antoinelouis
commited on
Commit
•
075dfca
1
Parent(s):
566931f
Create custom.py
Browse files
custom.py
ADDED
@@ -0,0 +1,148 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from langdetect import detect
|
3 |
+
import torch.multiprocessing as mp
|
4 |
+
|
5 |
+
from colbert import Indexer, Searcher
|
6 |
+
from colbert.infra import ColBERTConfig, Run
|
7 |
+
from colbert.utils.utils import print_message
|
8 |
+
from colbert.data.collection import Collection
|
9 |
+
from colbert.modeling.checkpoint import Checkpoint
|
10 |
+
from colbert.indexing.index_saver import IndexSaver
|
11 |
+
from colbert.search.index_storage import IndexScorer
|
12 |
+
from colbert.infra.launcher import Launcher, print_memory_stats
|
13 |
+
from colbert.indexing.collection_encoder import CollectionEncoder
|
14 |
+
from colbert.indexing.collection_indexer import CollectionIndexer
|
15 |
+
|
16 |
+
|
17 |
+
MMARCO_LANGUAGES = {
|
18 |
+
'ar': ('arabic', 'ar_AR'),
|
19 |
+
'de': ('german', 'de_DE'),
|
20 |
+
'en': ('english', 'en_XX'),
|
21 |
+
'es': ('spanish', 'es_XX'),
|
22 |
+
'fr': ('french', 'fr_XX'),
|
23 |
+
'hi': ('hindi', 'hi_IN'),
|
24 |
+
'id': ('indonesian', 'id_ID'),
|
25 |
+
'it': ('italian', 'it_IT'),
|
26 |
+
'ja': ('japanese', 'ja_XX'),
|
27 |
+
'nl': ('dutch', 'nl_XX'),
|
28 |
+
'pt': ('portuguese', 'pt_XX'),
|
29 |
+
'ru': ('russian', 'ru_RU'),
|
30 |
+
'vi': ('vietnamese', 'vi_VN'),
|
31 |
+
'zh': ('chinese', 'zh_CN'),
|
32 |
+
}
|
33 |
+
MRTYDI_LANGUAGES = {
|
34 |
+
'ar': ('arabic', 'ar_AR'),
|
35 |
+
'bn': ('bengali', 'bn_IN'),
|
36 |
+
'en': ('english', 'en_XX'),
|
37 |
+
'fi': ('finnish', 'fi_FI'),
|
38 |
+
'id': ('indonesian', 'id_ID'),
|
39 |
+
'ja': ('japanese', 'ja_XX'),
|
40 |
+
'ko': ('korean', 'ko_KR'),
|
41 |
+
'ru': ('russian', 'ru_RU'),
|
42 |
+
'sw': ('swahili', 'sw_KE'),
|
43 |
+
'te': ('telugu', 'te_IN'),
|
44 |
+
'th': ('thai', 'th_TH'),
|
45 |
+
}
|
46 |
+
MIRACL_LANGUAGES = {
|
47 |
+
'ar': ('arabic', 'ar_AR'),
|
48 |
+
'bn': ('bengali', 'bn_IN'),
|
49 |
+
'en': ('english', 'en_XX'),
|
50 |
+
'es': ('spanish', 'es_XX'),
|
51 |
+
'fa': ('persian', 'fa_IR'),
|
52 |
+
'fi': ('finnish', 'fi_FI'),
|
53 |
+
'fr': ('french', 'fr_XX'),
|
54 |
+
'hi': ('hindi', 'hi_IN'),
|
55 |
+
'id': ('indonesian', 'id_ID'),
|
56 |
+
'ja': ('japanese', 'ja_XX'),
|
57 |
+
'ko': ('korean', 'ko_KR'),
|
58 |
+
'ru': ('russian', 'ru_RU'),
|
59 |
+
'sw': ('swahili', 'sw_KE'),
|
60 |
+
'te': ('telugu', 'te_IN'),
|
61 |
+
'th': ('thai', 'th_TH'),
|
62 |
+
'zh': ('chinese', 'zh_CN'),
|
63 |
+
}
|
64 |
+
ALL_LANGUAGES = {**MMARCO_LANGUAGES, **MRTYDI_LANGUAGES, **MIRACL_LANGUAGES}
|
65 |
+
|
66 |
+
|
67 |
+
def set_xmod_language(model, lang:str):
|
68 |
+
"""
|
69 |
+
Set the default language code for the model. This is used when the language is not specified in the input.
|
70 |
+
Source: https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/xmod/modeling_xmod.py#L687
|
71 |
+
"""
|
72 |
+
lang = lang.split('-')[0]
|
73 |
+
if (value := ALL_LANGUAGES.get(lang)) is not None:
|
74 |
+
model.set_default_language(value[1])
|
75 |
+
else:
|
76 |
+
raise KeyError(f"Language {lang} not supported.")
|
77 |
+
|
78 |
+
#-----------------------------------------------------------------------------------------------------------------#
|
79 |
+
# INDEXER
|
80 |
+
#-----------------------------------------------------------------------------------------------------------------#
|
81 |
+
class CustomIndexer(Indexer):
|
82 |
+
def __launch(self, collection):
|
83 |
+
manager = mp.Manager()
|
84 |
+
shared_lists = [manager.list() for _ in range(self.config.nranks)]
|
85 |
+
shared_queues = [manager.Queue(maxsize=1) for _ in range(self.config.nranks)]
|
86 |
+
launcher = Launcher(custom_encode)
|
87 |
+
launcher.launch(self.config, collection, shared_lists, shared_queues, self.verbose)
|
88 |
+
|
89 |
+
def custom_encode(config, collection, shared_lists, shared_queues, verbose: int = 3):
|
90 |
+
encoder = CustomCollectionIndexer(config=config, collection=collection, verbose=verbose)
|
91 |
+
encoder.run(shared_lists)
|
92 |
+
|
93 |
+
class CustomCollectionIndexer(CollectionIndexer):
|
94 |
+
def __init__(self, config: ColBERTConfig, collection, verbose=2):
|
95 |
+
self.verbose = verbose
|
96 |
+
self.config = config
|
97 |
+
self.rank, self.nranks = self.config.rank, self.config.nranks
|
98 |
+
self.use_gpu = self.config.total_visible_gpus > 0
|
99 |
+
if self.config.rank == 0 and self.verbose > 1:
|
100 |
+
self.config.help()
|
101 |
+
self.collection = Collection.cast(collection)
|
102 |
+
self.checkpoint = Checkpoint(self.config.checkpoint, colbert_config=self.config)
|
103 |
+
if self.checkpoint.bert.__class__.__name__.lower().startswith("xmod"):
|
104 |
+
language = detect(self.collection.__getitem__(0))
|
105 |
+
Run().print_main(f"#> Setting X-MOD language adapters to {language}.")
|
106 |
+
set_xmod_language(self.checkpoint.bert, lang=language)
|
107 |
+
if self.use_gpu:
|
108 |
+
self.checkpoint = self.checkpoint.cuda()
|
109 |
+
self.encoder = CollectionEncoder(config, self.checkpoint)
|
110 |
+
self.saver = IndexSaver(config)
|
111 |
+
print_memory_stats(f'RANK:{self.rank}')
|
112 |
+
|
113 |
+
#-----------------------------------------------------------------------------------------------------------------#
|
114 |
+
# SEARCHER
|
115 |
+
#-----------------------------------------------------------------------------------------------------------------#
|
116 |
+
class CustomSearcher(Searcher):
|
117 |
+
def __init__(self, index, checkpoint=None, collection=None, config=None, index_root=None, verbose:int = 3):
|
118 |
+
self.verbose = verbose
|
119 |
+
if self.verbose > 1:
|
120 |
+
print_memory_stats()
|
121 |
+
|
122 |
+
initial_config = ColBERTConfig.from_existing(config, Run().config)
|
123 |
+
|
124 |
+
default_index_root = initial_config.index_root_
|
125 |
+
index_root = index_root if index_root else default_index_root
|
126 |
+
self.index = os.path.join(index_root, index)
|
127 |
+
self.index_config = ColBERTConfig.load_from_index(self.index)
|
128 |
+
|
129 |
+
self.checkpoint = checkpoint or self.index_config.checkpoint
|
130 |
+
self.checkpoint_config = ColBERTConfig.load_from_checkpoint(self.checkpoint)
|
131 |
+
self.config = ColBERTConfig.from_existing(self.checkpoint_config, self.index_config, initial_config)
|
132 |
+
|
133 |
+
self.collection = Collection.cast(collection or self.config.collection)
|
134 |
+
self.configure(checkpoint=self.checkpoint, collection=self.collection)
|
135 |
+
|
136 |
+
self.checkpoint = Checkpoint(self.checkpoint, colbert_config=self.config, verbose=self.verbose)
|
137 |
+
if self.checkpoint.bert.__class__.__name__.lower().startswith("xmod"):
|
138 |
+
language = detect(self.collection.__getitem__(0))
|
139 |
+
print_message(f"#> Setting X-MOD language adapters to {language}.")
|
140 |
+
set_xmod_language(self.checkpoint.bert, lang=language)
|
141 |
+
use_gpu = self.config.total_visible_gpus > 0
|
142 |
+
if use_gpu:
|
143 |
+
self.checkpoint = self.checkpoint.cuda()
|
144 |
+
load_index_with_mmap = self.config.load_index_with_mmap
|
145 |
+
if load_index_with_mmap and use_gpu:
|
146 |
+
raise ValueError(f"Memory-mapped index can only be used with CPU!")
|
147 |
+
self.ranker = IndexScorer(self.index, use_gpu, load_index_with_mmap)
|
148 |
+
print_memory_stats()
|