Closed KristinaUlicna closed 1 year ago
Intriguingly, this is the same run but its tf.events
file was opened from:
VSCode launch option | Tensorboard command line |
---|---|
As can be seen, the command line launch unwrap the entire run (full 500 epochs), so it seems better to use that option for future run monitoring. Instructions below 👇
Describe the bug
When GCN is trained for more than, say, 100 epochs, Tensorboard appears to save all the info but fails to render it correctly:
⬆️ There is no figure to visualise the loss curves, for example...
Python version
No response
What OS are you using?
No response
Comments
Possibly linked to #97