Hyperparameter Optimization with DDP #20043
Unanswered
gojkopero
asked this question in
DDP / multi-GPU / multi-node
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Is there any reliable HPO framework that works with Lightning DDP currently? Optuna seems to work only with ddp_spawn which has horrendously slow initialization. Lightning PyTorch Training Studio App is outdated, and Ray[Tune] guide as well.
Beta Was this translation helpful? Give feedback.
All reactions