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

[BUG] Save it with wanlab.log({"plot": swanlab.Image(plt)}) and load it all the time #667

Open
Puiching-Memory opened this issue Aug 9, 2024 · 6 comments
Labels
🐛 bug Something isn't working

Comments

@Puiching-Memory
Copy link
Contributor

I save a copy of the image at the end of each epoch using wanlab.log({"plot": swanlab.Image(plt)})
But the image part of the panel keeps saying "loading".

img_v3_02dj_e44e8b0e-9b79-4530-8d8e-846267e6160g

When I look at the request report, I find that plot is also refreshed once with each request.
I guess the reason is that the image didn't load in such a short time and was interrupted by the next request.

Supplement: I am using ssh and port forwarding

@Puiching-Memory Puiching-Memory added the 🐛 bug Something isn't working label Aug 9, 2024
@Zeyi-Lin
Copy link
Member

Let's take a look at that. Thanks for the feedback

@SAKURA-CAT
Copy link
Contributor

Your anticipation is correct; we have already completed the refactoring and fixing of the chart section in the cloud version. However, due to staffing issues, the local version still requires some time for refactoring 🥲

@SAKURA-CAT
Copy link
Contributor

Your anticipation is correct; we have already completed the refactoring and fixing of the chart section in the cloud version. However, due to staffing issues, the local version still requires some time for refactoring 🥲

By the way, we have split the local version of Swanboard: https://github.com/SwanHubX/SwanLab-Dashboard

As the content of the local version is largely derived from early concepts and differs significantly from the current architecture, we may need some time to complete its refactoring.

@Puiching-Memory
Copy link
Contributor Author

感谢Debug:

据你们的信息,似乎本地和云端使用的是两套不同的前端代码.
这其中是否存在一些历史原因,或者考量,导致不能使用统一的代码?

我也注意到了:
#586
提到的颜色修改功能,在本地版本中并不存在.

期待你们能完成前端重构和统一.
据我了解,存在一些前端开发框架,例如https://dcloud.io/index.html
可以实现一套代码打包到各个平台,这使得部署到移动端成为可能

@SAKURA-CAT
Copy link
Contributor

SAKURA-CAT commented Aug 15, 2024

感谢Debug:

据你们的信息,似乎本地和云端使用的是两套不同的前端代码. 这其中是否存在一些历史原因,或者考量,导致不能使用统一的代码?

我也注意到了: #586 提到的颜色修改功能,在本地版本中并不存在.

期待你们能完成前端重构和统一. 据我了解,存在一些前端开发框架,例如dcloud.io/index.html 可以实现一套代码打包到各个平台,这使得部署到移动端成为可能

核心在于,swanlab分为本地版(对应init(mode="local"))和云端版本(对应init(mode="cloud")),两者在实验指标数据的收集原理上不太相同,前者使用本机文件系统,后者使用了我们自己写/部署的诸多服务。此外,云端版本在设计上功能也更多(涉及协作等),请求的api也不一样,因此我们没有办法使用同一套前端代码。
事实上,在上一个版本,我们复用了前端的“图表部分”代码——这类似于一个依赖。但是在我重构的时候发现两者的区别还是太大,因此在当前(下一个)版本,我们单独维护了线上版本的前端图表。
至于框架,由于我们主要维护web端,因此移动端可以通过浏览器访问我们的前端界面,因此框架问题并不是十分棘手。
不过总而言之,从技术上,本地和云端两者出现体验上的区别的最大原因在于我们一开始只有本地版本,云端是后来居上的(并且应该在未来的一段时间内我们的主要精力都在云端上),因此本地版本在一开始有许多部分考虑并不是十分周到。
我们会完成对本地版本的重构,寻求高效的维护方式,不过应该不是现在(因为云端有许多功能还在迭代中)😀

@Puiching-Memory
Copy link
Contributor Author

Puiching-Memory commented Aug 31, 2024

感谢你们的工作!
目前我选择swanlab而不是wandb的很大原因是因为:
1.swanlab可以支持本地保存记录
2.swanlab可以支持本地查看记录(wandb似乎必须上传到云端)

为什么不是tensorborad?
1.相比之下,swanlab启动配置简洁优雅,我们不需要很多时间即可享受可视化过程

在我们的工作服务器上,一般无法直接访问外部网络,而且我们的数据也涉及到隐私性.
今天我尝试使用了云端版本的swanlab,但是因为网络限制,无法连接 :(

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
🐛 bug Something isn't working
Projects
None yet
Development

No branches or pull requests

3 participants