What is the recommended way to recreate the sac:pixel from the results? Modify Pytorch_sac, or sac_ae, or is code available? (I know changes are trivial in some sense but I'd just like to keep the implementation as similar as possible)
You can set --decoder_update_freq to some large number, ideally larger than the total number of training steps. This will prevent decoder from being updated ever.
What is the recommended way to recreate the sac:pixel from the results? Modify Pytorch_sac, or sac_ae, or is code available? (I know changes are trivial in some sense but I'd just like to keep the implementation as similar as possible)