I am in google colab and I can not run well SVD, although in this other colab free if you can run but it has the watermark stability filter and nsfw that bothers a lot, maybe you can get things from here to reduce vram usage, in my little knowledge with models I think a good option is to load the model in float16, but I better leave to those who know to do it.
I am in google colab and I can not run well SVD, although in this other colab free if you can run but it has the watermark stability filter and nsfw that bothers a lot, maybe you can get things from here to reduce vram usage, in my little knowledge with models I think a good option is to load the model in float16, but I better leave to those who know to do it.
colab mentioned: https://colab.research.google.com/github/mkshing/notebooks/blob/main/stable_video_diffusion_img2vid.ipynb