search

nateraw / stable-diffusion-videos

Create 🔥 videos with Stable Diffusion by exploring the latent space and morphing between text prompts
Apache License 2.0
4.36k stars 417 forks source link

Segmentation Fault when running run_app.py on GPU-backed Ubuntu server #117

Closed eolszewski closed 1 year ago

eolszewski commented 1 year ago

Was going to submit a PR to support a newer version of diffusers but got the following after cloning, pip3 installing, and attempting to run.

eolszewski@sdtest-2:~/stable-diffusion-videos/examples$ python3 run_app.py Segmentation fault

I have the package on the same server and working fine. Not sure what the issue might be...

nateraw commented 1 year ago

Any more logs available? did you run out of memory by chance?

eolszewski commented 1 year ago

Nothing more than Segmentation fault

I did see that it crashes on this import from stable_diffusion_videos import StableDiffusionWalkPipeline, Interface

This looks like it might be a local issue of sorts, hopefully I can resolve it soon and anyone who gets Segmentation Fault in the future will be able to use this as a resource (this is me asking you to not close this, just yet!).

nateraw commented 1 year ago

Any update here?

eolszewski commented 1 year ago

Sorry, yes, I don't remember the finer details but I was having issues with GCP instances failing to run the nvidia installer on boot if they were created with a machine image (even on machine restart), so, I had to find these scripts and run them manually. This and a number of other issues caused me to move off GCP to lambdalabs (much cheaper!) and I no longer had this issue. So, I don't recall exactly what was happening, but I imagine it to having been a combination of a few niche issues with Google's platform.

nateraw commented 1 year ago

ah yea, been there. The classic sudo python3 install_gpu_driver.py. They actually tell you to run that when ya first boot the machine, but sometimes (depending on how you ssh in) it can be missed.

Anyways, glad it worked out!