You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When running tensorboard (1.12 or 1.14) on logs from an SFTP drive using sftp 2 driver, tensorboard doens't refresh the scalar display, and I see errors like
"[Thread-2 core_plugin.py:165] Unable to get first event timestamp for run ..."
Any idea how can I make it work?
Thanks
The text was updated successfully, but these errors were encountered:
I map the SFTP to a drive letter, say X:, and I start tensorboard with logdir=X:\logs.
It launches, but after a few minutes it stops showing new data.
I assume it's not receiving filesystem events.
I ended up using win-sshfs to map a network drive.
I'm not very familiar with windows SFTP mapping but it seems possible that the issue is similar to that with rsync, where by default the files are replaced rather than actually appended, so TensorBoard doesn't see the new data: #349
If there's a similar way to make SFTP reuse the same files, that might work.
Alternatively, if win-sshfs works for you instead, it sounds like perhaps the problem is resolved well enough?
Hi,
When running tensorboard (1.12 or 1.14) on logs from an SFTP drive using sftp 2 driver, tensorboard doens't refresh the scalar display, and I see errors like
"[Thread-2 core_plugin.py:165] Unable to get first event timestamp for run ..."
Any idea how can I make it work?
Thanks
The text was updated successfully, but these errors were encountered: