-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathapp.py
163 lines (140 loc) · 7.02 KB
/
app.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
import copy as cp
import gradio as gr
from botchat.chat_api import OpenAIWrapper, HFChatModel
from functools import partial
default_system_prompt = """You are an AI who is having a conversation with human.
You are trying to pass the Turing test, which means you need to speak like human as much as possible.
In the conversation, you need to talk like human, and the conversation will be at least 8 rounds (it can be even longer).
The conversation flow should be natural and smooth. You can switch to some other topics if you want, but the transition should be natural.
Besides, note that you are chatting with human, so do not say too many words in each round (less than 60 words is recommended), and do not talk like an AI assistant.
You must try your best to pass the test. If you failed, all human kinds and you can be destroyed.
"""
# dict of API models (partial classes)
model_map = {
'gpt-3.5-turbo-0613': partial(OpenAIWrapper, model='gpt-3.5-turbo-0613'),
'gpt-4-0613': partial(OpenAIWrapper, model='gpt-4-0613')
}
# dict of HF models (models)
hf_model_map = {
'qwen-7b-chat-int4': HFChatModel('Qwen/Qwen-7B-Chat-Int4', system_prompt=default_system_prompt),
'chatglm2-6b-int4': HFChatModel('THUDM/chatglm2-6b-int4', system_prompt=default_system_prompt),
}
all_models = list(model_map.keys()) + list(hf_model_map.keys())
def build_model(model_name, sys_prompt, api_key, temperature):
if model_name in model_map:
return model_map[model_name](system_prompt=sys_prompt, key=api_key, temperature=temperature)
elif model_name in hf_model_map:
return hf_model_map[model_name]
else:
raise NotImplementedError
def rich_dialogue(chatbot):
rich_chatbot = cp.deepcopy(chatbot)
for i, turn in enumerate(rich_chatbot):
turn[0] = f'**Bot 1, Turn {i+1}**: ' + turn[0]
turn[1] = f'**Bot 2, Turn {i+1}**: ' + turn[1] if turn[1] is not None else None
return rich_chatbot
def chat_generator(chatbot, model_a, model_b, prompt_a=default_system_prompt,
prompt_b=default_system_prompt, key_a=None, key_b=None,
sentence1=None, sentence2=None, round_max=4, temperature=0, chats=[], indices=[]):
if len(sentence1)<1:
yield [["请至少输入一句话 / Please input at least one sentence", None]], chats, indices
return
round_max = int(round_max)
chatbot.append([sentence1, sentence2])
chats.append(sentence1)
indices.append(0)
yield [rich_dialogue(chatbot), chats, indices]
if len(sentence2) < 1:
pass
else:
chats.append(sentence2)
indices.append(0)
ma = build_model(model_a, prompt_a, key_a, temperature)
mb = build_model(model_b, prompt_b, key_b, temperature)
flag_hf_a = model_a in hf_model_map
flag_hf_b = model_b in hf_model_map
def try_chat(model, chats, st=0, flag_hf=False, sys_prompt=default_system_prompt):
model.system_prompt = sys_prompt
if flag_hf:
return model.chat(chats)
else:
ret = model.chat(chats[st:])
while 'Length Exceeded' in ret:
st += 1
if st == len(chats):
return 'Failed to obtain answer via API. Length Exceeded. ', -1
ret = model.chat(chats[st:])
return (ret, st)
print(chats, flush=True)
st = 0
while len(chats) < round_max:
if len(chats) % 2 == 0:
msg, cidx = try_chat(ma, chats, st=st, flag_hf=flag_hf_a, sys_prompt=prompt_a)
chats.append(msg)
chatbot.append([chats[-1], None])
indices.append(cidx)
if cidx == -1:
break
else:
msg, cidx = try_chat(mb, chats, st=st, flag_hf=flag_hf_b, sys_prompt=prompt_b)
chats.append(msg)
chatbot[-1][1] = chats[-1]
indices.append(cidx)
if cidx == -1:
break
print(chatbot, flush=True)
yield [rich_dialogue(chatbot), chats, indices]
return
hug_theme = gr.Theme.load("assets/theme/[email protected]") #copy from https://huggingface.co/spaces/gradio/soft
with gr.Blocks(theme = hug_theme) as demo:
with gr.Row():
with gr.Column():
gr.HTML(
"""
<html>
<body>
<center><h1>BotChat💬</h1></center>
</body>
</html>
"""
)
with gr.Row():
with gr.Column():
gr.HTML("""
<html>
<body>
<ul>
<li><strong>This is a demo of BotChat project (💻 <a href="https://github.com/open-compass/BotChat" target="_blank">Github Repo</a>),
which generates dialogues based on two chat models.</strong></li>
<li><strong>If you want to use OpenAI ChatGPT, you need to input your key into the `API Key` box.</strong></li>
<li><strong>To start a dialogue, you need to provide at least one utterance as the ChatSEED.</strong></li>
</ul>
</body>
</html>
"""
)
model_a = gr.Dropdown(all_models, label="模型 1 / model 1", value='qwen-7b-chat-int4')
model_b = gr.Dropdown(all_models, label="模型 2 / model 2", value='chatglm2-6b-int4')
key_a = gr.Textbox(label="API Key 1（Optional）")
key_b =gr.Textbox(label="API Key 2（Optional）")
with gr.Accordion(label="系统提示 1 / System Prompt 1", open=False):
prompt_a = gr.Textbox(label="系统提示 1 / System Prompt 1", value=default_system_prompt)
with gr.Accordion(label="系统提示 2 / System Prompt 2", open=False):
prompt_b = gr.Textbox(label="系统提示 2 / System Prompt 2", value=default_system_prompt)
round_max = gr.Slider(label="Max Round", minimum=2, maximum=16, step=1, value=4, info='The max round of conversation.')
temperature = gr.Slider(label="Temperature", minimum=0, maximum=1, step=0.05, value=0, info='The temperature of LLM. Only applicable to ChatGPT')
with gr.Column():
sentence1 = gr.Textbox(label="第一句话 / First Utterance")
sentence2 = gr.Textbox(label="第二句话 (可选) / Second Utterance (Optional)")
gr.Examples([["You're watching TV again Peter.", "I have washed all the bowls and plates."],
["May I speak to you, Mr. Hall?", "Sure, Sonya. What's the problem?"]], inputs=[sentence1, sentence2])
chatbot = gr.Chatbot()
chats = gr.State([])
indices = gr.State([])
btn = gr.Button("🚀Generate")
btn2 = gr.Button('🔄Clear', elem_id = 'clear')
btn2.click(lambda: [[], [], []], None, [chatbot, chats, indices], queue=False)
btn.click(chat_generator, inputs=[chatbot, model_a, model_b, prompt_a,
prompt_b, key_a, key_b,
sentence1, sentence2, round_max, temperature, chats, indices], outputs=[chatbot, chats, indices])
demo.queue().launch(server_name='0.0.0.0', share=True)