We started work on hyper-parameter tuning last year. This is handled by torchapp using optuna or scikit-optimize. I was fixing the number of MACC operations in the model and choosing a size for the initial features in the first CNN according to the other parameters chosen for that run in the hyper-parameter tuning session. We should try this again and try it for the diffusion model.
We started work on hyper-parameter tuning last year. This is handled by torchapp using optuna or scikit-optimize. I was fixing the number of MACC operations in the model and choosing a size for the initial features in the first CNN according to the other parameters chosen for that run in the hyper-parameter tuning session. We should try this again and try it for the diffusion model.