File size: 7,949 Bytes
d711bd7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
from transformers import AutoModel, AutoTokenizer
import os
import ipdb
import gradio as gr
import mdtex2html
from model.openllama import OpenLLAMAPEFTModel
import torch
import json
from header import TaskType, LoraConfig

# init the model
args = {
    'model': 'openllama_peft',
    'imagebind_ckpt_path': 'pretrained_ckpt/imagebind_ckpt',
    'vicuna_ckpt_path': 'openllmplayground/vicuna_7b_v0',
    'delta_ckpt_path': 'pretrained_ckpt/pandagpt_ckpt/7b/pytorch_model.pt',
    'stage': 2,
    'max_tgt_len': 128,
    'lora_r': 32,
    'lora_alpha': 32,
    'lora_dropout': 0.1,
}
model = OpenLLAMAPEFTModel(**args)
delta_ckpt = torch.load(args['delta_ckpt_path'], map_location=torch.device('cpu'))
model.load_state_dict(delta_ckpt, strict=False)
model = model.half().cuda().eval() if torch.cuda.is_available() else model.eval()
print(f'[!] init the model over ...')


"""Override Chatbot.postprocess"""


def postprocess(self, y):
    if y is None:
        return []
    for i, (message, response) in enumerate(y):
        y[i] = (
            None if message is None else mdtex2html.convert((message)),
            None if response is None else mdtex2html.convert(response),
        )
    return y


gr.Chatbot.postprocess = postprocess


def parse_text(text):
    """copy from https://github.com/GaiZhenbiao/ChuanhuChatGPT/"""
    lines = text.split("\n")
    lines = [line for line in lines if line != ""]
    count = 0
    for i, line in enumerate(lines):
        if "```" in line:
            count += 1
            items = line.split('`')
            if count % 2 == 1:
                lines[i] = f'<pre><code class="language-{items[-1]}">'
            else:
                lines[i] = f'<br></code></pre>'
        else:
            if i > 0:
                if count % 2 == 1:
                    line = line.replace("`", "\`")
                    line = line.replace("<", "&lt;")
                    line = line.replace(">", "&gt;")
                    line = line.replace(" ", "&nbsp;")
                    line = line.replace("*", "&ast;")
                    line = line.replace("_", "&lowbar;")
                    line = line.replace("-", "&#45;")
                    line = line.replace(".", "&#46;")
                    line = line.replace("!", "&#33;")
                    line = line.replace("(", "&#40;")
                    line = line.replace(")", "&#41;")
                    line = line.replace("$", "&#36;")
                lines[i] = "<br>"+line
    text = "".join(lines)
    return text


def predict(
    input, 
    image_path, 
    audio_path, 
    video_path, 
    thermal_path, 
    chatbot, 
    max_length, 
    top_p, 
    temperature, 
    history, 
    modality_cache, 
):
    if image_path is None and audio_path is None and video_path is None and thermal_path is None:
        return [(input, "There is no image/audio/video provided. Please upload the file to start a conversation.")]
    else:
        print(f'[!] image path: {image_path}\n[!] audio path: {audio_path}\n[!] video path: {video_path}\n[!] thermal pah: {thermal_path}')
    # prepare the prompt
    prompt_text = ''
    for idx, (q, a) in enumerate(history):
        if idx == 0:
            prompt_text += f'{q}\n### Assistant: {a}\n###'
        else:
            prompt_text += f' Human: {q}\n### Assistant: {a}\n###'
    if len(history) == 0:
        prompt_text += f'{input}'
    else:
        prompt_text += f' Human: {input}'

    response = model.generate({
        'prompt': prompt_text,
        'image_paths': [image_path] if image_path else [],
        'audio_paths': [audio_path] if audio_path else [],
        'video_paths': [video_path] if video_path else [],
        'thermal_paths': [thermal_path] if thermal_path else [],
        'top_p': top_p,
        'temperature': temperature,
        'max_tgt_len': max_length,
        'modality_embeds': modality_cache
    })
    chatbot.append((parse_text(input), parse_text(response)))
    history.append((input, response))
    return chatbot, history, modality_cache


def reset_user_input():
    return gr.update(value='')


def reset_state():
    return None, None, None, None, [], [], []


with gr.Blocks() as demo:
    gr.HTML("""<h1 align="center">PandaGPT</h1>""")
    gr.Markdown('''We note that the current online demo uses the 7B version of PandaGPT due to the limitation of computation resource. 

    Better results should be expected when switching to the 13B version of PandaGPT.

    For more details on how to run 13B PandaGPT, please refer to our [main project repository](https://github.com/yxuansu/PandaGPT).''')

    with gr.Row(scale=4):
        with gr.Column(scale=2):
            image_path = gr.Image(type="filepath", label="Image", value=None)

            gr.Examples(
                [ 
                    os.path.join(os.path.dirname(__file__), "assets/images/bird_image.jpg"),
                    os.path.join(os.path.dirname(__file__), "assets/images/dog_image.jpg"),
                    os.path.join(os.path.dirname(__file__), "assets/images/car_image.jpg"),
                ],
                image_path
            )
        with gr.Column(scale=2):
            audio_path = gr.Audio(type="filepath", label="Audio", value=None)
            gr.Examples(
                [ 
                    os.path.join(os.path.dirname(__file__), "assets/audios/bird_audio.wav"),
                    os.path.join(os.path.dirname(__file__), "assets/audios/dog_audio.wav"),
                    os.path.join(os.path.dirname(__file__), "assets/audios/car_audio.wav"),
                ],
                audio_path
            )
    with gr.Row(scale=4):
        with gr.Column(scale=2):
            video_path = gr.Video(type='file', label="Video")

            gr.Examples(
                [ 
                    os.path.join(os.path.dirname(__file__), "assets/videos/world.mp4"),
                    os.path.join(os.path.dirname(__file__), "assets/videos/a.mp4"),
                ],
                video_path
            )
        with gr.Column(scale=2):
            thermal_path = gr.Image(type="filepath", label="Thermal Image", value=None)

            gr.Examples(
                [ 
                    os.path.join(os.path.dirname(__file__), "assets/thermals/190662.jpg"),
                    os.path.join(os.path.dirname(__file__), "assets/thermals/210009.jpg"),
                ],
                thermal_path
            )

    chatbot = gr.Chatbot()
    with gr.Row():
        with gr.Column(scale=4):
            with gr.Column(scale=12):
                user_input = gr.Textbox(show_label=False, placeholder="Input...", lines=10).style(container=False)
            with gr.Column(min_width=32, scale=1):
                submitBtn = gr.Button("Submit", variant="primary")
        with gr.Column(scale=1):
            emptyBtn = gr.Button("Clear History")
            max_length = gr.Slider(0, 512, value=128, step=1.0, label="Maximum length", interactive=True)
            top_p = gr.Slider(0, 1, value=0.01, step=0.01, label="Top P", interactive=True)
            temperature = gr.Slider(0, 1, value=0.8, step=0.01, label="Temperature", interactive=True)

    history = gr.State([])
    modality_cache = gr.State([])

    submitBtn.click(
        predict, [
            user_input, 
            image_path, 
            audio_path, 
            video_path, 
            thermal_path, 
            chatbot, 
            max_length, 
            top_p, 
            temperature, 
            history, 
            modality_cache,
        ], [
            chatbot, 
            history,
            modality_cache
        ],
        show_progress=True
    )

    submitBtn.click(reset_user_input, [], [user_input])
    emptyBtn.click(reset_state, outputs=[
        image_path,
        audio_path,
        video_path,
        thermal_path,
        chatbot, 
        history, 
        modality_cache
    ], show_progress=True)


demo.launch(enable_queue=True)