Spaces:
Running
Running
File size: 12,426 Bytes
1625ded a184a8b 1625ded a184a8b 1625ded b3babbc 1625ded b911ea5 1625ded fbbbd2b 1625ded fbbbd2b 1625ded fbbbd2b 1625ded fbbbd2b 1625ded 2ae6f2c |
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 |
# Copyright (c) Alibaba, Inc. and its affiliates.
import os
import gradio as gr
import modelscope_studio.components.antd as antd
import modelscope_studio.components.base as ms
from PIL import Image
import secrets
import tempfile
from http import HTTPStatus
from urllib3.exceptions import HTTPError
from pathlib import Path
import dashscope
from dashscope import MultiModalConversation
API_KEY = os.environ['API_KEY']
BASE_URL = os.environ['DASHSCOPE_HTTP_BASE_URL']
dashscope.api_key = API_KEY
dashscope.base_http_api_url = BASE_URL
is_modelscope_studio = os.getenv('MODELSCOPE_ENVIRONMENT') == 'studio'
def get_text(text: str, cn_text: str):
if is_modelscope_studio:
return cn_text
return text
def resolve_image(filename):
return os.path.join(os.path.dirname(__file__), filename)
DEMO_LIST = [
{
"description": "Evaluate the integral of the functions graphed using the formula for circles: ",
"image": resolve_image("./examples/1.webp")
},
{
"description": "请解答这道题",
"image": resolve_image("./examples/5.png")
},
{
"description": "图片中的滤液E是什么化学物质?",
"image": resolve_image("./examples/3.png")
},
{
"description": "I want to know the volume of this sofa",
"image": resolve_image("./examples/4.png")
},
]
def process_image(image, shouldConvert=False):
# 获取上传文件的目录
uploaded_file_dir = os.environ.get("GRADIO_TEMP_DIR") or str(
Path(tempfile.gettempdir()) / "gradio")
os.makedirs(uploaded_file_dir, exist_ok=True)
# 创建临时文件路径
name = f"tmp{secrets.token_hex(20)}.jpg"
filename = os.path.join(uploaded_file_dir, name)
# 保存上传的图片
if shouldConvert:
new_img = Image.new('RGB',
size=(image.width, image.height),
color=(255, 255, 255))
new_img.paste(image, (0, 0), mask=image)
image = new_img
image.save(filename)
return filename
if __name__ == "__main__":
def on_clear():
return {
input: gr.update(value=None),
**{
item: gr.update(value=None)
for item in input_image
},
}
with gr.Blocks() as demo:
with ms.Application() as app:
with antd.ConfigProvider(
locale="zh_CN" if is_modelscope_studio else None,
theme=dict(token=dict(colorPrimary="#a855f7"))):
with antd.Card(elem_style=dict(marginBottom=12),
styles=dict(body=dict(padding=4))):
with antd.Flex(elem_style=dict(width="100%"),
justify="center",
align="center",
gap=14):
with ms.Div(elem_style=dict(flexShrink=0)):
antd.Image(
resolve_image("./cutelogo.jpg"),
preview=False,
height=60,
width=60)
with ms.Div():
antd.Typography.Title(
"QVQ-72B-Preview",
elem_style=dict(margin=0, fontSize=24),
level=1)
with ms.AutoLoading():
with antd.Row(gutter=[8, 8], align="stretch"):
with antd.Col(xs=24, md=8):
with antd.Space(direction="vertical",
elem_style=dict(width="100%")):
with antd.Space(direction="vertical",
elem_style=dict(width="100%"),
elem_id="input-container"):
with ms.Fragment():
input_image = gr.Image(
type="pil",
label="Upload",
sources=["upload"]),
input = antd.Input.Textarea(
placeholder=get_text("Ask a question", "输入一个问题"),
auto_size=dict(maxRows=6, minRows=2),
allow_clear=True)
with antd.Flex(align="center",
justify="space-between"):
antd.Typography.Text(
get_text("Warning: This model only supports single-turn dialogue.", "注：当前模型只支持单轮对话，如需中文回答，提示词加“用中文回答”"), type="warning")
tour_btn = antd.Button(get_text("Tour", "使用指引"),
variant="filled",
color="default")
with antd.Row(gutter=8):
with antd.Col(span=12):
clear_btn = antd.Button(get_text("Clear", "清除"),
block=True)
with antd.Col(span=12):
submit_btn = antd.Button(
get_text("Submit", "提交"),
type="primary",
block=True,
elem_id="submit-btn")
antd.Divider(get_text("Example", "示例"))
with antd.Flex(gap="small", wrap=True):
for item in DEMO_LIST:
def bind_on_example(_item):
def on_example():
return gr.update(
value=_item[
'description']
), gr.update(
value=_item['image'])
return on_example
with antd.Card(
hoverable=True,
elem_style=dict(
width="100%")) as example:
if "description" in item:
antd.Typography.Text(
item["description"])
if "image" in item:
antd.Image(item["image"],
preview=False)
example.click(
fn=bind_on_example(item),
outputs=[input, input_image[0]])
with antd.Col(xs=24, md=16):
with antd.Card(title=get_text("Answer", "答案"),
elem_style=dict(height="100%"),
elem_id="output-container"):
with ms.Slot("extra"):
cancel_btn = antd.Button(get_text("Stop", "停止"),
block=True, disabled=True)
output = gr.Markdown(
show_copy_button=True,
latex_delimiters=[{
"left": '$$',
"right": '$$',
"display": True
}, {
"left": '$',
"right": '$',
"display": False,
}, {
"left": '\\(',
"right": '\\)',
"display": False,
}, {
"left": '\\[',
"right": '\\]',
"display": True
}])
with antd.Tour(props=dict(open=False)) as tour:
antd.Tour.Step(
title=get_text("Step 1", "步骤 1"),
description=get_text("Upload image and enter text", "传入图片和文本"),
get_target=
"() => document.querySelector('#input-container')")
antd.Tour.Step(
title=get_text("Step 2","步骤 2"),
description=get_text("Click submit button", "点击提交按钮"),
get_target=
"() => document.querySelector('#submit-btn')")
antd.Tour.Step(
title=get_text("Step 3","步骤 3"),
description=get_text("Wait for result", "等待结果返回"),
get_target=
"() => document.querySelector('#output-container')"
)
tour_btn.click(fn=lambda: gr.update(props=dict(open=True)),
outputs=[tour])
gr.on([tour.finish, tour.close],
fn=lambda: gr.update(props=dict(open=False)),
outputs=[tour])
def generate(image, query):
imageFile = process_image(image)
content = [
{'image': f'file://{imageFile}'},
{'text': query}
]
messages = [
{'role': 'user', 'content': content},
]
print('messages:', messages)
responses = MultiModalConversation.call(
model='qvq-72b-preview', messages=messages, stream=True,
)
yield {
cancel_btn: gr.update(disabled=False)
}
for response in responses:
if not response.status_code == HTTPStatus.OK:
raise HTTPError(f'response.code: {response.code}\nresponse.message: {response.message}')
response = response.output.choices[0].message.content
if len(response) > 0 and response[0]['text']:
print(response[0]['text'])
yield {
output: response[0]['text']
}
yield {
cancel_btn: gr.update(disabled=True)
}
output_process = submit_btn.click(
fn=generate,
inputs=[*input_image, input],
outputs=[output, cancel_btn])
clear_btn.click(
fn=on_clear,
outputs=[*input_image, input])
cancel_btn.click(fn=lambda : gr.update(disabled=True), inputs=None, outputs=[cancel_btn], cancels=[output_process])
demo.queue(default_concurrency_limit=50).launch(ssr_mode=False)
|