tf.summary.text option to control whether text is treated as markdown

[drdozer]

I'm exporting text summaries from keras into tensorboard. The text contains various markdown sequences, but I don't intend this to be interpreted as markdown. However, tensorboard insists on trying to treat it as markdown. In the best case this gives entertainingly broken rendering in the text viewer, and worst case it causes the server to bomb in the text plugin. The text I'm exporting should never be treated as markdown. It isn't markdown.

Can we please have some way to tag a text summary as being not markdown? Say::

tf.summary.text('sometext', text_tensor, step, is_raw_text=True)

And then have this communicated through to tensorboard, and have tensorboard not attempt to treat it as markdown?

[nfelt]

Duplicate of #830

[nfelt]

Thanks for the request; there is an existing very similar feature request at #830, except that it talks about making this an option in the UI rather than an option at the summary write time.

Would a UI option work in your case? It's a bit more flexible and more easily compatible with cases when the tf.summary call is inside a library or some other place that isn't directly in control of the user.

[nfelt]

FWIW, #5378 implemented a UI option to toggle markdown rendering off (sticky for a given hostname/port). I'll rescope this FR to be just about having a summary-level control, which could still be helpful if users want either a mix of markdown and not-markdown, or want to ensure the data is always treated a certain way regardless of the setting's default value.