chavinlo

@lxj616 Thanks. One more question, in the preprocess function, you treat npz as if it had all the videos? because it itterates through it, and adds all the frames of...

@lxj616 Hello again, I got training working with bs 1 and 25 frames. Although I had to convert the model to BFloat16, because I got OOM with fp32 (80GB+), and...

Also, bf16 uses 44GB, but when using grad checkpointing, it decreases to 11GB

Training with 240 frames... it's really big. 74GB of VRAM usage WITH gradient checkpointing and bf16 

> > Training with 240 frames... it's really big. 74GB of VRAM usage WITH gradient checkpointing and bf16  > > For 13.73s/it , if you train on webvid-10m, you...

 This is after 20,000 steps. No upgrades Maybe you could post the script or how you initialized the extra layers? I saw that you use the inpainting model (9...

> Thank you for the reply. All logic lead me to the fact that I should learn it as well. Honestly it makes me a bit mad that I will...

It's been a while since I posted here. Since my last response, and after many trials and errors (and tremendous help by lopho) I got it trained on a small...

around 14GB with Xformers + 8Bit + Gradient Checkpointing + fp16 + resizing to 512

Thanks. I am asking because I plan on finetuning this on some data I got. I use A100s so memory won't be an issue.