Repository navigation
Expand file tree
/
Copy pathadapter_query.py
More file actions
237 lines (215 loc) · 9.19 KB
/
Copy pathadapter_query.py
File metadata and controls
237 lines (215 loc) · 9.19 KB
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
# -*- coding: utf-8 -*-
import asyncio
import logging
from modelcache.embedding import MetricType
from modelcache.utils.time import time_cal
from FlagEmbedding import FlagReranker
USE_RERANKER = False # 如果为 True 则启用 reranker,否则使用原有逻辑
async def adapt_query(cache_data_convert, *args, **kwargs):
# Extract query parameters
chat_cache = kwargs.pop("cache_obj")
scope = kwargs.pop("scope")
model = scope['model']
context = kwargs.pop("cache_context", {})
cache_factor = kwargs.pop("cache_factor", 1.0)
# Preprocess query for embedding generation
pre_embedding_data = chat_cache.query_pre_embedding_func(
kwargs,
extra_param=context.get("pre_embedding_func", None),
prompts=chat_cache.prompts,
)
# Generate embedding with performance monitoring
embedding_data = await time_cal(
chat_cache.embedding_func,
func_name="embedding",
report_func=chat_cache.report.embedding,
cache_obj=chat_cache
)(pre_embedding_data)
search_time_cal = time_cal(
chat_cache.data_manager.search,
func_name="vector_search",
report_func=chat_cache.report.search,
cache_obj=chat_cache
)
cache_data_list = await asyncio.to_thread(
search_time_cal,
embedding_data,
extra_param=context.get("search_func", None),
top_k=kwargs.pop("top_k", -1),
model=model
)
# Initialize result containers
cache_answers = []
cache_questions = []
cache_ids = []
cosine_similarity = None
# Similarity evaluation based on metric type
if chat_cache.similarity_metric_type == MetricType.COSINE:
cosine_similarity = cache_data_list[0][0]
# This code uses the built-in cosine similarity evaluation in milvus
if cosine_similarity < chat_cache.similarity_threshold:
return None # No suitable match found
elif chat_cache.similarity_metric_type == MetricType.L2:
# this is the code that uses L2 for similarity evaluation
similarity_threshold = chat_cache.similarity_threshold
similarity_threshold_long = chat_cache.similarity_threshold_long
min_rank, max_rank = chat_cache.similarity_evaluation.range()
rank_threshold = (max_rank - min_rank) * similarity_threshold * cache_factor
rank_threshold_long = (max_rank - min_rank) * similarity_threshold_long * cache_factor
# Clamp thresholds to valid range
rank_threshold = (
max_rank
if rank_threshold > max_rank
else min_rank
if rank_threshold < min_rank
else rank_threshold
)
rank_threshold_long = (
max_rank
if rank_threshold_long > max_rank
else min_rank
if rank_threshold_long < min_rank
else rank_threshold_long
)
# Evaluate similarity score
if cache_data_list is None or len(cache_data_list) == 0:
rank_pre = -1.0
else:
cache_data_dict = {'search_result': cache_data_list[0]}
rank_pre = chat_cache.similarity_evaluation.evaluation(
None,
cache_data_dict,
extra_param=context.get("evaluation_func", None),
)
if rank_pre < rank_threshold:
return None # Similarity too low
else:
raise ValueError(
f"Unsupported similarity metric type: {chat_cache.similarity_metric_type}"
)
# Process search results with optional reranking
if USE_RERANKER:
reranker = FlagReranker('BAAI/bge-reranker-v2-m3', use_fp16=False)
for cache_data in cache_data_list:
primary_id = cache_data[1]
ret = await asyncio.to_thread(
chat_cache.data_manager.get_scalar_data,
cache_data, extra_param=context.get("get_scalar_data", None), model=model
)
if ret is None:
continue
rank = reranker.compute_score([pre_embedding_data, ret[0]], normalize=True)[0]
if "deps" in context and hasattr(ret.question, "deps"):
eval_query_data = {
"question": context["deps"][0]["data"],
"embedding": None
}
eval_cache_data = {
"question": ret.question.deps[0].data,
"answer": ret.answers[0].answer,
"search_result": cache_data,
"embedding": None,
}
else:
eval_query_data = {
"question": pre_embedding_data,
"embedding": embedding_data,
}
eval_cache_data = {
"question": ret[0],
"answer": ret[1],
"search_result": cache_data,
"embedding": None
}
if len(pre_embedding_data) <= 256:
if rank_threshold <= rank:
cache_answers.append((rank, ret[0]))
cache_questions.append((rank, ret[1]))
cache_ids.append((rank, primary_id))
else:
if rank_threshold_long <= rank:
cache_answers.append((rank, ret[0]))
cache_questions.append((rank, ret[1]))
cache_ids.append((rank, primary_id))
else:
# Original logic without reranking
for cache_data in cache_data_list:
primary_id = cache_data[1]
# Retrieve full cache entry data
ret = await asyncio.to_thread(
chat_cache.data_manager.get_scalar_data,
cache_data, extra_param=context.get("get_scalar_data", None), model=model
)
if ret is None:
continue
if chat_cache.similarity_metric_type == MetricType.COSINE:
assert cosine_similarity is not None, "cosine_similarity should not be None"
cache_answers.append((cosine_similarity, ret[0]))
cache_questions.append((cosine_similarity, ret[1]))
cache_ids.append((cosine_similarity, primary_id))
elif chat_cache.similarity_metric_type == MetricType.L2:
if "deps" in context and hasattr(ret.question, "deps"):
eval_query_data = {
"question": context["deps"][0]["data"],
"embedding": None
}
eval_cache_data = {
"question": ret.question.deps[0].data,
"answer": ret.answers[0].answer,
"search_result": cache_data,
"embedding": None,
}
else:
eval_query_data = {
"question": pre_embedding_data,
"embedding": embedding_data,
}
eval_cache_data = {
"question": ret[0],
"answer": ret[1],
"search_result": cache_data,
"embedding": None
}
# Evaluate similarity for this specific result
rank = chat_cache.similarity_evaluation.evaluation(
eval_query_data,
eval_cache_data,
extra_param=context.get("evaluation_func", None),
)
if len(pre_embedding_data) <= 256:
if rank_threshold <= rank:
cache_answers.append((rank, ret[0]))
cache_questions.append((rank, ret[1]))
cache_ids.append((rank, primary_id))
else:
if rank_threshold_long <= rank:
cache_answers.append((rank, ret[0]))
cache_questions.append((rank, ret[1]))
cache_ids.append((rank, primary_id))
else:
raise ValueError(
f"Unsupported similarity metric type: {chat_cache.similarity_metric_type}"
)
# Sort results by similarity score (highest first)
cache_answers = sorted(cache_answers, key=lambda x: x[0], reverse=True)
cache_questions = sorted(cache_questions, key=lambda x: x[0], reverse=True)
cache_ids = sorted(cache_ids, key=lambda x: x[0], reverse=True)
if len(cache_answers) != 0:
return_message = chat_cache.post_process_messages_func(
[t[1] for t in cache_answers]
)
return_query = chat_cache.post_process_messages_func(
[t[1] for t in cache_questions]
)
return_id = chat_cache.post_process_messages_func(
[t[1] for t in cache_ids]
)
# Update hit count for analytics (async to avoid blocking)
try:
asyncio.create_task(asyncio.to_thread(chat_cache.data_manager.update_hit_count,return_id))
except Exception:
logging.info('update_hit_count except, please check!')
# Record cache hit for reporting
chat_cache.report.hint_cache()
return cache_data_convert(return_message, return_query)
return None