/
embeddings_cc_index_unikge.py
151 lines (125 loc) · 5.11 KB
/
embeddings_cc_index_unikge.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
# UniKGE/Procrustes & embeddings.cc index API
# https://github.com/dice-group/embeddings.cc
#
# Usage: time python3 api/embeddings_cc_index_unikge.py <PASSWORD>
# Local usage: time python3 api/embeddings_cc_index_unikge.py <PASSWORD> http://127.0.0.1:8008
#
# Authors (for questions): N'Dah Jean Kouagou, Adrian Wilke
#
# Note: To be available via the API, create the following aliases afterwards:
# Index Alias Comments
# dbpedia_en_fr_15k_procrustes dbpedia-en-fr-15k
# dbpedia_en_fr_100k_procrustes_v2 dbpedia-en-fr-100k
# dbpedia_en_de_100k_procrustes dbpedia-en-de-100k
# dbpedia_caligraph dbpedia-caligraph Only DBpedia URIs
import sys
import numpy as np
import json
import time
from embeddings_cc_index import EmbeddingsCcIndex
# Configuration
add_mappings_key = True
add_mappings_value = True
# UniKGE/Procrustes data: https://hobbitdata.informatik.uni-leipzig.de/UniKGE/
# Each folder has to contain: alignment.json, list_merged_entities.txt, Universal_Emb.npy
# Do not forget final '/'
es_indices = []
data_folders = []
if False:
es_indices.append("dbpedia_en_fr_15k_procrustes")
data_folders.append("/home/wilke/Data/UniKGE/dbpedia_en_fr_15k_procrustes/EnFr15KV1/")
if False:
es_indices.append("dbpedia_en_fr_100k_procrustes_v2")
data_folders.append("/home/wilke/Data/UniKGE/dbpedia_en_fr_100k_procrustes/EnFr100KV2/")
if False:
es_indices.append("dbpedia_en_de_100k_procrustes")
data_folders.append("/home/wilke/Data/UniKGE/dbpedia_en_de_100k_procrustes/EnDe100K/")
if False:
es_indices.append("dbpedia_caligraph_procrustes")
data_folders.append("/home/wilke/Data/UniKGE/caligraph_dbpedia_procrustes/")
if False:
es_indices.append("dbpedia_caligraph")
data_folders.append("/home/wilke/Data/UniKGE/caligraph_dbpedia_procrustes/")
add_mappings_key = False # Only add DBpedia URIs from caligraph
# Get password form CLI
if len(sys.argv) > 1:
password = sys.argv[1]
else:
print('Please provide a password')
sys.exit(1)
# Get webservice URL form CLI
webservice_url = None
if len(sys.argv) > 2:
webservice_url = sys.argv[2]
# Create instance
embeddings_cc_index = EmbeddingsCcIndex(webservice_url=webservice_url)
# Ping webservice
statusCode = embeddings_cc_index.ping(seconds=1)
if statusCode == 502:
print('502: Webservice unavailable')
sys.exit(1)
elif statusCode == 503:
print('503: Elasticsearch service unavailable')
sys.exit(1)
else:
print(statusCode)
# Add data
def add_embeddings(api, password, index, embeddings, i):
response = api.add(password, index, embeddings)
if response.status_code == 200:
print(i, end=' ')
sys.stdout.flush()
return True
else:
print('Could not add embeddings', response.status_code, response.text)
return False
# Wait some time if CPU usage is high
def optional_sleep(embeddings_cc_index, password):
if int(embeddings_cc_index.get_max_cpu_usage(password).text) > 50:
print('(s)', end=' ')
sys.stdout.flush()
time.sleep(30)
# recursive call:
optional_sleep(embeddings_cc_index, password)
# Parse files
for es_index, data_folder in zip(es_indices, data_folders):
# Delete existing index
print(f"\nUploading universal embeddings for {data_folder}\n")
response = embeddings_cc_index.delete_index(password, es_index)
if response.status_code == 200:
print('Deleting index', response.status_code, response.text)
# Read data
print('\nLoading embeddings...\n')
with open(data_folder + 'list_merged_entities.txt') as file:
entities = file.read().split('\t')
with open(data_folder + 'alignment.json') as file:
alignment = json.load(file)
embeddings = np.load(data_folder + 'Universal_Emb.npy', mmap_mode='r+')
print('Done!\n')
# Create index
response = embeddings_cc_index.create_index(password, es_index, embeddings.shape[1], shards=5)
print('Creating index', response.status_code, response.text)
# Add data
Embeddings = []
print("\nDimensions:", embeddings.shape[1], "\n")
assert len(entities) == embeddings.shape[0], f"{len(entities)}, {embeddings.shape[0]}"
for i, ent in enumerate(entities):
if add_mappings_key:
Embeddings.append([ent, list(embeddings[i])])
if add_mappings_value and (ent in alignment):
Embeddings.append([alignment[ent], list(embeddings[i])])
if len(Embeddings) >= 5000:
if not add_embeddings(embeddings_cc_index, password, es_index, Embeddings, i):
break
Embeddings = []
optional_sleep(embeddings_cc_index, password)
optional_sleep(embeddings_cc_index, password)
if len(Embeddings) > 0:
add_embeddings(embeddings_cc_index, password, es_index, Embeddings, 0)
# Print number of documents
# In tests, this resulted in 'httpcore.ReadTimeout: timed out'
if True:
optional_sleep(embeddings_cc_index, password)
for es_index, data_folder in zip(es_indices, data_folders):
response = embeddings_cc_index.count(es_index)
print(response.status_code, response.text)