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

Web UI: Some bugs in graph display #136

Closed
chrishkchris opened this issue Aug 28, 2020 · 6 comments
Closed

Web UI: Some bugs in graph display #136

chrishkchris opened this issue Aug 28, 2020 · 6 comments

Comments

@chrishkchris
Copy link
Contributor

When we look at the curve below, there are two errors:

  1. When the x data start from 1, the plot should also start at the same point (x=1). This is because the loss at epoch 0 is not 0, since there is no data it displays 0 (this is highly confusing to user).

  2. Y-axis is named "Loss Over Epochs", but actually there are two curves for loss and accuracy. In this case the Y-axis can be renamed as Value instead.

image

@chrishkchris
Copy link
Contributor Author

Also, since loss and accuracy are in a different scale, could consider to separate them in two graphs

@arielclj
Copy link
Contributor

arielclj commented Sep 1, 2020

@chrishkchris For this example, by having the plot start @ x=1 would mean that the graph presents a data point instead of a line. Is this the refinement you had in mind?

@chrishkchris
Copy link
Contributor Author

in usual case, there are 5-50 epochs so it appears as a line.
It is not common to train for only 1 epoch.
I think 1 epoch is only for test purpose
When there is only 1 epoch, I also prefer using a point instead of line (as this is not misleading)

@arielclj
Copy link
Contributor

arielclj commented Sep 1, 2020

SS2

For this trial, there is a point 0- epoch: 0, epoch_accuracy: 0, loss: 0.

In this scenario, would you want the point 0 to be plotted?

@chrishkchris
Copy link
Contributor Author

chrishkchris commented Sep 1, 2020

SS2

For this trial, there is a point 0- epoch: 0, epoch_accuracy: 0, loss: 0.

In this scenario, would you want the point 0 to be plotted?

No, loss should not be equal to 0 at epoch 0. This is the exact reason I raised the point 1

As you see the graph is correct except x=0, this is why we need to skip x = 0

NLGithubWP added a commit that referenced this issue Sep 7, 2020
@chrishkchris
Copy link
Contributor Author

solved

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

2 participants