metadata
license: mit
tags:
- text-to-image
- diffusion
- from-scratch
- pokemon
base_model: openai/clip-vit-base-patch32
LouV1GPU
T2I diffusion from scratch (VAE+UNet+DDIM, torch pur). Texte = openai/clip-vit-base-patch32 gele + projection. Dataset diffusers/pokemon-gpt4-captions. Res 64->2048. Classe Lou: train/generate/display/push.