You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Iris-3B: pixel-space generation & general vision learner. A 3B diffusion transformer that generates every pixel directly (no VAE); the same prior, fine-tuned, does monocular depth estimation and 4x image restoration. Paper, weights, training code and demo.
Linum's P-JiT 21K natively in ComfyUI: a 2B pixel-space text-to-image model with no VAE, on ComfyUI's own KSampler and memory management. bf16 and int8 ConvRot files, a refiner workflow, Linum's samples seed for seed.
P-JiT 21K on Apple silicon (MLX): exact port with seed parity to Linum's H100 gallery, about 2x PyTorch speed; the 8-bit tier runs in 3.6 GB (whole process)
Pixel-space diffusion with LLM-style mixture-of-experts in ~260 lines: no VAE, routed experts, hierarchical flow matching. Trains on one laptop GPU. Implements PixelMoE (arXiv:2610.18857).