Views
No views yet
1from lua import load_model, upscale_latent
2
3# Auto-downloads this checkpoint
4lua_model = load_model(device="cuda")
5
6# Upscale a FLUX VAE latent (B, 16, H, W)
7upscaled = upscale_latent(lua_model, latent, head="x2") # or "x4"| Architecture | SwinIR with multi-head upsampling |
| Parameters | ~250M |
| Input | 16-channel VAE latent (FLUX latent space) |
| Heads | x2 (2x upscaling), x4 (4x upscaling) |
| Training | Trained on high-quality image pairs in FLUX latent space |
| Precision | float32 (recommended), bfloat16 compatible |
1@article{razin2024lua,
2 title={One Small Step in Latent, One Giant Leap for Pixels: Fast Latent Upscale Adapter for Your Diffusion Models},
3 author={Razin, Aleksandr and Kazantsev, Danil and Makarov, Ilya},
4 journal={arXiv preprint arXiv:2511.10629},
5 year={2024}
6}