New features added to the WaveDiff model. Now it can be optimised using any number of cycles. Options for the new projected DD features scenario are added.
New click options:
--save_all_cycles: Make checkpoint at every cycle or just save the checkpoint at the end of the training.
--project_dd_features: Project NP DD features onto parametric model.
--eval_only_param: Use only the parametric model for evaluation.
--pretrained_model: Path to pretrained model checkpoint callback.
--reset_dd_features: Reset to random initialisation the non-parametric model after projecting.
New features added to the WaveDiff model. Now it can be optimised using any number of cycles. Options for the new projected DD features scenario are added.
New click options:
--save_all_cycles: Make checkpoint at every cycle or just save the checkpoint at the end of the training.--project_dd_features: Project NP DD features onto parametric model.--eval_only_param: Use only the parametric model for evaluation.--pretrained_model: Path to pretrained model checkpoint callback.--reset_dd_features: Reset to random initialisation the non-parametric model after projecting.