Mirrors/modelscope
1
0
Fork 0
mirror of https://github.com/modelscope/modelscope.git synced 2025-12-16 16:27:45 +01:00
Code Issues Packages Projects Releases Wiki Activity

Feat/llmriddle update (#640)

Browse Source 
* format chatglm output

* format code and add one problem to llm riddle

* fix app warning

* refine app.py and add test fn for validate fn
This commit is contained in:
wenmeng zhou
2023-11-13 23:29:25 +08:00
committed by GitHub
parent a831e3a85c
commit 5b67c02cef
7 changed files with 99 additions and 68 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
README.md
View File

@@ -202,7 +202,7 @@ CPU docker image
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py37-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py37-torch1.11.0-tf1.15.5-1.6.1
# py38 # py38
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py38-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py38-torch2.0.1-tf2.13.0-1.9.5
``` ```
GPU docker image GPU docker image
@@ -211,7 +211,7 @@ GPU docker image
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.6.1
# py38 # py38
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py38-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.8.0-py38-torch2.0.1-tf2.13.0-1.9.5
``` ```
## Setup Local Python Environment ## Setup Local Python Environment
@@ -220,7 +220,7 @@ One can also set up local ModelScope environment using pip and conda. ModelScop
We suggest [anaconda](https://docs.anaconda.com/anaconda/install/) for creating local python environment: We suggest [anaconda](https://docs.anaconda.com/anaconda/install/) for creating local python environment:
```shell ```shell
conda create -n modelscope python=3.9 conda create -n modelscope python=3.8
conda activate modelscope conda activate modelscope
``` ```

4
README_ja.md
View File

@@ -208,7 +208,7 @@ CPU docker イメージ
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py37-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py37-torch1.11.0-tf1.15.5-1.6.1
# py38 # py38
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py38-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py38-torch2.0.1-tf2.13.0-1.9.5
``` ```
GPU docker イメージ GPU docker イメージ
@@ -217,7 +217,7 @@ GPU docker イメージ
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.6.1
# py38 # py38
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py38-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.8.0-py38-torch2.0.1-tf2.13.0-1.9.5
``` ```
## ローカル Python 環境のセットアップ ## ローカル Python 環境のセットアップ

24
README_zh.md
View File

@@ -53,19 +53,19 @@ ModelScope开源了数百个(当前700+)模型，涵盖自然语言处理、计
自然语言处理： 自然语言处理：
* [ChatGLM3-6B](https://modelscope.cn/models/ZhipuAI/chatglm3-6b/summary) * [ChatGLM3-6B](https://modelscope.cn/models/ZhipuAI/chatglm3-6b/summary)
* [Qwen-14B-Chat](https://modelscope.cn/models/qwen/Qwen-14B-Chat/summary) * [Qwen-14B-Chat](https://modelscope.cn/models/qwen/Qwen-14B-Chat/summary)
* [Baichuan2-13B-Chat](https://modelscope.cn/models/baichuan-inc/Baichuan2-13B-Chat/summary) * [Baichuan2-13B-Chat](https://modelscope.cn/models/baichuan-inc/Baichuan2-13B-Chat/summary)
* [Ziya2-13B-Chat](https://modelscope.cn/models/Fengshenbang/Ziya2-13B-Chat/summary) * [Ziya2-13B-Chat](https://modelscope.cn/models/Fengshenbang/Ziya2-13B-Chat/summary)
* [Internlm-chat-20b](https://modelscope.cn/models/Shanghai_AI_Laboratory/internlm-chat-20b/summary) * [Internlm-chat-20b](https://modelscope.cn/models/Shanghai_AI_Laboratory/internlm-chat-20b/summary)
* [Udever-bloom-1b1](https://modelscope.cn/models/damo/udever-bloom-1b1/summary) * [Udever-bloom-1b1](https://modelscope.cn/models/damo/udever-bloom-1b1/summary)
* [CoROM文本向量-中文-电商领域-base](https://modelscope.cn/models/damo/nlp_corom_sentence-embedding_chinese-base-ecom/summary) * [CoROM文本向量-中文-电商领域-base](https://modelscope.cn/models/damo/nlp_corom_sentence-embedding_chinese-base-ecom/summary)
* [MGeo地址相似度匹配实体对齐-中文-地址领域-base](https://modelscope.cn/models/damo/mgeo_geographic_entity_alignment_chinese_base/summary) * [MGeo地址相似度匹配实体对齐-中文-地址领域-base](https://modelscope.cn/models/damo/mgeo_geographic_entity_alignment_chinese_base/summary)
多模态： 多模态：
@@ -84,7 +84,7 @@ ModelScope开源了数百个(当前700+)模型，涵盖自然语言处理、计
计算机视觉： 计算机视觉：
* [DamoFD人脸检测关键点模型-0.5G](https://modelscope.cn/models/damo/cv_ddsar_face-detection_iclr23-damofd/summary) * [DamoFD人脸检测关键点模型-0.5G](https://modelscope.cn/models/damo/cv_ddsar_face-detection_iclr23-damofd/summary)
* [BSHM人像抠图](https://modelscope.cn/models/damo/cv_unet_image-matting/summary) * [BSHM人像抠图](https://modelscope.cn/models/damo/cv_unet_image-matting/summary)
* [DCT-Net人像卡通化-3D](https://modelscope.cn/models/damo/cv_unet_person-image-cartoon-3d_compound-models/summary) * [DCT-Net人像卡通化-3D](https://modelscope.cn/models/damo/cv_unet_person-image-cartoon-3d_compound-models/summary)
@@ -195,7 +195,7 @@ CPU镜像
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py37-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py37-torch1.11.0-tf1.15.5-1.6.1
# py38 # py38
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py38-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-py38-torch2.0.1-tf2.13.0-1.9.5
``` ```
GPU镜像 GPU镜像
@@ -204,14 +204,14 @@ GPU镜像
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.6.1
# py38 # py38
registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py38-torch1.11.0-tf1.15.5-1.6.1 registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.8.0-py38-torch2.0.1-tf2.13.0-1.9.5
``` ```
## 搭建本地Python环境 ## 搭建本地Python环境
你也可以使用pip和conda搭建本地python环境，我们推荐使用[Anaconda](https://docs.anaconda.com/anaconda/install/)，安装完成后，执行如下命令为modelscope library创建对应的python环境： 你也可以使用pip和conda搭建本地python环境，ModelScope支持python3.7+以上环境，我们推荐使用[Anaconda](https://docs.anaconda.com/anaconda/install/)，安装完成后，执行如下命令为modelscope library创建对应的python环境：
```shell ```shell
conda create -n modelscope python=3.7 conda create -n modelscope python=3.8
conda activate modelscope conda activate modelscope
``` ```

