search

cloud-native-robotz-hackathon / infrastructure

0 stars 1 forks source link

Add Stable Diffusion Image Generation to DataScience Part #65

Open nexus-Six opened 2 months ago

nexus-Six commented 2 months ago

Allow the participants to create an avatar for their team with Stable Diffusion running on RHOAI. Samples here :
https://ai-on-openshift.io/demos/stable-diffusion/stable-diffusion https://github.com/opendatahub-io/distributed-workloads/tree/main/examples/stable-diffusion-dreambooth

nexus-Six commented 2 months ago

Got this to work in this notebook : https://github.com/cloud-native-robotz-hackathon/object-detection-notebooks/blob/master/generate-team-avatar.ipynb

based on this post: https://developers.redhat.com/articles/2024/09/04/use-stable-diffusion-create-images-red-hat-openshift-ai-rosa-cluster-gpu#2__setting_up_gpu_environment

Due to the lack of GPU, it takes roughly 30 min per image and they don't look that great yet. Need to evaluate if we want to include this or if it can be enhanced.

Image

nexus-Six commented 2 months ago

This notebook modifies the python packages which seem to prevent the Elyra Pipeline from starting after that. Needs to be investigated. We probably need a separate Workbench for this.

nexus-Six commented 1 month ago

Need to retest with new GPU-Node