Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

各大推理框架硬件适配代码量统计 #8342

Closed
hong19860320 opened this issue Jan 24, 2022 · 4 comments
Closed

各大推理框架硬件适配代码量统计 #8342

hong19860320 opened this issue Jan 24, 2022 · 4 comments

Comments

@hong19860320
Copy link
Collaborator

hong19860320 commented Jan 24, 2022

统计各大推理框架支持一款新的硬件跑通 ResNet-50 所需的代码量,涉及反映总工作量的『总代码量(千行)』和『文件新增/修改总数(个)』指标、以及反映与框架耦合度的『框架代码量(行)』和『框架文件新增/修改数(个)』指标,通过删除代码的方法,统计以上指标,结果如下:(如有问题,可以直接在该 issue 下回复)

  1. Paddle Lite:Support resnet-50 model on Paddle Lite TangYuan-Liu/Paddle-Lite#1
    • 示例硬件:华为麒麟 NPU
    • 是否基于子图支持 CPU+NPU 异构?:✓
    • 总代码量(千行):1.7
    • 文件新增/修改总数(个):20
    • 框架代码量(行):19
    • 框架文件新增/修改数(个):3
  2. Tengine-lite:Supports resnet-50 model on tengine-lite csy0225/Tengine#1
    • 示例硬件:瑞芯微 NPU(TIM-VX)
    • 是否基于子图支持CPU+NPU异构?:✓
    • 总代码量(千行):3.0
    • 文件新增/修改总数(个):20
    • 框架代码量(行):18
    • 框架文件新增/修改数(个):2
  3. TNN:Supports resnet-50 model on tnn csy0225/TNN#1
    • 示例硬件:华为麒麟 NPU
    • 是否基于子图支持CPU+NPU异构?:✓
    • 总代码量(千行):2.6
    • 文件新增/修改总数(个):24
    • 框架代码量(行):46
    • 框架文件新增/修改数(个):3
  4. Tensorflow Lite: Supports resnet-50 model on Tensorflow Lite zhupengyang/tensorflow#1
    • 示例硬件:高通 DSP
    • 是否基于子图支持 CPU+NPU 异构?:✓
    • 总代码量(千行):4.6
    • 文件新增/修改总数(个):36
    • 框架代码量(行):1
    • 框架文件新增/修改数(个):1
  5. MACE: Supports resnet-50 model on mace zhupengyang/mace#2
    • 示例硬件:联发科 APU
    • 是否基于子图支持 CPU+NPU 异构?:✓
    • 总代码量(千行):2.5
    • 文件新增/修改总数(个):36
    • 框架代码量(行):123
    • 框架文件新增/修改数(个):17
  6. MNN: Supports resnet-50 model on MNN shentanyue/MNN#1
    • 示例硬件:华为麒麟 NPU
    • 是否基于子图支持 CPU+NPU 异构?:✓
    • 总代码量(千行):1.8
    • 文件新增/修改总数(个):18
    • 框架代码量(行):14
    • 框架文件新增/修改数(个):1
  7. Mindspore-Lite:Support resnet-50 model on Mindspore-Lite. TangYuan-Liu/mindspore#1
    - 示例硬件:华为麒麟 NPU
    - 是否基于子图支持 CPU+NPU 异构?:✓
    - 总代码量(千行):5.2
    - 文件新增/修改总数(个):48
    - 框架代码量(行):116
    - 框架文件新增/修改数(个):5
@paddle-bot-old
Copy link

您好,我们已经收到了您的问题,会安排技术人员尽快解答您的问题,请耐心等待。请您再次检查是否提供了清晰的问题描述、复现代码、环境&版本、报错信息等。同时,您也可以通过查看官网文档常见问题历史Issue来寻求解答。祝您生活愉快~

Hi! We've received your issue and please be patient to get responded. We will arrange technicians to answer your questions as soon as possible. Please make sure that you have posted enough message to demo your request. You may also check out the APIFAQ and Github Issue to get the answer.Have a nice day!

@hong19860320 hong19860320 changed the title 各大框架硬件适配代码量统计 各大推理框架硬件适配代码量统计 Jan 24, 2022
@wzyforgit
Copy link

在龙芯上面在坐的各位都是蜗牛。。。或者都编不过

@hong19860320
Copy link
Collaborator Author

在龙芯上面在坐的各位都是蜗牛。。。或者都编不过

龙芯是通过paddleinference支持的哈,不是paddle lite呢https://www.paddlepaddle.org.cn/documentation/docs/zh/guides/09_hardware_support/hardware_info_cn.html#paddle-inference

@hong19860320
Copy link
Collaborator Author

在龙芯上面在坐的各位都是蜗牛。。。或者都编不过

一般各大框架没有针对mips单独写kernel进行优化,走的都是c++ reference的实现,所以会比较慢的。

@qili93 qili93 closed this as completed Feb 5, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants