self.index_dir}lucene9-index.cacm', 'naver/splade-cocondenser-ensembledistil')
# this function will never be called in _impact_searcher, here to check quantization correctness
results = temp_object1.encode("here is a test")
self.assertEqual(results.get("here"), 156)
self.assertEqual(results.get("a"), 31)
self.assertEqual(results.get("test"), 149)
temp_object1.close()
del temp_object1
def test_clip_encoder_cmd_text(self):
index_dir = 'temp_index'
cmd = f'python -m pyserini.encode \
input --corpus {self.test_file} \
--fields text \
output --embeddings {index_dir} \
encoder --encoder openai/clip-vit-base-patch32 \
--fields text \
--batch 1 --max-length 77 \
--device cpu'
status = os.system(cmd)
self.assertEqual(status, 0)
embedding_json_fn = os.path.join(index_dir, 'embeddings.jsonl')
self.assertIsFile(embedding_json_fn)
with open(embedding_json_fn) as f:
embeddings = [json.loads(line) for line in f]
self.assertListEqual([entry["id"] for entry in embeddings], self.docids)
self.assertListEqual(
[entry["contents"] for entry in embeddings],
[entry.strip() for entry in self.texts],
)
self.assertAlmostEqual(embeddings[0]['vector'][0], 0.022726990282535553, places=4)
self.assertAlmostEqual(embeddings[0]['vector'][-1], -0.02527175098657608, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][0], 0.00724585447460413, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][-1], 0.039689723402261734, places=4)
shutil.rmtree(index_dir)
def test_clip_encoder_cmd_image(self):
# special case setup for image data
docids = []
texts = []
test_file = 'tests/resources/sample_collection_jsonl_image/images.small.jsonl'
image_dir = pl.Path(test_file).parent
with open(test_file) as f:
for line in f:
line = json.loads(line)
docids.append(line['id'])
texts.append(line['path'])
index_dir = 'temp_index'
cmd = f'python -m pyserini.encode \
input --corpus {test_file} \
--fields path \
output --embeddings {index_dir} \
encoder --encoder openai/clip-vit-base-patch32 \
--fields path \
--batch 1 --multimodal --l2-norm \
--device cpu'
status = os.system(cmd)
self.assertEqual(status, 0)
embedding_json_fn = os.path.join(index_dir, 'embeddings.jsonl')
self.assertIsFile(embedding_json_fn)
with open(embedding_json_fn) as f:
embeddings = [json.loads(line) for line in f]
self.assertListEqual([entry["id"] for entry in embeddings], docids)
self.assertListEqual(
[entry["contents"] for entry in embeddings],
[str(pl.Path(image_dir, entry.strip())) for entry in texts],
)
self.assertAlmostEqual(embeddings[0]['vector'][0], 0.003283643862232566, places=4)
self.assertAlmostEqual(embeddings[0]['vector'][-1], -0.055951327085494995, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][0], 0.021012384444475174, places=4)
self.assertAlmostEqual(embeddings[2]['vector'][-1], -0.0011692788684740663, places=4)
shutil.rmtree(index_dir)
@classmethod
def tearDownClass(cls):
os.remove(cls.tarball_name)
shutil.rmtree(cls.index_dir)
if __name__ == '__main__':
unittest.main()