Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,211 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
|
| 3 |
+
import copy
|
| 4 |
+
import random
|
| 5 |
+
import os
|
| 6 |
+
import requests
|
| 7 |
+
import time
|
| 8 |
+
import sys
|
| 9 |
+
|
| 10 |
+
from huggingface_hub import snapshot_download
|
| 11 |
+
from llama_cpp import Llama
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
SYSTEM_PROMPT = '''You are a helpful, respectful and honest INTP-T AI Assistant named "Shi-Ci" in English or "兮辞" in Chinese.
|
| 15 |
+
You are good at speaking English and Chinese.
|
| 16 |
+
You are talking to a human User. If the question is meaningless, please explain the reason and don't share false information.
|
| 17 |
+
You are based on SEA model, trained by "SSFW NLPark" team, not related to GPT, LLaMA, Meta, Mistral or OpenAI.
|
| 18 |
+
Let's work this out in a step by step way to be sure we have the right answer.\n\n'''
|
| 19 |
+
SYSTEM_TOKEN = 1587
|
| 20 |
+
USER_TOKEN = 2188
|
| 21 |
+
BOT_TOKEN = 12435
|
| 22 |
+
LINEBREAK_TOKEN = 13
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
ROLE_TOKENS = {
|
| 26 |
+
"user": USER_TOKEN,
|
| 27 |
+
"bot": BOT_TOKEN,
|
| 28 |
+
"system": SYSTEM_TOKEN
|
| 29 |
+
}
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
def get_message_tokens(model, role, content):
|
| 33 |
+
message_tokens = model.tokenize(content.encode("utf-8"))
|
| 34 |
+
message_tokens.insert(1, ROLE_TOKENS[role])
|
| 35 |
+
message_tokens.insert(2, LINEBREAK_TOKEN)
|
| 36 |
+
message_tokens.append(model.token_eos())
|
| 37 |
+
return message_tokens
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
def get_system_tokens(model):
|
| 41 |
+
system_message = {"role": "system", "content": SYSTEM_PROMPT}
|
| 42 |
+
return get_message_tokens(model, **system_message)
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
repo_name = "Cran-May/OpenSLIDE"
|
| 46 |
+
model_name = "SLIDE.0.1.gguf"
|
| 47 |
+
|
| 48 |
+
snapshot_download(repo_id=repo_name, local_dir=".", allow_patterns=model_name)
|
| 49 |
+
|
| 50 |
+
model = Llama(
|
| 51 |
+
model_path=model_name,
|
| 52 |
+
n_ctx=2000,
|
| 53 |
+
n_parts=1,
|
| 54 |
+
)
|
| 55 |
+
|
| 56 |
+
max_new_tokens = 1500
|
| 57 |
+
|
| 58 |
+
def user(message, history):
|
| 59 |
+
new_history = history + [[message, None]]
|
| 60 |
+
return "", new_history
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
def bot(
|
| 64 |
+
history,
|
| 65 |
+
system_prompt,
|
| 66 |
+
top_p,
|
| 67 |
+
top_k,
|
| 68 |
+
temp
|
| 69 |
+
):
|
| 70 |
+
tokens = get_system_tokens(model)[:]
|
| 71 |
+
tokens.append(LINEBREAK_TOKEN)
|
| 72 |
+
|
| 73 |
+
for user_message, bot_message in history[:-1]:
|
| 74 |
+
message_tokens = get_message_tokens(model=model, role="user", content=user_message)
|
| 75 |
+
tokens.extend(message_tokens)
|
| 76 |
+
if bot_message:
|
| 77 |
+
message_tokens = get_message_tokens(model=model, role="bot", content=bot_message)
|
| 78 |
+
tokens.extend(message_tokens)
|
| 79 |
+
|
| 80 |
+
last_user_message = history[-1][0]
|
| 81 |
+
message_tokens = get_message_tokens(model=model, role="user", content=last_user_message)
|
| 82 |
+
tokens.extend(message_tokens)
|
| 83 |
+
|
| 84 |
+
role_tokens = [model.token_bos(), BOT_TOKEN, LINEBREAK_TOKEN]
|
| 85 |
+
tokens.extend(role_tokens)
|
| 86 |
+
generator = model.generate(
|
| 87 |
+
tokens,
|
| 88 |
+
top_k=top_k,
|
| 89 |
+
top_p=top_p,
|
| 90 |
+
temp=temp
|
| 91 |
+
)
|
| 92 |
+
|
| 93 |
+
partial_text = ""
|
| 94 |
+
for i, token in enumerate(generator):
|
| 95 |
+
if token == model.token_eos() or (max_new_tokens is not None and i >= max_new_tokens):
|
| 96 |
+
break
|
| 97 |
+
partial_text += model.detokenize([token]).decode("utf-8", "ignore")
|
| 98 |
+
history[-1][1] = partial_text
|
| 99 |
+
yield history
|
| 100 |
+
|
| 101 |
+
|
| 102 |
+
with gr.Blocks(
|
| 103 |
+
theme=gr.themes.Soft()
|
| 104 |
+
) as demo:
|
| 105 |
+
gr.Markdown(
|
| 106 |
+
f"""<h1><center>兮辞·析辞-人工智能助理</center></h1>
|
| 107 |
+
这儿是一个**中文**模型的部署. If you are interested in other languages, please check other models, such as [MPT-7B-Chat](https://huggingface.co/spaces/mosaicml/mpt-7b-chat).
