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
{{ message }}
This repository has been archived by the owner on Sep 18, 2024. It is now read-only.
What would you like to be added: Add an optional field portList to configurations when using PAI mode.
Why is this needed: I want to use see my tensorboard on PAI. Currently I follow the practice below: launch a tensorboard in the script submitted to PAI in the background, then launch my training. However, I cannot access my tensorboard without a open port. PAI has provided an option, so that I can access the port using something like $PAI_PORT_LIST_main_0_tensorboard, but this option is not accessible from the interface that NNI has provided.
Without this feature, how does current nni work: Indeed we can find a workaround like downloading the event files and view it locally; but I can imagine adding this feature will provide great convenience for not only tensorboard, but also visdom and other web-dashboard-based tools.
Components that may involve changes: trainingServices logic, configuration settings.
Brief description of your proposal if any:
The text was updated successfully, but these errors were encountered:
What would you like to be added: Add an optional field
portList
to configurations when using PAI mode.Why is this needed: I want to use see my tensorboard on PAI. Currently I follow the practice below: launch a tensorboard in the script submitted to PAI in the background, then launch my training. However, I cannot access my tensorboard without a open port. PAI has provided an option, so that I can access the port using something like
$PAI_PORT_LIST_main_0_tensorboard
, but this option is not accessible from the interface that NNI has provided.Without this feature, how does current nni work: Indeed we can find a workaround like downloading the event files and view it locally; but I can imagine adding this feature will provide great convenience for not only tensorboard, but also visdom and other web-dashboard-based tools.
Components that may involve changes: trainingServices logic, configuration settings.
Brief description of your proposal if any:
The text was updated successfully, but these errors were encountered: