-
Notifications
You must be signed in to change notification settings - Fork 0
/
source_selection.py
298 lines (262 loc) · 13.7 KB
/
source_selection.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
from models.openai_models import GPT3_5Model, GPT4Model
from models.llama_models import LLAMA2_CHATMODEL
from models.flan_models import FLANModel
from models.mixtral_models import MIXTRALModel
from models.falcon_models import FALCONModel
from models.solar_models import SOLARModel
from source.base_source import BaseSource
import argparse
import csv
import os
import json
from tqdm import tqdm
api_key_path = "api_key"
def main(queries_file, engines_file, query_representation, llm_type, source_representation, batch, model_path, out_file, search_query_result=None, snippet_folder=None, search_top_n=None):
queries, sources = load_dataset(queries_file, engines_file)
# Initialize LLM based on the type
api_key = open(api_key_path, "r").read()
if llm_type == 'gpt-4':
model = GPT4Model(api_key=api_key)
model.load_model()
# Add other models here based on llm_type
elif llm_type == 'gpt-3.5':
model = GPT3_5Model(api_key=api_key)
model.load_model()
elif "llama2" in llm_type:
import random
random.seed(0)
model = LLAMA2_CHATMODEL()
model.load_model(model_path=model_path)
elif "flan" in llm_type:
import random
random.seed(0)
model = FLANModel()
model.load_model(model_path=model_path)
elif "mixtral" in llm_type:
import random
random.seed(0)
model = MIXTRALModel()
model.load_model(model_path=model_path)
elif "falcon" in llm_type:
import random
random.seed(0)
model = FALCONModel()
model.load_model(model_path=model_path)
elif "solar" in llm_type:
import random
random.seed(0)
model = SOLARModel()
model.load_model(model_path=model_path)
else:
raise ValueError("Unsupported LLM type")
representation_dict = {}
for source_id in sources:
# Initialize source based on the representation type
source_info_dict = sources[source_id]
source_info = BaseSource(source_info_dict)
representation = source_info.get_representation(source_representation, llm_type)
print(representation)
representation_dict[source_id] = representation
if source_representation == "example_snippet" or source_representation == "name_snippet":
real_docs_dict = {}
search_query_dict = load_example_result(search_query_result, search_top_n)
search_content_dict = load_search_content(snippet_folder, "example_doc")
for qid in search_query_dict:
real_docs_dict[qid] = {}
for engine_id in search_query_dict[qid]:
real_docs_dict[qid][engine_id] = []
for tem_id in search_query_dict[qid][engine_id]:
real_docs_dict[qid][engine_id].append(search_content_dict[engine_id][tem_id])
elif source_representation == "example_querydoc" or source_representation == "name_querydoc" or source_representation=="realquery_querydoc":
real_docs_dict = {}
search_query_dict = load_example_result(search_query_result, 1)
search_content_dict = load_search_content(snippet_folder, "example_querydoc")
for qid in search_query_dict:
real_docs_dict[qid] = {}
for engine_id in search_query_dict[qid]:
real_docs_dict[qid][engine_id] = {}
tem_id = search_query_dict[qid][engine_id][0]
real_docs_dict[qid][engine_id]["query"] = search_content_dict[engine_id][tem_id]["query"]
if len(search_content_dict[engine_id][tem_id]["snippets"])>search_top_n:
real_docs_dict[qid][engine_id]["snippets"] = search_content_dict[engine_id][tem_id]["snippets"][0:search_top_n]
else:
real_docs_dict[qid][engine_id]["snippets"] = search_content_dict[engine_id][tem_id]["snippets"]
already_finished_qids = check_already_finished_qids(out_file)
chunked_representation_dict = chunk_dict(representation_dict, int(batch))
for qid in tqdm(queries):
# Decide source for each query versus source representation
if qid in already_finished_qids:
continue
query = queries[qid][query_representation]
score_dict = {}
for chunk in chunked_representation_dict:
if (llm_type == 'gpt-4') or (llm_type == 'gpt-3.5'):
for source_id in chunk:
decision_score = model.predict(query, representation_dict[source_id])
#print(f"Query: {query}\nSource: {representation_dict[source_id]}\nDecision Score: {decision_score}\n")
score_dict[source_id] = decision_score
# Sort sources based on the decision score
else:
if source_representation == "example_snippet" or source_representation == "name_snippet":
result_dict = model.batch_predict(query, chunk, real_docs_dict[qid])
elif source_representation == "example_querydoc" or source_representation == "name_querydoc" or source_representation == "realquery_querydoc":
result_dict = model.batch_predict(query, chunk, real_docs_dict[qid], True)
else:
result_dict = model.batch_predict(query, chunk)
for source_id in result_dict:
score_dict[source_id] = result_dict[source_id]
sorted_sources = sorted(score_dict.items(), key=lambda x: x[1], reverse=True)
# Write the results to a file
write_trec_results(qid, sorted_sources, llm_type, out_file)
def chunk_dict(input_dict, chunk_size):
# Sort the dictionary based on the length of the value
sorted_items = sorted(input_dict.items(), key=lambda x: len(x[1].split()))
# Create chunks
chunks = []
current_chunk = {}
current_size = 0
for key, value in sorted_items:
if current_size + 1 > chunk_size and current_chunk:
chunks.append(current_chunk)
current_chunk = {}
current_size = 0
current_chunk[key] = value
current_size += 1
# Add the last chunk if it's not empty
if current_chunk:
chunks.append(current_chunk)
print(f"Number of chunks: {len(chunks)}")
return chunks
def check_already_finished_qids(result_file):
# This function get all qids from the result filem in trec format
# You can use this function to check if the query has been processed
qids = []
if not os.path.exists(result_file):
return set(qids)
with open(result_file, "r") as f:
for line in f:
qid = line.split()[0]
qids.append(qid)
return set(qids)
def write_trec_results(qid, sorted_sources, label, out_file):
# This function writes the results to a file
# The results should be in TREC format
# The file name should be {qid}.trec
# The format of each line should be {qid} Q0 {source_id} {rank} {decision_score} {run_name}
if not os.path.exists(out_file):
#check if the folder exist first
out_folder = os.path.dirname(out_file)
if not os.path.exists(out_folder):
os.makedirs(out_folder)
#create the file
fw = open(out_file, "w")
else:
fw = open(out_file, "a")
rank = 1
for source_id, decision_score in sorted_sources:
# if decision_score <= 0:
# continue
fw.write(f"{qid} Q0 {source_id} {rank} {decision_score} {label}\n")
rank += 1
fw.close()
def load_dataset(queries_file, engines_file):
queries = {}
engines = {}
# Load queries
with open(queries_file, 'r') as file:
reader = csv.DictReader(file)
for row in reader:
query_id = row['qid']
queries[query_id] = row
# Load engines
with open(engines_file, 'r') as file:
reader = csv.DictReader(file)
for row in reader:
engine_id = row['engineID']
engines[engine_id] = row
return queries, engines
def load_example_result(example_result_file, n):
# this is a dictionary
# key is the query id
initial_dict = json.load(open(example_result_file, "r"))
result_dict = {}
for qid in initial_dict:
result_dict[qid] = {}
for engine_id in initial_dict[qid]:
result_dict[qid][engine_id] = [tem_id for tem_id, score in initial_dict[qid][engine_id][:n]]
return result_dict
import glob
def load_search_content(search_folder, search_type):
"""
Load search content from subfolders within the specified search folder.
Depending on the search_type, load different fields into a dictionary.
"""
all_content = {}
prefile_data = ""
if "FW13" in search_folder:
prefile_data = "FW13-"
elif "FW14" in search_folder:
prefile_data = "FW14-"
# Iterate over subfolders in the search folder
for subfolder in os.listdir(search_folder):
search_content = {}
subfolder_path = os.path.join(search_folder, subfolder)
if os.path.isdir(subfolder_path):
jsonl_files = glob.glob(os.path.join(subfolder_path + "/*.jsonl"))
if len(jsonl_files) == 0:
continue
for json_file in jsonl_files:
if os.path.isfile(json_file):
with open(json_file, 'r') as file:
data = json.load(file)
for key, item in data.items():
if (search_type == 'example_querydoc'):
# Extract the 'query' field
search_content[key] = {}
search_content[key]['query'] = item['query']
search_content[key]['snippets'] = []
for snippet in item['snippets']:
for snippet_id, snippet_details in snippet.items():
tem_content = snippet_details['title'] if snippet_details['title'] is not None else 'None'
tem_content += ' '
tem_content += snippet_details['description'] if snippet_details['description'] is not None else 'None'
search_content[key]['snippets'].append(tem_content)
elif search_type == 'example_doc':
# Extract the 'title' from each snippet
for snippet in item['snippets']:
for snippet_id, snippet_details in snippet.items():
tem_content = "Title: "
tem_content += snippet_details['title'].strip() if snippet_details[
'title'] is not None else 'None'
tem_content += '\n'
tem_content = "Description: "
tem_content += snippet_details['description'] if snippet_details[
'description'] is not None else 'None'
search_content[snippet_id] = tem_content
else:
raise ValueError("Invalid search type. Must be 'query' or 'snippet'.")
all_content[prefile_data + subfolder.replace('/', "")] = search_content
return all_content
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Federated Search Source Selection')
parser.add_argument('--queries', type=str, required=True, help='Path to the queries file')
parser.add_argument('--engines', type=str, required=True, help='Path to the engines file')
parser.add_argument('--llm', type=str, choices=['gpt-4', 'gpt-3.5', 'llama2-7b-chat', 'llama2-13b-chat', "flan-xxl", "flan-large", "flan-xl", "mixtral-13b", "falcon-7b", "solar-11b", "flan-large-tuned","flan-xl-tuned", "llama2-7b-chat-tuned"], required=True, help='Type of Large Language Model to use')
parser.add_argument('--query-representation', type=str, choices=['query', 'description','narrative'], required=True,
help='Type of query representation')
parser.add_argument('--source-representation', type=str, choices=['name', 'name_description', "example_querydoc", "example_snippet", "name_querydoc", "name_snippet", "realquery_querydoc", "querylog"], required=True, help='Type of source representation')
parser.add_argument('--batch', type=int, default=1, help='the batch size')
parser.add_argument('--model_path', type=str, default='llama2-7b-chat', help='the path of the model')
parser.add_argument('--out_file', type=str, required=True, help='output file')
args, _ = parser.parse_known_args()
# Conditional arguments based on the value of source-representation
if (args.source_representation == "example_snippet") or (args.source_representation == "example_querydoc") or (args.source_representation == "name_snippet") or (args.source_representation == "name_querydoc") or (args.source_representation == "realquery_querydoc"):
parser.add_argument('--search_query_result', type=str, required=True,
help='Path to the search query result file')
parser.add_argument('--snippet_folder', type=str, required=True, help='Path to the snippet folder')
parser.add_argument('--search_top_n', type=int, required=True, help='top n example in the search query result file')
# Parse again including the new arguments
args = parser.parse_args()
main(args.queries, args.engines, args.query_representation, args.llm, args.source_representation, args.batch,
args.model_path, args.out_file, getattr(args, 'search_query_result', None),
getattr(args, 'snippet_folder', None), getattr(args, 'search_top_n', None))