Did you use the pretrained 14B ODE checkpoint in casual-vid-plus repo, to train the self-forcing DMD model?

#11
by lj1995 - opened

How can we train the OED init checkpoint in the self forcing repo?

We directly use the original Wan2.1 model for dmd distillation without ode initialization.

Thanks!

lj1995 changed discussion status to closed

Sign up or log in to comment