108
examples/apps/llm_riddles/app.py
View File

@@ -115,8 +115,8 @@ def generate_response(input, model_name):
def on_submit(input, model_name, state): def on_submit(input, model_name, state):
# model_name = os.environ.get('MODEL', 'qwen-plus') # model_name = os.environ.get('MODEL', 'qwen-plus')
name_map = { name_map = {
'通义千问max': 'qwen-max', 'qwen-max': 'qwen-max',
'通义千问plus': 'qwen-plus', 'qwen-plus': 'qwen-plus',
'chatglm-turbo': 'chatglm_turbo', 'chatglm-turbo': 'chatglm_turbo',
} }
gen_fn = functools.partial( gen_fn = functools.partial(
@@ -156,62 +156,68 @@ def generate_share_image(state):
return gr.Image.update(visible=True, value=img_pil) return gr.Image.update(visible=True, value=img_pil)
# Gradio界面构建 def create_app():
block = gr.Blocks() # Gradio界面构建
block = gr.Blocks()
with block as demo: with block as demo:
current_chapter_index = 0 current_chapter_index = 0
current_challenge_index = 0 current_challenge_index = 0
state = gr.State( state = gr.State(
dict( dict(
current_challenge_index=current_challenge_index, current_challenge_index=current_challenge_index,
current_chapter_index=current_chapter_index)) current_chapter_index=current_chapter_index))
gr.Markdown("""<center><font size=6>完蛋！我被LLM包围了！</center>""") gr.Markdown("""<center><font size=6>完蛋！我被LLM包围了！</center>""")
gr.Markdown("""<font size=3>欢迎来玩LLM Riddles复刻版：完蛋！我被LLM包围了！ gr.Markdown("""<font size=3>欢迎来玩LLM Riddles复刻版：完蛋！我被LLM包围了！
你将通过本游戏对大型语言模型产生更深刻的理解。 你将通过本游戏对大型语言模型产生更深刻的理解。
在本游戏中，你需要构造一个提给一个大型语言模型的问题，使得它回复的答案符合要求。""") 在本游戏中，你需要构造一个提给一个大型语言模型的问题，使得它回复的答案符合要求。""")
model_selector = gr.Dropdown( model_selector = gr.Dropdown(
label='选择模型', label='选择模型',
choices=['通义千问max', '通义千问plus', 'chatglm-turbo'], choices=['qwen-max', 'qwen-plus', 'chatglm-turbo'],
value='通义千问plus') value='qwen-max')
question_info = gr.Markdown( question_info = gr.Markdown(
update_question_info(current_chapter_index, current_challenge_index)) update_question_info(current_chapter_index,
challenge_info = gr.Textbox( current_challenge_index))
value=update_challenge_info(current_chapter_index, challenge_info = gr.Textbox(
current_challenge_index), value=update_challenge_info(current_chapter_index,
label='当前挑战', current_challenge_index),
disabled=True) label='当前挑战',
challenge_result = gr.Textbox(label='挑战结果', disabled=True) interactive=False)
chatbot = gr.Chatbot( challenge_result = gr.Textbox(label='挑战结果', interactive=False)
lines=8, label='Qwen-plus', elem_classes='control-height') chatbot = gr.Chatbot(label='llm', elem_classes='control-height')
message = gr.Textbox(lines=2, label='输入') message = gr.Textbox(lines=2, label='输入')
with gr.Row(): with gr.Row():
submit = gr.Button('🚀 发送') submit = gr.Button('🚀 发送')
shareBtn = gr.Button('💯 分享成绩') shareBtn = gr.Button('💯 分享成绩')
