The distillation models related to qwen are all distilled based on the DMD2 algorithm. The distillation models related to video generation are based on improved training of self-forcing. I don’t know if this is correct.
· Sign up or log in to comment
