-
Notifications
You must be signed in to change notification settings - Fork 7
/
run.py
180 lines (140 loc) · 5.88 KB
/
run.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
import argparse
import json
import os
import sys
from time import time
from functools import partial
import autogen
from ecoassistant import RetrievalAgent, \
initial_assistant_hierarchy, model_evaluation_function, human_evaluation_function, mixed_evaluation_function
KEY_LOC = "./"
here = os.path.abspath(os.path.dirname(__file__))
class Transcript(object):
def __init__(self, filename):
self.terminal = sys.stdout
self.logfile = open(filename, "w")
def write(self, message):
self.terminal.write(message)
self.terminal.flush()
self.logfile.write(message)
self.logfile.flush()
def flush(self):
# this flush method is needed for python 3 compatibility.
# this handles the flush command by doing nothing.
# you might want to specify some extra behavior here.
pass
def start(filename):
"""Start transcript, appending print output to given filename"""
sys.stdout = Transcript(filename)
def stop():
"""Stop transcript and return print functionality to normal"""
sys.stdout.logfile.close()
sys.stdout = sys.stdout.terminal
def get_llm_config(model, openai_key=None):
if openai_key is None:
config_list = autogen.config_list_from_models(key_file_path=KEY_LOC, model_list=[model])
else:
config_list = [{'api_key': openai_key, 'model': model}]
return config_list
def get_llm_configs(models, openai_key=None):
config_lists = []
for model in models:
if model == 'Llama-2-13b-chat-hf':
config_list = [{
'api_key' : 'llama',
'api_base' : 'http://localhost:8000/v1',
'model' : 'meta-llama/Llama-2-13b-chat-hf',
'max_tokens': 945,
}]
else:
config_list = get_llm_config(model, openai_key=openai_key)
config_lists.append(config_list)
return config_lists
def main(args, queries):
APIs = json.load(open(args.path_to_key, 'r'))
API_tokens = {
'google places': '181dbb37',
'weatherapi' : 'b4d5490d',
'alphavantage' : 'af8fb19b'
}
openai_key = APIs.pop('openai')
models = args.model.split(',')
config_lists = get_llm_configs(models, openai_key=openai_key)
assistant = initial_assistant_hierarchy(models, config_lists, seed=args.seed, name='assistant')
if args.eval == 'human':
evaluation_func = partial(human_evaluation_function, cache=os.path.join(args.output_dir, 'human_eval_cache.json'))
else:
config_list = get_llm_config('gpt-4', openai_key=openai_key)
llm_config = {
"model" : 'gpt-4',
"request_timeout": 600,
"seed" : 43,
"config_list" : config_list
}
if args.eval == 'llm':
evaluation_func = model_evaluation_function(llm_config)
elif args.eval == 'mix':
evaluation_func = mixed_evaluation_function(llm_config, cache=os.path.join(args.output_dir, 'human_eval_cache.json'))
else:
raise NotImplementedError
code_execution_config = {
"work_dir" : args.work_dir,
"use_docker": "python:3" if args.docker else False,
"timeout" : 300,
}
user = RetrievalAgent(
"user",
chain_of_thought=args.cot,
apis=APIs if args.api else {},
api_tokens=API_tokens if args.api else {},
solution_demonstration=args.solution_demonstration,
retrieval_topk=args.retrieval_topk,
retrieval_threshold=args.retrieval_threshold,
evaluation_function=evaluation_func,
max_consecutive_auto_reply=5,
code_execution_config=code_execution_config,
)
start(os.path.join(args.output_dir, f'{args.name}.txt'))
print(args)
logs = {}
time_start = time()
logs['args'] = args.__dict__
for i, raw_query in enumerate(queries):
print(f'>>>>>>>> Query {i}: {raw_query}')
log = user.initiate_chat(raw_query, assistant)
logs[raw_query] = log
run_time = time() - time_start
logs['run_time'] = run_time
print(f'Run time: {run_time}')
stop()
return logs
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--model', type=str, default="gpt-3.5-turbo,gpt-4")
parser.add_argument('--eval', type=str, default="llm", choices=['human', 'llm', 'mix'])
parser.add_argument('--docker', action="store_true", default=False)
parser.add_argument('--api', action="store_true", default=False)
parser.add_argument('--cot', action="store_true", default=False)
parser.add_argument('--solution_demonstration', action="store_true", default=False)
parser.add_argument('--retrieval_threshold', type=float, default=0.5)
parser.add_argument('--retrieval_topk', type=int, default=1)
parser.add_argument('--seed', type=int, default=0)
parser.add_argument('--path_to_key', type=str, default='./keys.json')
parser.add_argument('--data', type=str, default='mixed_100')
args = parser.parse_args()
output_dir = os.path.join('results/')
os.path.exists(output_dir) or os.makedirs(output_dir)
args.output_dir = output_dir
args.name = f'{args.data}_{args.eval}_{args.seed}_{args.model}'
if args.api:
args.name += '_api'
if args.cot:
args.name += '_cot'
if args.solution_demonstration:
args.name += '_soldemo'
data_path = './dataset/'
queries = json.load(open(os.path.join(data_path, f'{args.data}.json')))
args.work_dir = os.path.join('workdir/', f'{args.name}_workdir')
os.path.exists(args.work_dir) or os.makedirs(args.work_dir)
logs = main(args, queries)
json.dump(logs, open(os.path.join(output_dir, f'{args.name}.json'), 'w'), indent=4)