File size: 6,550 Bytes
8ec41f0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0164d1b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8ec41f0
 
 
0439a5a
8ec41f0
 
551cd0e
8ec41f0
 
 
 
 
 
 
dd1fecf
8ec41f0
 
 
 
 
 
0164d1b
 
 
8ec41f0
 
 
0164d1b
8ec41f0
0164d1b
 
8ec41f0
0164d1b
8ec41f0
0164d1b
8ec41f0
dc0dba4
 
 
 
 
8ec41f0
 
 
 
 
 
0164d1b
8ec41f0
551cd0e
 
8ec41f0
551cd0e
8ec41f0
0164d1b
8ec41f0
 
 
 
 
 
dc0dba4
0164d1b
8ec41f0
 
 
 
 
 
dc0dba4
0164d1b
8ec41f0
0164d1b
 
 
 
 
 
 
 
 
8ec41f0
 
0164d1b
 
8ec41f0
 
 
0164d1b
 
8ec41f0
0164d1b
 
8ec41f0
0164d1b
 
 
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
import os
import re
import pickle
import base64
import requests
import argparse
import numpy as np
import gradio as gr

from functools import partial
from PIL import Image

SERVER_URL = os.getenv('SERVER_URL')


def get_images(state):
    history = ''
    for i in range(len(state)):
        for j in range(len(state[i])):
            history += state[i][j] + '\n'
    for image_path in re.findall('image/[0-9,a-z]+\.png', history):
        if os.path.exists(image_path):
            continue
        data = {'method': 'get_image', 'args': [image_path], 'kwargs': {}}
        data = base64.b64encode(pickle.dumps(data)).decode('utf-8')
        response = requests.post(SERVER_URL, json=data)
        image = pickle.loads(base64.b64decode(response.json().encode('utf-8')))
        image.save(image_path)


def bot_request(method, *args, **kwargs):
    data = {'method': method, 'args': args, 'kwargs': kwargs}
    data = base64.b64encode(pickle.dumps(data)).decode('utf-8')
    response = requests.post(SERVER_URL, json=data)
    response = pickle.loads(base64.b64decode(response.json().encode('utf-8')))
    if response is not None:
        state = response[0]
        get_images(state)
    return response


def run_image(image, *args, **kwargs):
    if image is not None:
        width, height = image.size
        ratio = min(512 / width, 512 / height)
        width_new, height_new = (round(width * ratio), round(height * ratio))
        width_new = int(np.round(width_new / 64.0)) * 64
        height_new = int(np.round(height_new / 64.0)) * 64
        image = image.resize((width_new, height_new))
        image = image.convert('RGB')
    return bot_request('run_image', image, *args, **kwargs)


def predict_example(temperature, top_p, max_new_token, keep_last_n_paragraphs, image, text):
    state = []
    buffer = ''
    chatbot, state, text, buffer = run_image(image, state, text, buffer)
    chatbot, state, text, buffer = bot_request(
        'run_text', text, state, temperature, top_p,
        max_new_token, keep_last_n_paragraphs, buffer)
    return chatbot, state, text, None, buffer


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--temperature', type=float, default=0.0, help='temperature for the llm model')
    parser.add_argument('--max_new_tokens', type=int, default=256, help='max number of new tokens to generate')
    parser.add_argument('--top_p', type=float, default=1.0, help='top_p for the llm model')
    parser.add_argument('--top_k', type=int, default=40, help='top_k for the llm model')
    parser.add_argument('--keep_last_n_paragraphs', type=int, default=0, help='keep last n paragraphs in the memory')
    args = parser.parse_args()

    examples = [
        ['images/example-1.jpg', 'What is unusual about this image?'],
        ['images/example-2.jpg', 'Make the image look like a cartoon.'],
        ['images/example-3.jpg', 'Segment the tie in the image.'],
        ['images/example-4.jpg', 'Generate a man watching a sea based on the pose of the woman.'],
        ['images/example-5.jpg', 'Replace the dog with a monkey.'],
    ]

    if not os.path.exists('image'):
        os.makedirs('image')

    with gr.Blocks() as demo:
        state = gr.Chatbot([], visible=False)
        buffer = gr.Textbox('', visible=False)

        with gr.Row():
            with gr.Column(scale=0.3):
                with gr.Row():
                    image = gr.Image(type='pil', label='input image')
                with gr.Row():
                    txt = gr.Textbox(lines=7, show_label=False, elem_id='textbox',
                                     placeholder='Enter text and press submit, or upload an image').style(container=False)
                with gr.Row():
                    submit = gr.Button('Submit')
                with gr.Row():
                    clear = gr.Button('Clear')
                with gr.Row():
                    llm_name = gr.Radio(
                        ["Vicuna-13B"],
                        label="LLM Backend",
                        value="Vicuna-13B",
                        interactive=True)
                    keep_last_n_paragraphs = gr.Slider(
                        minimum=0,
                        maximum=3,
                        value=args.keep_last_n_paragraphs,
                        step=1,
                        interactive=True,
                        label='Remember Last N Paragraphs')
                    max_new_token = gr.Slider(
                        minimum=64,
                        maximum=512,
                        value=args.max_new_tokens,
                        step=1,
                        interactive=True,
                        label='Max New Tokens')
                    temperature = gr.Slider(
                        minimum=0.0,
                        maximum=1.0,
                        value=args.temperature,
                        step=0.1,
                        interactive=True,
                        visible=False,
                        label='Temperature')
                    top_p = gr.Slider(
                        minimum=0.0,
                        maximum=1.0,
                        value=args.top_p,
                        step=0.1,
                        interactive=True,
                        visible=False,
                        label='Top P')
            with gr.Column(scale=0.7):
                chatbot = gr.Chatbot(elem_id='chatbot', label='🦙 GPT4Tools').style(height=690)

            image.upload(lambda: '', None, txt)
            submit.click(run_image,
                         [image, state, txt, buffer],
                         [chatbot, state, txt, buffer]).then(
                partial(bot_request, 'run_text'),
                        [txt, state, temperature, top_p, max_new_token, keep_last_n_paragraphs, buffer],
                        [chatbot, state, txt, buffer]).then(
                    lambda: None, None, image)
            clear.click(partial(bot_request, 'clear'))
            clear.click(lambda: [[], [], '', ''], None, [chatbot, state, txt, buffer])

        with gr.Row():
            gr.Examples(
                examples=examples,
                fn=partial(predict_example, args.temperature, args.top_p,
                           args.max_new_tokens, args.keep_last_n_paragraphs),
                inputs=[image, txt],
                outputs=[chatbot, state, txt, image, buffer],
                cache_examples=True,
            )

        demo.queue(concurrency_count=6)
        demo.launch()