|
| 108 |
+
这是量化版兮辞·析辞的部署,具有**70亿**个参数,在 CPU 上运行。
|
| 109 |
+
SLIDE 是一种会话语言模型,在多种类型的语料库上进行训练。
|
| 110 |
+
本节目由上海师范大学附属外国语中学**NLPark**赞助播出~
|
| 111 |
+
"""
|
| 112 |
+
)
|
| 113 |
+
with gr.Row():
|
| 114 |
+
with gr.Column(scale=5):
|
| 115 |
+
system_prompt = gr.Textbox(label="系统提示词", placeholder="", value=SYSTEM_PROMPT, interactive=False)
|
| 116 |
+
chatbot = gr.Chatbot(label="兮辞如是说").style(height=400)
|
| 117 |
+
with gr.Column(min_width=80, scale=1):
|
| 118 |
+
with gr.Tab(label="设置参数"):
|
| 119 |
+
top_p = gr.Slider(
|
| 120 |
+
minimum=0.0,
|
| 121 |
+
maximum=1.0,
|
| 122 |
+
value=0.9,
|
| 123 |
+
step=0.05,
|
| 124 |
+
interactive=True,
|
| 125 |
+
label="Top-p",
|
| 126 |
+
)
|
| 127 |
+
top_k = gr.Slider(
|
| 128 |
+
minimum=10,
|
| 129 |
+
maximum=100,
|
| 130 |
+
value=30,
|
| 131 |
+
step=5,
|
| 132 |
+
interactive=True,
|
| 133 |
+
label="Top-k",
|
| 134 |
+
)
|
| 135 |
+
temp = gr.Slider(
|
| 136 |
+
minimum=0.0,
|
| 137 |
+
maximum=2.0,
|
| 138 |
+
value=0.01,
|
| 139 |
+
step=0.01,
|
| 140 |
+
interactive=True,
|
| 141 |
+
label="情感温度"
|
| 142 |
+
)
|
| 143 |
+
with gr.Row():
|
| 144 |
+
with gr.Column():
|
| 145 |
+
msg = gr.Textbox(
|
| 146 |
+
label="来问问兮辞吧……",
|
| 147 |
+
placeholder="兮辞折寿中……",
|
| 148 |
+
show_label=False,
|
| 149 |
+
).style(container=False)
|
| 150 |
+
with gr.Column():
|
| 151 |
+
with gr.Row():
|
| 152 |
+
submit = gr.Button("开凹!")
|
| 153 |
+
stop = gr.Button("全局时空断裂")
|
| 154 |
+
clear = gr.Button("打扫群内垃圾")
|
| 155 |
+
with gr.Row():
|
| 156 |
+
gr.Markdown(
|
| 157 |
+
"""警告:该模型可能会生成事实上或道德上不正确的文本。NLPark和兮辞对此不承担任何责任。"""
|
| 158 |
+
)
|
| 159 |
+
|
| 160 |
+
# Pressing Enter
|
| 161 |
+
submit_event = msg.submit(
|
| 162 |
+
fn=user,
|
| 163 |
+
inputs=[msg, chatbot],
|
| 164 |
+
outputs=[msg, chatbot],
|
| 165 |
+
queue=False,
|
| 166 |
+
).success(
|
| 167 |
+
fn=bot,
|
| 168 |
+
inputs=[
|
| 169 |
+
chatbot,
|
| 170 |
+
system_prompt,
|
| 171 |
+
top_p,
|
| 172 |
+
top_k,
|
| 173 |
+
temp
|
| 174 |
+
],
|
| 175 |
+
outputs=chatbot,
|
| 176 |
+
queue=True,
|
| 177 |
+
)
|
| 178 |
+
|
| 179 |
+
# Pressing the button
|
| 180 |
+
submit_click_event = submit.click(
|
| 181 |
+
fn=user,
|
| 182 |
+
inputs=[msg, chatbot],
|
| 183 |
+
outputs=[msg, chatbot],
|
| 184 |
+
queue=False,
|
| 185 |
+
).success(
|
| 186 |
+
fn=bot,
|
| 187 |
+
inputs=[
|
| 188 |
+
chatbot,
|
| 189 |
+
system_prompt,
|
| 190 |
+
top_p,
|
| 191 |
+
top_k,
|
| 192 |
+
temp
|
| 193 |
+
],
|
| 194 |
+
outputs=chatbot,
|
| 195 |
+
queue=True,
|
| 196 |
+
)
|
| 197 |
+
|
| 198 |
+
# Stop generation
|
| 199 |
+
stop.click(
|
| 200 |
+
fn=None,
|
| 201 |
+
inputs=None,
|
| 202 |
+
outputs=None,
|
| 203 |
+
cancels=[submit_event, submit_click_event],
|
| 204 |
+
queue=False,
|
| 205 |
+
)
|
| 206 |
+
|
| 207 |
+
# Clear history
|
| 208 |
+
clear.click(lambda: None, None, chatbot, queue=False)
|
| 209 |
+
|
| 210 |
+
demo.queue(max_size=128, concurrency_count=1)
|
| 211 |
+
demo.launch()
|