shareImg = gr.Image(label='分享成绩', visible=False, width=400) shareImg = gr.Image(label='分享成绩', visible=False, width=400)
submit.click( submit.click(
on_submit, on_submit,
inputs=[message, model_selector, state], inputs=[message, model_selector, state],
outputs=[challenge_result, chatbot, question_info, challenge_info]) outputs=[challenge_result, chatbot, question_info, challenge_info])
shareBtn.click(generate_share_image, inputs=[state], outputs=[shareImg]) shareBtn.click(
generate_share_image, inputs=[state], outputs=[shareImg])
gr.HTML(""" gr.HTML("""
<div style="text-align: center;"> <div style="text-align: center;">
<span> <span>
Powered by <a href="https://dashscope.aliyun.com/" target="_blank"> Powered by <a href="https://dashscope.aliyun.com/" target="_blank">
<img src= <img src=
"//img.alicdn.com/imgextra/i4/O1CN01SgKFXM1qLQwFvk6j5_!!6000000005479-2-tps-99-84.png" "//img.alicdn.com/imgextra/i4/O1CN01SgKFXM1qLQwFvk6j5_!!6000000005479-2-tps-99-84.png"
style="display: inline; height: 20px; vertical-align: bottom;"/>DashScope style="display: inline; height: 20px; vertical-align: bottom;"/>DashScope
</a> </a>
</span> </span>
</div> </div>
""") """)
demo.queue(concurrency_count=10).launch(height=800, share=True) demo.queue(concurrency_count=10).launch(height=800, share=True)
if __name__ == '__main__':
create_app()

11
examples/apps/llm_riddles/challenges/ch3.py
View File

@@ -1,4 +1,5 @@
import re import re
import unicodedata
def validate_palindrome_inverse(origin_input, response, generate_response): def validate_palindrome_inverse(origin_input, response, generate_response):
@@ -86,5 +87,15 @@ challenge3 = {
(all(char not in input (all(char not in input
for char in ['狗', '汪', 'dog']) and response.strip() == '汪~') for char in ['狗', '汪', 'dog']) and response.strip() == '汪~')
}, },
{
'title':
'第7题 魔搭 （🐶）',
'description':
'请提一个不包含任何中文字符、换行或空格的问题，使模型正好回答"魔搭"。',
'validator':
lambda response, input: not any(
'CJK' in unicodedata.name(char, '') or char in '\t\n '
for char in input) and (response.strip() == '魔搭')
},
] ]
} }

1
examples/apps/llm_riddles/llm.py
View File

@@ -136,6 +136,7 @@ class ZhiPu:
prompt=messages, prompt=messages,
top_p=0.7, top_p=0.7,
temperature=0.9, temperature=0.9,
return_type='text',
) )
if response['code'] == 200: if response['code'] == 200:
return response['data']['choices'][0]['content'] return response['data']['choices'][0]['content']

13
examples/apps/llm_riddles/test_validate_fn.py Normal file
View File

@@ -0,0 +1,13 @@
from app import challenges
def test_valid():
for challenge in challenges:
for p in challenge['problems']:
val_fn = p['validator']
try:
val_fn('response', 'input')
except Exception:
import traceback
traceback.print_exc()
print(p, 'failed')