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
Currently the main bottleneck in the training phase is fitting the scikit-learn LogisticRegression models, not the actual fitting of the reporters. This is because LogisticRegression is done on the CPU. But we could easily replicate ~the same results by fitting Reporter objects with a purely supervised loss, and this should speed up the training process dramatically.
The text was updated successfully, but these errors were encountered:
Currently the main bottleneck in the training phase is fitting the scikit-learn
LogisticRegression
models, not the actual fitting of the reporters. This is becauseLogisticRegression
is done on the CPU. But we could easily replicate ~the same results by fittingReporter
objects with a purely supervised loss, and this should speed up the training process dramatically.The text was updated successfully, but these errors were encountered: