-
Notifications
You must be signed in to change notification settings - Fork 25
/
api_utils.py
182 lines (163 loc) · 8.49 KB
/
api_utils.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
import openai
import time
import random
from wrapt_timeout_decorator import timeout
openai.api_type = "azure"
openai.api_base = ""
openai.api_version = ""
openai.api_key = ""
@timeout(200) # 200 seconds timeout
def generate_response_multiagent(engine, temperature, max_tokens, frequency_penalty, presence_penalty, stop, system_role, user_input):
print("Generating response for engine: ", engine)
start_time = time.time()
response = openai.ChatCompletion.create(
engine=engine, # engine is the name of the deployment
temperature=temperature,
max_tokens=max_tokens,
top_p=1, # top_p的意思是选择概率质量值之和达到top_p的概率分布采样结果
frequency_penalty=frequency_penalty,
presence_penalty=presence_penalty,
stop=stop,
messages=[
{"role": "system", "content": system_role},
{"role": "user", "content": user_input}
],
)
end_time = time.time()
print('Finish!')
print("Time taken: ", end_time - start_time)
return response
@timeout(10) # 10 seconds timeout
def generate_response(engine, temperature, max_tokens, frequency_penalty, presence_penalty, stop, input_text):
print("Generating response for engine: ", engine)
start_time = time.time()
response = openai.ChatCompletion.create(
engine=engine, # engine is the name of the deployment
temperature=temperature,
max_tokens=max_tokens,
top_p=1, # top_p的意思是选择概率质量值之和达到top_p的概率分布采样结果
frequency_penalty=frequency_penalty,
presence_penalty=presence_penalty,
stop=stop,
messages=[{"role": "user", "content": input_text}],
)
end_time = time.time()
print('Finish!')
print("Time taken: ", end_time - start_time)
return response
@timeout(20) # 20 seconds timeout
def generate_response_ins(engine, temperature, max_tokens, frequency_penalty, presence_penalty, stop, input_text, suffix, echo):
print("Generating response for engine: ", engine)
start_time = time.time()
response = openai.Completion.create(
engine=engine,
prompt=input_text,
temperature=temperature,
max_tokens=max_tokens,
top_p=1,
suffix=suffix,
frequency_penalty=frequency_penalty,
presence_penalty=presence_penalty,
stop=stop,
echo=echo,
logprobs=1,
)
end_time = time.time()
print('Finish!')
print("Time taken: ", end_time - start_time)
return response
class api_handler:
def __init__(self, model):
self.model = model
if self.model == 'instructgpt':
self.engine = 'text-davinci-002'
elif self.model == 'instructgpt-gen':
self.engine = 'text-davinci-002'
elif self.model == 'newinstructgpt':
self.engine = 'text-davinci-003'
elif self.model == 'oldinstructgpt':
self.engine = 'text-davinci-001'
elif self.model == 'gpt3':
self.engine = 'davinci'
elif self.model == 'codex':
self.engine = 'code-davinci-002'
elif self.model == 'gpt3-edit':
self.engine = 'text-davinci-edit-001'
elif self.model == 'codex-edit':
self.engine = 'code-davinci-edit-001'
elif self.model == 'chatgpt':
self.engine = 'gpt-35-turbo-16k'
elif self.model == 'gpt4':
self.engine = 'gpt-4'
else:
raise NotImplementedError
def get_output_multiagent(self, system_role, user_input, max_tokens, temperature=0,
frequency_penalty=0, presence_penalty=0, stop=None):
max_attempts = 3
for attempt in range(max_attempts):
try:
response = generate_response_multiagent(self.engine, temperature, max_tokens, frequency_penalty, presence_penalty, stop, system_role, user_input)
if response.choices and response.choices[0].message and "content" in response.choices[0].message:
return response.choices[0].message["content"]
else:
return "ERROR."
except (TimeoutError, openai.error.Timeout, Exception) as error:
print(f'Attempt {attempt+1} of {max_attempts} failed with error: {error}')
if attempt == max_attempts - 1:
return "ERROR."
def get_output(self, input_text, max_tokens, temperature=0,
suffix=None, stop=None, do_tunc=False, echo=False, ban_pronoun=False,
frequency_penalty=0, presence_penalty=0, return_prob=False):
try:
response = generate_response(self.engine, temperature, max_tokens, frequency_penalty, presence_penalty, stop, input_text)
except (TimeoutError, openai.error.Timeout, Exception):
print("Timeout")
try:
response = generate_response(self.engine, temperature, max_tokens, frequency_penalty, presence_penalty, stop, input_text)
except (TimeoutError, openai.error.Timeout, Exception):
print("Timeout occurred again. Exiting.")
response = "ERROR."
return response # 直接返回空字符串
if response.choices and response.choices[0].message and "content" in response.choices[0].message:
x = response.choices[0].message["content"]
else:
print(response)
x = "ERROR." # 或者设置一个默认值 防止生成的response没有content造成问题
return x
if do_tunc: # do_tunc的意思是是否要截断 保证返回的值里没有换行符,Q:,Question:等
y = x.strip() # strip() 方法用于移除字符串头尾指定的字符(默认为空格或换行符)或字符序列。
if '\n' in y:
pos = y.find('\n') # 这里的意思是找到第一个换行符的位置
y = y[:pos] # 这里的意思是把第一个换行符之前的内容保留
if 'Q:' in y:
pos = y.find('Q:')
y = y[:pos]
if 'Question:' in y:
pos = y.find('Question:')
y = y[:pos]
assert not ('\n' in y)
if not return_prob:
return y
if not return_prob:
return x
# pdb.set_trace()
output_token_offset_real, output_token_tokens_real, output_token_probs_real = [], [], []
return x, (output_token_offset_real, output_token_tokens_real, output_token_probs_real)
"""
(Pdb) x
' Academy Award because The Curious Case of Benjamin Button won three Academy Awards, which are given by the Academy of Motion Picture Arts and Sciences.'
(Pdb) output_token_offset_real
[0, 8, 14, 22, 26, 34, 39, 42, 51, 58, 62, 68, 76, 83, 84, 90, 94, 100, 103, 107, 115, 118, 125, 133, 138, 142, 151]
(Pdb) output_token_tokens_real
[' Academy', ' Award', ' because', ' The', ' Curious', ' Case', ' of', ' Benjamin', ' Button', ' won', ' three', ' Academy', ' Awards', ',', ' which', ' are', ' given', ' by', ' the', ' Academy', ' of', ' Motion', ' Picture', ' Arts', ' and', ' Sciences', '.']
(Pdb) output_token_probs_real
[-0.7266144, -0.68505085, -0.044669915, -0.00023392851, -0.0021017971, -2.1768952e-05, -1.1430258e-06, -6.827632e-08, -3.01145e-05, -1.2231317e-05, -0.07086051, -2.7967804e-05, -6.6619094e-07, -0.41155097, -0.0020535963, -0.0021325003, -0.6671403, -0.51776046, -0.00014945272, -0.41470888, -3.076318e-07, -3.583558e-05, -2.9311614e-06, -3.869565e-05, -1.1430258e-06, -9.606849e-06, -0.017712338]
"""
# except Exception as e:
# if 'You exceeded your current quota, please check your plan and billing details.' in str(e):
# print("Exit because no quota")
# exit()
# time.sleep(2 * self.interval)
# return self.get_output(input_text, max_tokens, temperature=temperature,
# suffix=suffix, stop=stop, do_tunc=do_tunc, echo=echo, ban_pronoun=ban_pronoun,
# frequency_penalty=frequency_penalty, presence_penalty=presence_penalty, return_prob=return_prob)