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
Could the extension use multiple optimizers at the same time instead of averaging the results of the scorers? You could have 1 optimizer per payload, or 1 optimizer per image, or an arbitrary number of optimizers maybe?
There needs to be a way of conciliating the optimizer predictions before merging a model I think.
I am not sure this is a good idea, but I wanted to suggest it in case it could enhance the quality of merges.
The text was updated successfully, but these errors were encountered:
ljleb
changed the title
Ensemble of optimizers
[Feature Request] Ensemble of optimizers
Jun 14, 2023
Hyperband could be used to conciliate the optimizer predictions. Run multiple optimizers in parallel continuously feeding the results to each other. At every n steps cut the number of optimizers in half, but double the payload number. Simple, but it should work.
Could the extension use multiple optimizers at the same time instead of averaging the results of the scorers? You could have 1 optimizer per payload, or 1 optimizer per image, or an arbitrary number of optimizers maybe?
There needs to be a way of conciliating the optimizer predictions before merging a model I think.
I am not sure this is a good idea, but I wanted to suggest it in case it could enhance the quality of merges.
The text was updated successfully, but these errors were encountered: