how to use the model
the model seems not complete, how to use it ? replace the file svd_xt.safetensors in stabilityai/stable-video-diffusion-img2vid-xt ??
Yes, I just swapped them out and it worked.
Simply replacing the checkpoint works for me, but I am wondering why the model becomes smaller?
The original model files were released in FP32, but we're moving towards releasing more things in FP16 (the extra precision is redundant and doesn't change results). (Note that we're not stripping out precision data here - we just don't bother having it from the start. 1.1 was built with 16bit parameters... all fp32 does is waste vram).
is there a code on how to use this
is there a code on how to use this
Are you familiar with ComfyUI? If so, you can download the SVD workflow and use this model.
is your colab notebook similar to animatediff on comfyUI?
测试：生成视频视频，请问解决报错如下：
使用 EulerEDMSampler 进行 26 个步骤的采样：96%|████████████████████████████████████████ ████████████▉ | 25/26 [00:56<00:02, 2.27s/it]
2024-03-25 09:18:24.440 未捕获的应用程序异常
回溯（最近一次调用）：
文件“/usr/local/lib/python3.10 /dist-packages/streamlit/runtime/scriptrunner/script_runner.py”，第 542 行，在 _run_script
exec(code, module.dict)
文件“/home/generative-models/scripts/demo/video_sampling.py”，第 280 行，在
save_video_as_grid_and_mp4(samples, save_path, T, fps= saving_fps)
文件“/home/generative-models/scripts/demo/streamlit_helpers.py”中，第 912 行，在 save_video_as_grid_and_mp4
imageio.mimwrite(video_path, vid, fps=fps)
文件“ /usr/local/lib/python3.10/dist-packages/imageio/v2.py”，第 495 行，在 mimwrite
返回 file.write(ims, is_batch=True, **kwargs)
文件“/usr/local/lib /python3.10/dist-packages/imageio/plugins/tifffile_v3.py”，第 224 行，写入
self._fh.write(image, **kwargs)
TypeError：TiffWriter.write() 得到意外的关键字参数 'fps'