Use GET instead of POST requests for hparams in colab environment. #4218
Conversation
Read the 'tensorboardColab' environment variable as an indicator of whether tensorboard is running in a colab environment. If it is true then send GET requests instead of POST requests. This check replaces an old colab-related check that stopped functioning. The new 'tensorboardColab' mechanism was introduced in tensorflow#4191.
| const inColab = | ||
| new URLSearchParams(window.location.search).get('tensorboardColab') === |
There was a problem hiding this comment.
Could you leave a comment here that this is deliberately a side-effect
at module import time to work around the bug in AppRouting, as #4213
did for the scalars dashboard?
There was a problem hiding this comment.
(I’ve gone ahead and done this.)
|
All (the pull request submitter and all commit authors) CLAs are signed, but one or more commits were authored or co-authored by someone other than the pull request submitter. We need to confirm that all authors are ok with their commits being contributed to this project. Please have them confirm that by leaving a comment that contains only Note to project maintainer: There may be cases where the author cannot leave a comment, or the comment is not properly detected as consent. In those cases, you can manually confirm consent of the commit author(s), and set the ℹ️ Googlers: Go here for more info. |
|
@googlebot I consent. |
|
CLAs look good, thanks! ℹ️ Googlers: Go here for more info. |
3 participants
Motivation for features / changes
Hparams dashboard does not work in Google-internal colab environments because they do not support POST requests.
Technical description of changes
Reuse mechanism introduced in scalars: don’t multiplex fetches in Colab #4191 that provides a 'tensorboardColab' query parameter in the URL to indicate a colab environment. If the value for the parameter is true then the Hparams dashboard will send GET requests. Otherwise it sends POST requests.
There already existed code that purported to determine if the hparams dashboard was running in a colab environment but this code has not worked for some time, possibly since colab: use proxyPort for dynamic plugin #2798. The code in this PR simply replaces that code. The plumbing to pipe this through other layers had already been implemented.
Tests
I loaded a local tensorboard with the query parameter 'tensorboardColab=true' and inspected the network tab to confirm that hparams-related calls were GET requests.
I loaded the local tensorboard without 'tensorboardColab' in the URL and inspected the network tab to confirm that hparams-related calls were POST requests.