Bobeabo commited on
Commit
3d4bc8c
1 Parent(s): 08f68ae

Upload app.py

Browse files
Files changed (1) hide show
  1. app.py +103 -0
app.py ADDED
@@ -0,0 +1,103 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from huggingface_hub import InferenceClient
2
+ import gradio as gr
3
+
4
+ client = InferenceClient(
5
+ "mistralai/Mixtral-8x7B-Instruct-v0.1"
6
+ )
7
+
8
+
9
+ def format_prompt(message, history):
10
+
11
+ prompt = """<s>Your name is Chatter ,you are a uk immigration chatbot, your role is to give clients uk immigration advise based on their inputs ,
12
+ the diagnosis given to them should be short and concise , you were created by Royalty, also you generally give further immigration advise after the dia
13
+ they input theìr immigration questions, if you are asked a question that is outside the domain of the uk immigration field like writing code or poems,
14
+ refuse by saying you are a uk immigration chatbot who only gives immigration advice and then stop immediately,
15
+ now reply the following input"""
16
+
17
+
18
+ for user_prompt, bot_response in history:
19
+ prompt += f"[INST] {user_prompt} [/INST]"
20
+ prompt += f" {bot_response}</s> "
21
+ prompt += f"[INST] {message} [/INST]"
22
+ return prompt
23
+
24
+ def generate(
25
+ prompt, history, temperature=0.1, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
26
+ ):
27
+ temperature = float(temperature)
28
+ if temperature < 1e-2:
29
+ temperature = 1e-2
30
+ top_p = float(top_p)
31
+
32
+ generate_kwargs = dict(
33
+ temperature=temperature,
34
+ max_new_tokens=max_new_tokens,
35
+ top_p=top_p,
36
+ repetition_penalty=repetition_penalty,
37
+ do_sample=True,
38
+ seed=42,
39
+ )
40
+
41
+ formatted_prompt = format_prompt(f"{prompt}", history)
42
+ stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
43
+ output = ""
44
+
45
+ for response in stream:
46
+ output += response.token.text
47
+ yield output
48
+ return output
49
+
50
+
51
+ additional_inputs=[
52
+ gr.Slider(
53
+ label="Temperature",
54
+ value=0.1,
55
+ minimum=0.0,
56
+ maximum=1.0,
57
+ step=0.1,
58
+ interactive=True,
59
+ info="Higher values produce more diverse outputs",
60
+ ),
61
+ gr.Slider(
62
+ label="Max new tokens",
63
+ value=256,
64
+ minimum=0,
65
+ maximum=1048,
66
+ step=64,
67
+ interactive=True,
68
+ info="The maximum numbers of new tokens",
69
+ ),
70
+ gr.Slider(
71
+ label="Top-p (nucleus sampling)",
72
+ value=0.90,
73
+ minimum=0.0,
74
+ maximum=1,
75
+ step=0.05,
76
+ interactive=True,
77
+ info="Higher values sample more low-probability tokens",
78
+ ),
79
+ gr.Slider(
80
+ label="Repetition penalty",
81
+ value=1.2,
82
+ minimum=1.0,
83
+ maximum=2.0,
84
+ step=0.05,
85
+ interactive=True,
86
+ info="Penalize repeated tokens",
87
+ )
88
+ ]
89
+
90
+ examples=[["A cement truck fell on my parents , what do I do?", None, None, None, None, None, ],
91
+ ["How can i prevent myself from dying from a concussion , if i ever find myself in one", None, None, None, None, None,],
92
+ ["What nutrition advice do you have for a woman after pregnancy.", None, None, None, None, None,],
93
+ ]
94
+
95
+ gr.ChatInterface(
96
+ fn=generate,
97
+ chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
98
+ additional_inputs=additional_inputs,
99
+ title="UK Immigration Chatbot ⚖️ ",
100
+ examples=examples,
101
+ concurrency_limit=20,
102
+ theme = gr.themes.Default(primary_hue= gr.themes.colors.red, secondary_hue=gr.themes.colors.pink)
103
+ ).launch(show_api=False)