File size: 5,836 Bytes
e9f296d
 
 
 
f69bb10
e9f296d
 
 
 
 
 
 
 
 
f69bb10
e9f296d
 
f69bb10
e9f296d
 
 
 
 
 
f69bb10
 
e9f296d
 
f69bb10
e9f296d
 
 
f69bb10
e9f296d
 
 
f69bb10
e9f296d
 
 
 
 
f69bb10
 
e9f296d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f69bb10
e9f296d
 
 
 
 
f69bb10
e9f296d
 
 
f69bb10
e9f296d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f69bb10
 
e9f296d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f69bb10
e9f296d
 
 
 
 
 
 
 
 
 
 
 
 
 
f69bb10
e9f296d
f69bb10
e9f296d
 
 
 
f69bb10
e9f296d
 
 
 
 
 
 
 
f69bb10
 
 
e9f296d
 
 
 
 
f69bb10
e9f296d
 
 
 
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
import os
from threading import Thread
from typing import Iterator

import gradio as gr
import spaces
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
import time

MAX_MAX_NEW_TOKENS = 2048
DEFAULT_MAX_NEW_TOKENS = 1024
MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))


DESCRIPTION = """\
# Dorna-Llama3-8B-Instruct Chat
"""

PLACEHOLDER = """
<div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">
   <img src="https://avatars.githubusercontent.com/u/39557177?v=4" style="width: 80%; max-width: 550px; height: auto; opacity: 0.80;  "> 
   <h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">Dorna-Llama3-8B-Instruct</h1>
</div>
"""

custom_css = """
@import url('https://fonts.googleapis.com/css2?family=Vazirmatn&display=swap');

body, .gradio-container, .gr-button, .gr-input, .gr-slider, .gr-dropdown, .gr-markdown {
    font-family: 'Vazirmatn', sans-serif !important;
}

._button {
    font-size: 20px;
}

pre, code {
    direction: ltr !important;
    unicode-bidi: plaintext !important;
}
"""


system_prompt = str(os.getenv("SYSTEM_PROMPT"))


def execution_time_calculator(start_time, log=True):
    delta = time.time() - start_time
    if log:
        print("--- %s seconds ---" % (delta))
    return delta

def token_per_second_calculator(tokens_count, time_delta):
    return tokens_count/time_delta

if not torch.cuda.is_available():
    DESCRIPTION = "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"


if torch.cuda.is_available():
    model_id = "PartAI/Dorna-Llama3-8B-Instruct"
    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
    tokenizer = AutoTokenizer.from_pretrained(model_id)
    
generation_speed = 0

def get_generation_speed():
    global generation_speed

    return generation_speed


@spaces.GPU
def generate(
    message: str,
    chat_history: list[tuple[str, str]],
    max_new_tokens: int = 1024,
    temperature: float = 0.6,
    top_p: float = 0.9,
    top_k: int = 50,
    repetition_penalty: float = 1.2,
    do_sample: bool =True,
) -> Iterator[str]:
    global generation_speed
    global system_prompt

    conversation = []
    if system_prompt:
        conversation.append({"role": "system", "content": system_prompt})
    for user, assistant in chat_history:
        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
    conversation.append({"role": "user", "content": message})

    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
    input_ids = input_ids.to(model.device)

    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        {"input_ids": input_ids},
        streamer=streamer,
        max_new_tokens=max_new_tokens,
        do_sample=do_sample,
        top_p=top_p,
        top_k=top_k,
        temperature=temperature,
        num_beams=1,
        repetition_penalty=repetition_penalty,
    )

    start_time = time.time()
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()

    outputs = []
    sum_tokens = 0
    for text in streamer:
        num_tokens = len(tokenizer.tokenize(text))
        sum_tokens += num_tokens
        
        outputs.append(text)
        yield "".join(outputs)

    time_delta = execution_time_calculator(start_time, log=False)

    generation_speed = token_per_second_calculator(sum_tokens, time_delta)

    print(f"generation_speed: {generation_speed}")


chatbot = gr.Chatbot(placeholder=PLACEHOLDER, scale=1, show_copy_button=True, height="68%", rtl=True) #,  elem_classes=["chatbot"])
chat_input = gr.Textbox(show_label=False, lines=2, rtl=True, placeholder="ورودی", show_copy_button=True, scale=4)
submit_btn = gr.Button(variant="primary", value="ارسال", size="sm", scale=1, elem_classes=["_button"])


chat_interface = gr.ChatInterface(
    fn=generate,
    additional_inputs_accordion=gr.Accordion(label="ورودی‌های اضافی", open=False),
    additional_inputs=[
        gr.Slider(
            label="حداکثر تعداد توکن ها",
            minimum=1,
            maximum=MAX_MAX_NEW_TOKENS,
            step=1,
            value=DEFAULT_MAX_NEW_TOKENS,
        ),
        gr.Slider(
            label="Temperature",
            minimum=0.01,
            maximum=4.0,
            step=0.01,
            value=0.6,
        ),
        gr.Slider(
            label="Top-p",
            minimum=0.05,
            maximum=1.0,
            step=0.01,
            value=0.65,
        ),
        gr.Slider(
            label="Top-k",
            minimum=1,
            maximum=1000,
            step=1,
            value=40,
        ),
        gr.Slider(
            label="جریمه تکرار",
            minimum=1.0,
            maximum=2.0,
            step=0.05,
            value=1.2,
        ),
        gr.Dropdown(
            label="نمونه‌گیری",
            choices=[False, True],
            value=True)
    ],
    stop_btn="توقف",
    chatbot=chatbot,
    textbox=chat_input,
    submit_btn=submit_btn,
    retry_btn="🔄 تلاش مجدد",
    undo_btn="↩️ بازگشت",
    clear_btn="🗑️ پاک کردن",
    title="درنا، محصول مرکز تحقیقات هوش مصنوعی پارت"
)


with gr.Blocks(css=custom_css, fill_height=False) as demo:
    gr.Markdown(DESCRIPTION)
    chat_interface.render()


if __name__ == "__main__":
    demo.queue(max_size=20).launch()