Class-conditional Image Synthesis on ImageNet 256x256 (train test)
1.96FIDSiT-XL/2 + REPA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SiT-XL/2 + REPAEpochs=2002026.04 | 1.96 | 4.49 | 264 | 82 | 60 | |
| PFT-XL/2 + REPA + look-aheadEpochs=2002026.04 | 2 | 4.32 | 284.1 | 81 | 61 | |
| SiT-XL/2Epochs=14002026.04 | 2.06 | 4.5 | 270.3 | 82 | 59 | |
| DiT-XL/2Epochs=14002026.04 | 2.27 | 4.6 | 278.2 | 83 | 57 | |
| MaskDiTEpochs=16002026.04 | 2.28 | 5.67 | 276.6 | 80 | 61 | |
| SD-DiTEpochs=4802026.04 | 3.23 | — | — | — | — |