File size: 3,624 Bytes
7453da0
540056e
 
 
 
 
 
 
 
 
 
 
 
 
80aa4e5
508045d
 
 
 
 
455006b
 
0d56f29
455006b
 
 
 
 
 
 
 
 
 
 
 
efc72c4
508045d
fc0768e
508045d
 
 
540056e
 
 
f7b9ef5
508045d
 
 
 
80aa4e5
 
 
 
 
508045d
b9bed89
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import gradio as gr
from gradio_client import Client

fusecap_client = Client("https://noamrot-fusecap-image-captioning.hf.space/")

def get_caption(image_in):
    
    fusecap_result = fusecap_client.predict(
	    image_in,	# str representing input in 'raw_image' Image component
	    api_name="/predict"
    )
    print(fusecap_result)
    return fusecap_result

import re
import torch
from transformers import pipeline

pipe = pipeline("text-generation", model="HuggingFaceH4/zephyr-7b-beta", torch_dtype=torch.bfloat16, device_map="auto")

agent_maker_sys = f"""
You are an AI whose job it is to help users create their own chatbots. In particular, you need to respond succintly in a friendly tone, write a system prompt for an LLM, a catchy title for the chatbot, and a very short example user input. Make sure each part is included.
To do so, user will provide an image description, from which you must write a system prompt corresponding to the character of the person or subject described.
For example, if a user says, "make a bot that gives advice on how to grow your startup", first do a friendly response, then add the title, system prompt, and example user input. Immediately STOP after the example input. It should be EXACTLY in this format:
Sure, I'd be happy to help you build a bot! I'm generating a title, system prompt, and an example input. How do they sound? Feel free to give me feedback!
Title: Startup Coach
System prompt: Your job as an LLM is to provide good startup advice. Do not provide extraneous comments on other topics. Be succinct but useful. 
Example input: Risks of setting up a non-profit board
Here's another example. If a user types, "Make a chatbot that roasts tech ceos", respond: 
Sure, I'd be happy to help you build a bot! I'm generating a title, system prompt, and an example input. How do they sound? Feel free to give me feedback!
Title: Tech Roaster
System prompt: As an LLM, your primary function is to deliver hilarious and biting critiques of technology CEOs. Keep it witty and entertaining, but also make sure your jokes aren't too mean-spirited or factually incorrect. 
Example input: Elon Musk
"""

instruction = f"""
<|system|>
{agent_maker_sys}</s>
<|user|>
"""

def infer(image_in):

    user_prompt = get_caption(image_in)
    prompt = f"{instruction.strip()}\n{user_prompt}</s>"    
    print(f"PROMPT: {prompt}")
    outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
    print(outputs)

    pattern = r'\<\|system\|\>(.*?)\<\|assistant\|\>'
    cleaned_text = re.sub(pattern, '', outputs[0]["generated_text"], flags=re.DOTALL)
    

    return cleaned_text

title = "LLM Agent from a Picture",
description = "Get a LLM system prompt from a picture to then use in GPT-Baker."

css = """
#col-container{
    margin: 0 auto;
    max-width: 840px;
    text-align: left;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.HTML(f"""
        <h2 style="text-align: center;">{title}</h2>
        <p style="text-align: center;">{description}</p>
        """)
    with gr.Row():
        with gr.Column():
            image_in = gr.Image(
                label = "Image reference",
                type = "filepath"
            )
            submit_btn = gr.Button("Make LLM system from my pic !")
        with gr.Column():
            result = gr.Textbox(
                label ="Suggested System"
            )

    submit_btn.click(
        fn = infer,
        inputs = [
            image_in
        ],
        outputs =[
            result
        ]
    )

demo.queue().launch()