Repository navigation
Feature request: improve (repeated) TensorBoard startup #15476
Description
After starting up VS Code, I regularly start up TensorBoard. Two things are annoying that do not need to be:
- The prompt for the log directory could be omitted if the log directory could be remembered (or even better, configured similar to
python.pythonPath). - The process could be made persistent*. I have a large log-file directory, and initial parsing takes several minutes until all data are available. I would not mind a TensorBoard process sticking around (and being reused) after closing (and restarting) VS Code.
(*) I usually do this from Python using
with open(log_file, "w") as stdout:
proc = subprocess.Popen(argv, stdout=stdout, stderr=subprocess.STDOUT)
Activity
joyceerhl commentedon Feb 25, 2021
Thanks for the suggestions bersbersbers, we definitely want to do the first one. Our team will discuss prioritization for both.
joyceerhl commentedon Mar 8, 2021
bersbersbers, would the python.tensorBoard.logDirectory setting proposed in #15588 be what you are looking for with your first request?
bersbersbers commentedon Mar 8, 2021
Looks very much like it, yes. Thank you!
joyceerhl commentedon Mar 8, 2021
Great, thanks for your feedback 😊
joyceerhl commentedon Mar 12, 2021
bersbersbers we've added
python.tensorBoard.logDirectoryas a user-level setting (since there's not a good way to pick the right workspace settings if the current workspace is a multiroot workspace). This will become available in the March release of the Python extension. Thanks again for the feature request!bersbersbers commentedon Mar 12, 2021
Thanks! Is a persistent process covered someplace else, or is that off the table for now?
joyceerhl commentedon Mar 12, 2021
Oh yes, my apologies. Just created a discussion thread for that here: #15658