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
Is your feature request related to a problem? Please describe.
There is limited support currently within minerva for proper downstream validation of a model as it is being fitted. This is particularly problematic for self-supervised training as the current Weighted KNN validation is a poor test of the model's true performance.
Describe the solution you'd like
A custom task that extracts the encoder, turns its gradients off, builds a model around the encoder then trains this model on a downstream task for several epochs, then tests it with separate validation data and logs the result.
Describe alternatives you've considered
Continue to rely on the Weighted KNN.
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem? Please describe.
There is limited support currently within
minerva
for proper downstream validation of a model as it is being fitted. This is particularly problematic for self-supervised training as the current Weighted KNN validation is a poor test of the model's true performance.Describe the solution you'd like
A custom
task
that extracts the encoder, turns its gradients off, builds a model around the encoder then trains this model on a downstream task for several epochs, then tests it with separate validation data and logs the result.Describe alternatives you've considered
Continue to rely on the Weighted KNN.
The text was updated successfully, but these errors were encountered: