torch.OutOfMemoryError: CUDA out of memory ？

[juntaosun]

torch.OutOfMemoryError: CUDA out of memory. See documentation for Memory Management  (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

After testing, the input video is only 10 seconds. As expected, the GPU memory crashed. The model is based on Stable Video Diffusion (SVD), which requires a very high computer configuration.

For video processing, at least 24GB~48GB VRAM is required. What is a good solution?

[Nightmare-n]

We recommend reducing the num_frames, num_interp_frames, max_resolution, and decode_chunk_size to decrease memory usage.

[juntaosun]

For cartoon images, depth doesn't seem to work?

[juntaosun]

It is best to get rid of Stable Video Diffusion, which requires too much VRAM. The general computer configuration cannot run video inference.