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

tf.summary.text support for styling/annotating words #1740

Open
silky opened this issue Jan 3, 2019 · 6 comments

Comments

Projects
None yet
3 participants
@silky
Copy link

commented Jan 3, 2019

see #830

it would be great to log summary text that can produce highlighing like so:

image

i.e. some markdown support for something like

Optimising [quantum]{"color: #ff0000aa"} [whatever]{"color: #ff000000"} ...

@nfelt nfelt changed the title add markdown support for mild styling of words for attention models tf.summary.text support for styling/annotating words Mar 28, 2019

@nfelt

This comment has been minimized.

Copy link
Collaborator

commented Mar 28, 2019

Generalizing this issue to encompass styling words not just to indicate attention but also other use cases like tagging words. It seems like at a minimum, this requires some way to highlight text with a custom color value. If you think this requires more than just highlighting (versus what could be done by preprocessing the text and using other markdown features that are already supported), please chime in.

@silky

This comment has been minimized.

Copy link
Author

commented Mar 28, 2019

i think it's unfortunate to not allow html styling. i understand that google may not want that, but i'd say that most companies actually don't have the same security concerns as google.

so i think a fine suggestion is just a plugin, that is off by default, but enable-able via the command line, that allows this, would be great.

alternatively, it seems your only other option is to add a new method, like tf.summary.text_highlighted, to specialise it to all the cases you deem appropriate.

🤷‍♀

for what it's worth, i just worked around it ages ago, by doing highlighting in an image, and logging the image instead.

@nfelt

This comment has been minimized.

Copy link
Collaborator

commented Mar 28, 2019

Understood that there's still interest in the full flexibility and attendant risk of allowing arbitrary HTML.

For what it's worth, we are tentatively working on making it easier to develop/distribute/install third-party plugins that don't require any approval/vetting from the core team, so it's quite possible that other parties who have a different sense of the security/convenience tradeoff could create a plugin for arbitrary HTML rendering, even if it doesn't get merged into core TensorBoard.

@silky

This comment has been minimized.

Copy link
Author

commented Jul 2, 2019

what ever happened here? anthing? @nfelt

@wchargin

This comment has been minimized.

Copy link
Member

commented Jul 2, 2019

@silky: Nope, there hasn’t been any work on this issue.

@nfelt

This comment has been minimized.

Copy link
Collaborator

commented Jul 10, 2019

FWIW, there has been work on the second paragraph of my previous comment, for third-party dynamic plugins: https://github.com/tensorflow/community/blob/master/rfcs/20190411-tensorboard-improved-plugin-ext.md

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.