| Dataset | Code Resolution | # layers | # feature dim | # codebook | Top-k, Temperature | rFID (stage1) | FID (stage2) | Link |
|---|---|---|---|---|---|---|---|---|
| ImageNet (cIN) | 8x8 + 16x16 | 12 | 1536 | 8192 + 8192 | 2048, 0.95 | 2.61 | 9.36 | link |
| ImageNet (cIN) | 8x8 + 16x16 | 24 | 1536 | 8192 + 8192 | 2048, 0.95 | 2.61 | 8.46 | link |
| ImageNet (cIN) | 8x8 + 16x16 | 42 | 1536 | 8192 + 8192 | 2048, 0.95 | 2.61 | 7.15 | link |
| CC-15M | 8x8 + 16x16 | 12 | 1536 | 8192 + 8192 | 8192, 0.9 | 5.76 (CC3M) | 12.86 | link |
| FFHQ | 8x8 + 16x16 | 24 | 1024 | 8192 + 8192 | 4096, 1.0 | 5.53 | 10.21 | link |