ComboStoc-XL/2 — Class-conditional ImageNet 256×256

Generate class-conditional 256×256 images with ComboStoc-XL/2 (SiT-style diffusion transformer, UNSYNC_ALL combinatorial stochasticity, 800K steps), decoded with the stabilityai/sd-vae-ft-mse VAE.

Pick an ImageNet class (or type any id 0–999), choose ODE / SDE sampling, classifier-free-guidance scale, number of steps, and a seed.

Weights: suvadityamuk/ComboStoc-XL-2-256x256 · License: AGPL-3.0 · Paper: arXiv:2405.13729 · Project page

ImageNet class
Well-known classes. Overridden by the class id box below when set.
1 8
Sampler
ODE = fixed-step Euler flow; SDE = Euler–Maruyama.
1 10
4 250
Examples