Hi!
I am interested in further post-training the released LongLive model and was wondering if you could share the checkpoint of the critic model (the fake score network used in DMD training). This would greatly help ensure training stability and reproducibility.
Thanks for your great work!