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

[PaddlePaddle Hackathon] add WideResNet #36952

Merged
merged 11 commits into from
Nov 11, 2021

Conversation

SigureMo
Copy link
Member

@SigureMo SigureMo commented Nov 2, 2021

PR types

New features

PR changes

APIs

Describe

status: Pending Review

Performance

AI Studio 测试详情:https://aistudio.baidu.com/aistudio/projectdetail/2561949
基准参考(torchvision):https://pytorch.org/vision/stable/models.html

Model Top-1 Top-5
wide_resnet50_2 0.78468(=) 0.94086(=)
wide_resnet101_2 0.78846(-0.00002) 0.94284(=)

括号中为以 torchvision 性能基准为参考的偏差值

由于题目要求中给出的预训练权重貌似是 CIFAR10 上预训练得到的,这与 paddle.vision 中其他模型均为 imagenet 预训练得到不同,因此本 PR 所使用的预训练权重为自行从 torchvision 转换得来(存储在 AI Studio 中),不过转换后的模型经自测精度损失相对于前几次高一点由于 OpenCV 与 Pillow 在 Resize 预处理时效果不一致,会导致精度降低,将预处理统一改成 Pillow 后精度与基准基本完全一致~

同时还对 ResNet 英文文档进行了一定的修改,以保持与中文文档一致~

@paddle-bot-old
Copy link

paddle-bot-old bot commented Nov 2, 2021

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

@SigureMo
Copy link
Member Author

SigureMo commented Nov 8, 2021

@LielinJiang 可以麻烦 review 一下这个嘛?

LielinJiang
LielinJiang previously approved these changes Nov 8, 2021
@SigureMo
Copy link
Member Author

SigureMo commented Nov 8, 2021

@LielinJiang 请问我自行转换的权重是否需要更换成 paddle 服务器上的链接呢?目前只是利用我个人 AI Studio 公开数据集生成的链接

pretrained (bool): If True, returns a model pre-trained on ImageNet

Examples:
.. code-block:: python
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

.. code-block 与 import 间需要有一个空行,否则会有奇怪的显示bug;上面的修改同理哈 都需要有的~

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

已修改~~~

@LielinJiang
Copy link
Contributor

@LielinJiang 请问我自行转换的权重是否需要更换成 paddle 服务器上的链接呢?目前只是利用我个人 AI Studio 公开数据集生成的链接

已经上传了,麻烦更新一下链接:https://paddle-hapi.bj.bcebos.com/models/wide_resnet101_2.pdparams,
https://paddle-hapi.bj.bcebos.com/models/wide_resnet50_2.pdparams

@SigureMo
Copy link
Member Author

SigureMo commented Nov 9, 2021

@LielinJiang 请问我自行转换的权重是否需要更换成 paddle 服务器上的链接呢?目前只是利用我个人 AI Studio 公开数据集生成的链接

已经上传了,麻烦更新一下链接:https://paddle-hapi.bj.bcebos.com/models/wide_resnet101_2.pdparams, https://paddle-hapi.bj.bcebos.com/models/wide_resnet50_2.pdparams

好哒~已更新~~~

@SigureMo
Copy link
Member Author

@XiaoguangHu01 可以麻烦 review 一下嘛?

Copy link
Contributor

@XiaoguangHu01 XiaoguangHu01 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

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

Successfully merging this pull request may close these issues.

4 participants