File size: 7,459 Bytes
86d1d69
 
 
 
c04f5fc
86d1d69
 
 
 
 
 
 
 
052595e
c04f5fc
86d1d69
 
 
 
 
 
 
 
 
 
dbc96cf
86d1d69
 
 
 
 
 
 
 
8757e5a
6f11ed6
86d1d69
5bb345f
86d1d69
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6f11ed6
bdc902b
f275cee
5171ca7
 
 
 
 
 
f275cee
 
5171ca7
 
97c89d2
86d1d69
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
12fe850
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
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
import torch
from PIL import Image
import gradio as gr
import spaces
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
import os
from threading import Thread

import pymupdf
import docx
from pptx import Presentation


MODEL_LIST = ["nikravan/glm-4vq"]

HF_TOKEN = os.environ.get("HF_TOKEN", None)
MODEL_ID = MODEL_LIST[0]
MODEL_NAME = "GLM-4vq"

TITLE = "<h1>3ML-bot</h1>"

DESCRIPTION = f"""
<center>
<p>😊 A Multi-Modal Multi-Lingual(3ML) Chat. 
<br>
🚀 MODEL NOW: <a href="https://hf.co/nikravan/glm-4vq">{MODEL_NAME}</a>
</center>"""

CSS = """
h1 {
    text-align: center;
    display: block;
}
"""


tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)



def extract_text(path):
    return open(path, 'r').read()


def extract_pdf(path):
    doc = pymupdf.open(path)
    text = ""
    for page in doc:
        text += page.get_text()
    return text


def extract_docx(path):
    doc = docx.Document(path)
    data = []
    for paragraph in doc.paragraphs:
        data.append(paragraph.text)
    content = '\n\n'.join(data)
    return content


def extract_pptx(path):
    prs = Presentation(path)
    text = ""
    for slide in prs.slides:
        for shape in slide.shapes:
            if hasattr(shape, "text"):
                text += shape.text + "\n"
    return text


def mode_load(path):
    choice = ""
    file_type = path.split(".")[-1]
    print(file_type)
    if file_type in ["pdf", "txt", "py", "docx", "pptx", "json", "cpp", "md"]:
        if file_type.endswith("pdf"):
            content = extract_pdf(path)
        elif file_type.endswith("docx"):
            content = extract_docx(path)
        elif file_type.endswith("pptx"):
            content = extract_pptx(path)
        else:
            content = extract_text(path)
        choice = "doc"
        print(content[:100])
        return choice, content[:5000]


    elif file_type in ["png", "jpg", "jpeg", "bmp", "tiff", "webp"]:
        content = Image.open(path).convert('RGB')
        choice = "image"
        return choice, content

    else:
        raise gr.Error("Oops, unsupported files.")


@spaces.GPU()
flag=False   
def stream_chat(message, history: list, temperature: float, max_length: int, top_p: float, top_k: int, penalty: float):
    if flag==False:
        model = AutoModelForCausalLM.from_pretrained(
            MODEL_ID,
            torch_dtype=torch.bfloat16,
            low_cpu_mem_usage=True,
            trust_remote_code=True
        )
        model.eval()
        flag=True
    
        
    
    print(f'message is - {message}')
    print(f'history is - {history}')
    conversation = []
    prompt_files = []
    if message["files"]:
        choice, contents = mode_load(message["files"][-1])
        if choice == "image":
            conversation.append({"role": "user", "image": contents, "content": message['text']})
        elif choice == "doc":
            format_msg = contents + "\n\n\n" + "{} files uploaded.\n" + message['text']
            conversation.append({"role": "user", "content": format_msg})
    else:
        if len(history) == 0:
            # raise gr.Error("Please upload an image first.")
            contents = None
            conversation.append({"role": "user", "content": message['text']})
        else:
            # image = Image.open(history[0][0][0])
            for prompt, answer in history:
                if answer is None:
                    prompt_files.append(prompt[0])
                    conversation.extend([{"role": "user", "content": ""}, {"role": "assistant", "content": ""}])
                else:
                    conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
            if len(prompt_files) > 0:
                choice, contents = mode_load(prompt_files[-1])
            else:
                choice = ""
                conversation.append({"role": "user", "image": "", "content": message['text']})


            if choice == "image":
                conversation.append({"role": "user", "image": contents, "content": message['text']})
            elif choice == "doc":
                format_msg = contents + "\n\n\n" + "{} files uploaded.\n" + message['text']
                conversation.append({"role": "user", "content": format_msg})
    print(f"Conversation is -\n{conversation}")

    input_ids = tokenizer.apply_chat_template(conversation, tokenize=True, add_generation_prompt=True,
                                              return_tensors="pt", return_dict=True).to(model.device)
    streamer = TextIteratorStreamer(tokenizer, timeout=60.0, skip_prompt=True, skip_special_tokens=True)

    generate_kwargs = dict(
        max_length=max_length,
        streamer=streamer,
        do_sample=True,
        top_p=top_p,
        top_k=top_k,
        temperature=temperature,
        repetition_penalty=penalty,
        eos_token_id=[151329, 151336, 151338],
    )
    gen_kwargs = {**input_ids, **generate_kwargs}

    with torch.no_grad():
        thread = Thread(target=model.generate, kwargs=gen_kwargs)
        thread.start()
        buffer = ""
        for new_text in streamer:
            buffer += new_text
            yield buffer


chatbot = gr.Chatbot(
    #rtl=True,
)
chat_input = gr.MultimodalTextbox(
    interactive=True,
    placeholder="Enter message or upload a file ...",
    show_label=False,
    #rtl=True,



)
EXAMPLES = [
    [{"text": "Write a poem about spring season in French Language", }],
    [{"text": "what does this chart mean?", "files": ["sales.png"]}],
    [{"text": "¿Qué está escrito a mano en esta foto?", "files": ["receipt1.png"]}],
    [{"text": "در مورد این عکس توضیح بده و بگو این چه فصلی می تواند باشد", "files": ["nature.jpg"]}]
]

with gr.Blocks(css=CSS, theme="soft", fill_height=True) as demo:
    gr.HTML(TITLE)
    gr.HTML(DESCRIPTION)
    gr.ChatInterface(
        fn=stream_chat,
        multimodal=True,


        textbox=chat_input,
        chatbot=chatbot,
        fill_height=True,
        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
        additional_inputs=[
            gr.Slider(
                minimum=0,
                maximum=1,
                step=0.1,
                value=0.8,
                label="Temperature",
                render=False,
            ),
            gr.Slider(
                minimum=1024,
                maximum=8192,
                step=1,
                value=4096,
                label="Max Length",
                render=False,
            ),
            gr.Slider(
                minimum=0.0,
                maximum=1.0,
                step=0.1,
                value=1.0,
                label="top_p",
                render=False,
            ),
            gr.Slider(
                minimum=1,
                maximum=20,
                step=1,
                value=10,
                label="top_k",
                render=False,
            ),
            gr.Slider(
                minimum=0.0,
                maximum=2.0,
                step=0.1,
                value=1.0,
                label="Repetition penalty",
                render=False,
            ),
        ],
    ),
    gr.Examples(EXAMPLES, [chat_input])

if __name__ == "__main__":
    demo.queue(api_open=False).launch(show_api=False, share=False, )#server_name="0.0.0.0", )