File size: 9,192 Bytes
7d1962a
9aa8f5f
ab13bd6
 
55029ff
ef219f6
9aa8f5f
 
 
 
 
 
 
 
 
 
 
4d31b4c
 
7d1962a
4ecef00
 
 
 
f1ada87
4ecef00
 
 
 
c63dc5b
f85425d
4ecef00
3440524
4d31b4c
 
 
3440524
4d31b4c
 
 
 
 
 
 
 
 
 
 
ea5ec2c
 
 
 
4d31b4c
 
 
3440524
4d31b4c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
67df941
ab13bd6
05eef7a
ca877b2
ab13bd6
ec98626
 
4d31b4c
ec98626
ab13bd6
835f9a2
9aa8f5f
ab13bd6
 
4d31b4c
 
ab13bd6
4d31b4c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9b61b41
4d31b4c
 
 
 
 
 
9b61b41
4d31b4c
 
 
 
 
 
 
 
ab13bd6
ca877b2
 
ab13bd6
4d31b4c
 
55029ff
05eef7a
4d31b4c
05eef7a
 
 
 
ef219f6
 
 
4d31b4c
ab13bd6
ca877b2
05eef7a
 
4d31b4c
 
 
 
9b61b41
05eef7a
 
 
 
 
 
9aa8f5f
 
 
 
 
 
 
 
 
 
 
 
4d31b4c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
05eef7a
 
4d31b4c
 
 
 
 
 
 
ab13bd6
05eef7a
4d31b4c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ca877b2
4d31b4c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ca877b2
ab13bd6
4d31b4c
 
7a3d937
9aa8f5f
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
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
import os

import gradio as gr
from text_generation import Client
from conversation import get_conv_template
from transformers import AutoTokenizer
from pymongo import MongoClient

DB_NAME = os.getenv("MONGO_DBNAME", "taiwan-llm")
USER = os.getenv("MONGO_USER")
PASSWORD = os.getenv("MONGO_PASSWORD")

uri = f"mongodb+srv://{USER}:{PASSWORD}@{DB_NAME}.kvwjiok.mongodb.net/?retryWrites=true&w=majority"
mongo_client = MongoClient(uri)
db = mongo_client[DB_NAME]
conversations_collection = db['conversations']

DESCRIPTION = """
# Language Models for Taiwanese Culture

<p align="center">
✍️ <a href="https://huggingface.co/spaces/yentinglin/Taiwan-LLaMa2" target="_blank">Online Demo</a>  

🤗 <a href="https://huggingface.co/yentinglin" target="_blank">HF Repo</a> • 🐦 <a href="https://twitter.com/yentinglin56" target="_blank">Twitter</a> • 📃 <a href="https://arxiv.org/pdf/2305.13711.pdf" target="_blank">[Paper Coming Soon]</a>  
• 👨️ <a href="https://github.com/MiuLab/Taiwan-LLaMa/tree/main" target="_blank">Github Repo</a> 
    <br/><br/>
    <img src="https://www.csie.ntu.edu.tw/~miulab/taiwan-llama/logo-v2.png" width="100"> <br/>
</p>

# 🌟 Checkout New [Taiwan-LLM UI](http://www.twllm.com) 🌟


Taiwan-LLaMa is a fine-tuned model specifically designed for traditional mandarin applications. It is built upon the LLaMa 2 architecture and includes a pretraining phase with over 5 billion tokens and fine-tuning with over 490k multi-turn conversational data in Traditional Mandarin.

## Key Features

1. **Traditional Mandarin Support**: The model is fine-tuned to understand and generate text in Traditional Mandarin, making it suitable for Taiwanese culture and related applications.

2. **Instruction-Tuned**: Further fine-tuned on conversational data to offer context-aware and instruction-following responses.

3. **Performance on Vicuna Benchmark**: Taiwan-LLaMa's relative performance on Vicuna Benchmark is measured against models like GPT-4 and ChatGPT. It's particularly optimized for Taiwanese culture.

4. **Flexible Customization**: Advanced options for controlling the model's behavior like system prompt, temperature, top-p, and top-k are available in the demo.

## Model Versions

Different versions of Taiwan-LLaMa are available:

- **Taiwan-LLM v2.0 (This demo)**: Cleaner pretraining, Better post-training
- **Taiwan-LLM v1.0**: Optimized for Taiwanese Culture
- **Taiwan-LLM v0.9**: Partial instruction set
- **Taiwan-LLM v0.0**: No Traditional Mandarin pretraining

The models can be accessed from the provided links in the Hugging Face repository.

Try out the demo to interact with Taiwan-LLaMa and experience its capabilities in handling Traditional Mandarin!
"""

LICENSE = """
## Licenses

- Code is licensed under Apache 2.0 License.
- Models are licensed under the LLAMA 2 Community License.
- By using this model, you agree to the terms and conditions specified in the license.
- By using this demo, you agree to share your input utterances with us to improve the model.

