wenqsun / DimensionX

DimensionX: Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusion
Apache License 2.0
832 stars 49 forks source link

Please provide GPU requirements! #21

Open jikkuatwork opened 2 days ago

jikkuatwork commented 2 days ago

Amazing work guys, but why most projects totally skip the hardware requirements needed to run? Isn't it the most important information for a project that can be run locally?

chenshuo20 commented 2 days ago

Thank you for your advice! In my tests on an A100 GPU, the VRAM usage is approximately 26.38GB. However, you can reduce it to around 17GB by using the offloading method with the following code after loading the pipeline:

pipe.enable_model_cpu_offload()
jikkuatwork commented 2 days ago

Thanks a lot for posting it! Will wait to test till it can run on poor GPU (12GB) ... Best wishes guys!

chenshuo20 commented 2 days ago

You can also test it using the Hugging Face online demo, available here.

jikkuatwork commented 2 days ago

Sadly too much traffic; wait time it over 2 hours :(

mp3pintyo commented 1 day ago

Then use my template on Runpod.io. https://runpod.io/console/deploy?template=8pqb3dg7lq&ref=2pdhmpu1 1 video generate time: 8 min