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
After creating my generation strategy and experiment, I'm able to receive the next recommended parameter values via get_next_trial. I've been trying to understand how candidates are being generated under the hood, and from my search it appears to use optimize_acqf from BoTorch (although correct me if I'm wrong). My question is: let's say I have some experimental design with predetermined candidates, and I want to evaluate only those candidates for potential selection. In BoTorch this is as simple as building an acquisition function over a trained model or models, and then evaluating the acquisition function over all the candidates. However, I'm having trouble finding a way to do this in the Ax framework. Is this currently supported? Thanks for your help.
The text was updated successfully, but these errors were encountered:
After seeing another issue posted here, I realized that the Torch Model Bridge has an evaluate_acquisition_function method, so my question is answered now. Thanks again
After creating my generation strategy and experiment, I'm able to receive the next recommended parameter values via get_next_trial. I've been trying to understand how candidates are being generated under the hood, and from my search it appears to use optimize_acqf from BoTorch (although correct me if I'm wrong). My question is: let's say I have some experimental design with predetermined candidates, and I want to evaluate only those candidates for potential selection. In BoTorch this is as simple as building an acquisition function over a trained model or models, and then evaluating the acquisition function over all the candidates. However, I'm having trouble finding a way to do this in the Ax framework. Is this currently supported? Thanks for your help.
The text was updated successfully, but these errors were encountered: