Class-conditional Image Synthesis on ImageNet (with MACs reporting)
278.2Inception ScoreDiT-XL/2
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DiT-XL/2cfg=1.5, Sampler=DDPM, #Steps=2502025.05 | 278.2 | 2.27 | 4.6 | 83 | 57 | 59.31 | |
| ICC (SVD)r=256, p=2, Sampler=DDIM, #Steps=442025.05 | 258.7 | 2.3 | 4.31 | 83 | 57 | 7.09 | |
| ICC (SVD)r=256, p=2, Sampler=DDIM, #Steps=582025.05 | 257.8 | 2.14 | 4.29 | 82 | 58 | 9.35 | |
| ICC (SVD)r=256, p=2, Sampler=DDIM, #Steps=282025.05 | 257.1 | 2.66 | 4.47 | 82 | 56 | 4.51 | |
| ICC (CD-SVD)r=192, p=2, Sampler=DDIM, #Steps=462025.05 | 253 | 2.28 | 4.26 | 82 | 57 | 6.98 | |
| ICC (CD-SVD)r=192, p=2, Sampler=DDIM, #Steps=622025.05 | 251.7 | 2.1 | 4.25 | 81 | 59 | 9.4 | |
| ICC (CD-SVD)r=192, p=2, Sampler=DDIM, #Steps=302025.05 | 250.6 | 2.53 | 4.34 | 81 | 57 | 4.55 | |
| ICC (CA-SVD)r=128, p=2, Sampler=DDIM, #Steps=662025.05 | 244 | 2.23 | 4.35 | 80 | 59 | 9.4 | |
| Naive Cachingp=3, Sampler=DDIM, #Steps=902025.05 | 243.8 | 2.52 | 4.91 | 81 | 57 | 7.15 | |
| Naive Cachingp=2, Sampler=DDIM, #Steps=802025.05 | 243.4 | 2.29 | 4.46 | 81 | 59 | 9.51 | |
| Naive Cachingp=3, Sampler=DDIM, #Steps=1202025.05 | 242.7 | 2.33 | 4.62 | 81 | 59 | 9.53 | |
| ICC (CA-SVD)r=128, p=2, Sampler=DDIM, #Steps=502025.05 | 241.7 | 2.31 | 4.33 | 80 | 60 | 7.11 | |
| Naive Cachingp=2, Sampler=DDIM, #Steps=602025.05 | 240 | 2.48 | 4.6 | 81 | 59 | 7.13 | |
| DiT-XL/2cfg=1.5, Sampler=DDIM, #Steps=402025.05 | 239.9 | 2.35 | 4.26 | 80 | 59 | 9.49 | |
| ICC (CA-SVD)r=128, p=2, Sampler=DDIM, #Steps=322025.05 | 236.8 | 2.77 | 4.35 | 80 | 58 | 4.55 | |
| Naive Cachingp=2, Sampler=DDIM, #Steps=402025.05 | 235.4 | 2.95 | 4.93 | 80 | 57 | 4.75 | |
| DiT-XL/2cfg=1.5, Sampler=DDIM, #Steps=302025.05 | 235 | 2.64 | 4.38 | 80 | 59 | 7.12 | |
| Naive Cachingp=3, Sampler=DDIM, #Steps=602025.05 | 234.4 | 3 | 5.54 | 80 | 57 | 4.76 | |
| DiT-XL/2cfg=1.5, Sampler=DDIM, #Steps=202025.05 | 224.3 | 3.49 | 4.93 | 79 | 58 | 4.75 | |
| ICC (SVD)r=256, p=2, Sampler=DDIM, #Steps=142025.05 | 219.5 | 5.1 | 5.59 | 78 | 52 | 2.26 | |
| ICC (CD-SVD)r=192, p=2, Sampler=DDIM, #Steps=142025.05 | 217.6 | 5.05 | 5.46 | 77 | 53 | 2.12 | |
| ICC (CA-SVD)r=128, p=2, Sampler=DDIM, #Steps=162025.05 | 205.9 | 5.56 | 5.4 | 75 | 55 | 2.27 | |
| Naive Cachingp=3, Sampler=DDIM, #Steps=302025.05 | 196.9 | 6.27 | 9.48 | 76 | 51 | 2.38 | |
| Naive Cachingp=2, Sampler=DDIM, #Steps=202025.05 | 193.33 | 6.73 | 8.77 | 74 | 53 | 2.38 | |
| DiT-XL/2cfg=1.5, Sampler=DDIM, #Steps=102025.05 | 160.4 | 12.33 | 11.34 | 67 | 51 | 2.37 |