search

JoePenna / Dreambooth-Stable-Diffusion

Implementation of Dreambooth (https://arxiv.org/abs/2208.12242) by way of Textual Inversion (https://arxiv.org/abs/2208.01618) for Stable Diffusion (https://arxiv.org/abs/2112.10752). Tweaks focused on training faces, objects, and styles.
MIT License
3.19k stars 558 forks source link

Fix Docker image name in README.md for RunPod execution #197

Closed lawrence-law closed 9 months ago

lawrence-law commented 1 year ago

An obvious fix to the Docker image name for the recommended PyTorch 1.13.1 version in RunPod.

As an aside, do you know if the PyTorch 2.0.0 and 2.0.1 versions require more VRAM? I hit a CUDA out of memory issue using runpod/pytorch:3.10-2.0.1-120-devel on a 24GB VRAM pod (RTX 3090 or A5000; can't quite remember which).

Do you have any suggestion on when to use which Docker image/PyTorch image?

Edit: it's suggested in https://github.com/JoePenna/Dreambooth-Stable-Diffusion/issues/194 that perhaps 1.13.1 consumes less VRAM, albeit in a different environment.

yushan777 commented 1 year ago

I had tested the following in June 2023 with a 3090, and all trained okay.

runpod/pytorch:3.10-2.0.1-120-devel
runpod/pytorch:3.10-2.0.1-118-runtime
runpod/pytorch:3.10-2.0.0-117
runpod/pytorch-3.10-1.13.1-116
lawrence-law commented 1 year ago

Strange - wonder if I got a pod with some kind of strange issue. Anyway, hopefully someone can accept the PR suggestion since the typo tripped me up.