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
What would you like to be added:
Improve current reusable environment design, and cover more training platform and provide more rich features.
Why is this needed:
It not only refactoring on code, also speed up feature integration on training platform, stablize quality.
Support all training service, including local, remote, k8s, and others. In local mode, file channel is better, and some user asks not to request any port, so that they can run NNI in colab.
Support file command channel by auto detect, or configurable. So that the experiment can pass through firewall with central storage.
Show environment information in web ui.
Support to set max env limitation for platform like OpenPAI. If user has some soft limitation, which is asked by admin.
Support mixed environments for one experiment, for example, remote machines, local machine, several different openpai environmentsin a same experiment.
Design better configuration experience, like useActiveGpu, gpuIndices, maxTrialPerGpu.
Move metric data and request new parameters commands out from stdout message.
Consider to delay send non-important message like stdout, to reduce network traffic.
Support to check GPU utilization on web ui.
The text was updated successfully, but these errors were encountered:
What would you like to be added:
Improve current reusable environment design, and cover more training platform and provide more rich features.
Why is this needed:
It not only refactoring on code, also speed up feature integration on training platform, stablize quality.
The text was updated successfully, but these errors were encountered: