Spaces:
Runtime error
Runtime error
File size: 9,060 Bytes
d6585f5 |
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 |
#
# 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.
#
"""Integration tests for ANCE and ANCE PRF using on-the-fly query encoding."""
import os
import socket
import unittest
from integrations.utils import clean_files, run_command, parse_score, parse_score_qa, parse_score_msmarco
from pyserini.search import QueryEncoder
from pyserini.search import get_topics
class TestSearchIntegration(unittest.TestCase):
def setUp(self):
self.temp_files = []
self.threads = 16
self.batch_size = 256
self.rocchio_alpha = 0.4
self.rocchio_beta = 0.6
# Hard-code larger values for internal servers
if socket.gethostname().startswith('damiano') or socket.gethostname().startswith('orca'):
self.threads = 36
self.batch_size = 144
def test_ance_encoded_queries(self):
encoded = QueryEncoder.load_encoded_queries('ance-msmarco-passage-dev-subset')
topics = get_topics('msmarco-passage-dev-subset')
for t in topics:
self.assertTrue(topics[t]['title'] in encoded.embedding)
encoded = QueryEncoder.load_encoded_queries('ance-dl19-passage')
topics = get_topics('dl19-passage')
for t in topics:
self.assertTrue(topics[t]['title'] in encoded.embedding)
encoded = QueryEncoder.load_encoded_queries('ance-dl20')
topics = get_topics('dl20')
for t in topics:
self.assertTrue(topics[t]['title'] in encoded.embedding)
def test_msmarco_passage_ance_avg_prf_otf(self):
output_file = 'test_run.dl2019.ance.avg-prf.otf.trec'
self.temp_files.append(output_file)
cmd1 = f'python -m pyserini.search.faiss --topics dl19-passage \
--index msmarco-passage-ance-bf \
--encoder castorini/ance-msmarco-passage \
--batch-size {self.batch_size} \
--threads {self.threads} \
--output {output_file} \
--prf-depth 3 \
--prf-method avg'
cmd2 = f'python -m pyserini.eval.trec_eval -l 2 -m map dl19-passage {output_file}'
status = os.system(cmd1)
stdout, stderr = run_command(cmd2)
score = parse_score(stdout, 'map')
self.assertEqual(status, 0)
self.assertAlmostEqual(score, 0.4247, delta=0.0001)
def test_msmarco_passage_ance_rocchio_prf_otf(self):
output_file = 'test_run.dl2019.ance.rocchio-prf.otf.trec'
self.temp_files.append(output_file)
cmd1 = f'python -m pyserini.search.faiss --topics dl19-passage \
--index msmarco-passage-ance-bf \
--encoder castorini/ance-msmarco-passage \
--batch-size {self.batch_size} \
--threads {self.threads} \
--output {output_file} \
--prf-depth 5 \
--prf-method rocchio \
--rocchio-topk 5 \
--threads {self.threads} \
--rocchio-alpha {self.rocchio_alpha} \
--rocchio-beta {self.rocchio_beta}'
cmd2 = f'python -m pyserini.eval.trec_eval -l 2 -m map dl19-passage {output_file}'
status = os.system(cmd1)
stdout, stderr = run_command(cmd2)
score = parse_score(stdout, 'map')
self.assertEqual(status, 0)
self.assertAlmostEqual(score, 0.4211, delta=0.0001)
def test_msmarco_doc_ance_bf_otf(self):
output_file = 'test_run.msmarco-doc.passage.ance-maxp.otf.txt'
self.temp_files.append(output_file)
cmd1 = f'python -m pyserini.search.faiss --topics msmarco-doc-dev \
--index msmarco-doc-ance-maxp-bf \
--encoder castorini/ance-msmarco-doc-maxp \
--output {output_file}\
--hits 1000 \
--max-passage \
--max-passage-hits 100 \
--output-format msmarco \
--batch-size {self.batch_size} \
--threads {self.threads}'
cmd2 = f'python -m pyserini.eval.msmarco_doc_eval --judgments msmarco-doc-dev --run {output_file}'
status = os.system(cmd1)
stdout, stderr = run_command(cmd2)
score = parse_score_msmarco(stdout, 'MRR @100')
self.assertEqual(status, 0)
# We get a small difference, 0.3794 on macOS.
self.assertAlmostEqual(score, 0.3796, delta=0.0002)
def test_msmarco_doc_ance_bf_encoded_queries(self):
encoder = QueryEncoder.load_encoded_queries('ance_maxp-msmarco-doc-dev')
topics = get_topics('msmarco-doc-dev')
for t in topics:
self.assertTrue(topics[t]['title'] in encoder.embedding)
def test_nq_test_ance_bf_otf(self):
output_file = 'test_run.ance.nq-test.multi.bf.otf.trec'
retrieval_file = 'test_run.ance.nq-test.multi.bf.otf.json'
self.temp_files.extend([output_file, retrieval_file])
cmd1 = f'python -m pyserini.search.faiss --topics dpr-nq-test \
--index wikipedia-ance-multi-bf \
--encoder castorini/ance-dpr-question-multi \
--output {output_file} \
--batch-size {self.batch_size} --threads {self.threads}'
cmd2 = f'python -m pyserini.eval.convert_trec_run_to_dpr_retrieval_run --topics dpr-nq-test \
--index wikipedia-dpr \
--input {output_file} \
--output {retrieval_file}'
cmd3 = f'python -m pyserini.eval.evaluate_dpr_retrieval --retrieval {retrieval_file} --topk 20'
status1 = os.system(cmd1)
status2 = os.system(cmd2)
stdout, stderr = run_command(cmd3)
score = parse_score_qa(stdout, 'Top20')
self.assertEqual(status1, 0)
self.assertEqual(status2, 0)
self.assertAlmostEqual(score, 0.8224, places=4)
def test_nq_test_ance_encoded_queries(self):
encoder = QueryEncoder.load_encoded_queries('dpr_multi-nq-test')
topics = get_topics('dpr-nq-test')
for t in topics:
self.assertTrue(topics[t]['title'] in encoder.embedding)
def test_trivia_test_ance_bf_otf(self):
output_file = 'test_run.ance.trivia-test.multi.bf.otf.trec'
retrieval_file = 'test_run.ance.trivia-test.multi.bf.otf.json'
self.temp_files.extend([output_file, retrieval_file])
cmd1 = f'python -m pyserini.search.faiss --topics dpr-trivia-test \
--index wikipedia-ance-multi-bf \
--encoder castorini/ance-dpr-question-multi \
--output {output_file} \
--batch-size {self.batch_size} --threads {self.threads}'
cmd2 = f'python -m pyserini.eval.convert_trec_run_to_dpr_retrieval_run --topics dpr-trivia-test \
--index wikipedia-dpr \
--input {output_file} \
--output {retrieval_file}'
cmd3 = f'python -m pyserini.eval.evaluate_dpr_retrieval --retrieval {retrieval_file} --topk 20'
status1 = os.system(cmd1)
status2 = os.system(cmd2)
stdout, stderr = run_command(cmd3)
score = parse_score_qa(stdout, 'Top20')
self.assertEqual(status1, 0)
self.assertEqual(status2, 0)
self.assertAlmostEqual(score, 0.8010, places=4)
def test_trivia_test_ance_encoded_queries(self):
encoder = QueryEncoder.load_encoded_queries('dpr_multi-trivia-test')
topics = get_topics('dpr-trivia-test')
for t in topics:
self.assertTrue(topics[t]['title'] in encoder.embedding)
def tearDown(self):
clean_files(self.temp_files)
if __name__ == '__main__':
unittest.main()
|