## Acknowledgements

Taiwan-LLaMa project acknowledges the efforts of the [Meta LLaMa team](https://github.com/facebookresearch/llama) and [Vicuna team](https://github.com/lm-sys/FastChat) in democratizing large language models.
"""

DEFAULT_SYSTEM_PROMPT = "你是人工智慧助理,以下是用戶和人工智能助理之間的對話。你要對用戶的問題提供有用、安全、詳細和禮貌的回答。 您是由國立臺灣大學的林彥廷博士生為研究目的而建造的。"

endpoint_url = os.environ.get("ENDPOINT_URL", "http://127.0.0.1:8080")
client = Client(endpoint_url, timeout=120)
eos_token = "</s>"
MAX_MAX_NEW_TOKENS = 4096
DEFAULT_MAX_NEW_TOKENS = 1536

max_prompt_length = 8192 - MAX_MAX_NEW_TOKENS - 10

model_name = "yentinglin/Taiwan-LLM-7B-v2.0-chat"
tokenizer = AutoTokenizer.from_pretrained(model_name)

with gr.Blocks() as demo:
    gr.Markdown(DESCRIPTION)

    chatbot = gr.Chatbot()
    with gr.Row():
        msg = gr.Textbox(
            container=False,
            show_label=False,
            placeholder='Type a message...',
            scale=10,
        )
        submit_button = gr.Button('Submit',
                                  variant='primary',
                                  scale=1,
                                  min_width=0)

    with gr.Row():
        retry_button = gr.Button('🔄  Retry', variant='secondary')
        undo_button = gr.Button('↩️ Undo', variant='secondary')
        clear = gr.Button('🗑️  Clear', variant='secondary')

    saved_input = gr.State()

    with gr.Accordion(label='Advanced options', open=False):
        system_prompt = gr.Textbox(label='System prompt',
                                   value=DEFAULT_SYSTEM_PROMPT,
                                   lines=6)
        max_new_tokens = gr.Slider(
            label='Max new tokens',
            minimum=1,
            maximum=MAX_MAX_NEW_TOKENS,
            step=1,
            value=DEFAULT_MAX_NEW_TOKENS,
        )
        temperature = gr.Slider(
            label='Temperature',
            minimum=0.1,
            maximum=1.0,
            step=0.1,
            value=0.3,
        )
        top_p = gr.Slider(
            label='Top-p (nucleus sampling)',
            minimum=0.05,
            maximum=1.0,
            step=0.05,
            value=0.95,
        )
        top_k = gr.Slider(
            label='Top-k',
            minimum=1,
            maximum=1000,
            step=1,
            value=50,
        )

    def user(user_message, history):
        return "", history + [[user_message, None]]


    def bot(history, max_new_tokens, temperature, top_p, top_k, system_prompt):
        conv = get_conv_template("twllm_v2").copy()
        roles = {"human": conv.roles[0], "gpt": conv.roles[1]}  # map human to USER and gpt to ASSISTANT
        conv.system = system_prompt
        for user, bot in history:
            conv.append_message(roles['human'], user)
            conv.append_message(roles["gpt"], bot)
        msg = conv.get_prompt()
        prompt_tokens = tokenizer.encode(msg)
        length_of_prompt = len(prompt_tokens)
        if length_of_prompt > max_prompt_length:
            msg = tokenizer.decode(prompt_tokens[-max_prompt_length + 1:])

        history[-1][1] = ""
        for response in client.generate_stream(
                msg,
                max_new_tokens=max_new_tokens,
                temperature=temperature,
                top_p=top_p,
                top_k=top_k,
                repetition_penalty=1.1,
        ):
            if not response.token.special:
                character = response.token.text
                history[-1][1] += character
                yield history

        # After generating the response, store the conversation history in MongoDB
        conversation_document = {
            "model_name": model_name,
            "history": history,
            "system_prompt": system_prompt,
            "max_new_tokens": max_new_tokens,
            "temperature": temperature,
            "top_p": top_p,
            "top_k": top_k,
        }
        conversations_collection.insert_one(conversation_document)

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        fn=bot,
        inputs=[
            chatbot,
            max_new_tokens,
            temperature,
            top_p,
            top_k,
            system_prompt,
        ],
        outputs=chatbot
    )
    submit_button.click(
        user, [msg, chatbot], [msg, chatbot], queue=False
    ).then(
        fn=bot,
        inputs=[
            chatbot,
            max_new_tokens,
            temperature,
            top_p,
            top_k,
            system_prompt,
        ],
        outputs=chatbot
    )


    def delete_prev_fn(
            history: list[tuple[str, str]]) -> tuple[list[tuple[str, str]], str]:
        try:
            message, _ = history.pop()
        except IndexError:
            message = ''
        return history, message or ''


    def display_input(message: str,
                      history: list[tuple[str, str]]) -> list[tuple[str, str]]:
        history.append((message, ''))
        return history

    retry_button.click(
        fn=delete_prev_fn,
        inputs=chatbot,
        outputs=[chatbot, saved_input],
        api_name=False,
        queue=False,
    ).then(
        fn=display_input,
        inputs=[saved_input, chatbot],
        outputs=chatbot,
        api_name=False,
        queue=False,
    ).then(
        fn=bot,
        inputs=[
            chatbot,
            max_new_tokens,
            temperature,
            top_p,
            top_k,
            system_prompt,
        ],
        outputs=chatbot,
    )

    undo_button.click(
        fn=delete_prev_fn,
        inputs=chatbot,
        outputs=[chatbot, saved_input],
        api_name=False,
        queue=False,
    ).then(
        fn=lambda x: x,
        inputs=[saved_input],
        outputs=msg,
        api_name=False,
        queue=False,
    )

    clear.click(lambda: None, None, chatbot, queue=False)

    gr.Markdown(LICENSE)

demo.queue(concurrency_count=4, max_size=128)
demo.launch()