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
Hi! I would like to do hyperparameter search for an ABSA model (polarity only to be precise). Since there is no hyperparameter_search method for AbsaTrainer, I do this like in the code below, however, I am not certain if I do this right/optimally.
Are sentence transformer weights re-initialized from scratch at the start of each trial? Should I set force_download=True in AbsaModel?
When I only train polarity and try evaluating results with trainer.evaluate(eval), I get the following error: NotFittedError: This LogisticRegression instance is not fitted yet. Call 'fit' with appropriate arguments before using this estimator. However, the model clearly has a trained LR head and model.predict(eval) works just fine.
Less of a technical question, but would it make sense to try different multi-target strategies for a 3-way classification (positive, neutral, negative)? If so, would I be correct to instantiate classes inside objective like this:
Hi! I would like to do hyperparameter search for an ABSA model (polarity only to be precise). Since there is no
hyperparameter_search
method forAbsaTrainer
, I do this like in the code below, however, I am not certain if I do this right/optimally.Some questions:
force_download=True
in AbsaModel?trainer.evaluate(eval)
, I get the following error:NotFittedError: This LogisticRegression instance is not fitted yet. Call 'fit' with appropriate arguments before using this estimator.
However, the model clearly has a trained LR head and model.predict(eval) works just fine.objective
like this:The text was updated successfully, but these errors were encountered: