Class-conditional image generation on ImageNet-1K 256x256 (test)
1.5FIDRAR-XL
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RAR-XLType=AR, Param.=955 M, Steps=2562025.03 | 1.5 | 8 | 10.55 | 306.9 | 80 | 62 | — | |
| RAR-LType=AR, Param.=461 M, Steps=2562025.03 | 1.7 | 12.08 | 6.37 | 299.5 | 81 | 60 | — | |
| MAR-LType=Mask, Param.=479 M, Steps=1002025.03 | 1.78 | 1.27 | 2.32 | 296 | 81 | 60 | — | |
| ARPG-XXLType=AR (Parallel), Param.=1.3 B, Steps=322025.03 | 1.83 | 55.28 | 7.22 | 336.1 | 80 | 60 | — | |
| ARPG-XXLType=AR (Parallel), Param.=1.3 B, Steps=642025.03 | 1.86 | 31.65 | 7.18 | 339.7 | 81 | 59 | — | |
| ARPG-XLType=AR (Parallel), Param.=719 M, Steps=322025.03 | 1.93 | 80.56 | 4.65 | 349.2 | 80 | 61 | — | |
| RAR-BType=AR, Param.=261 M, Steps=2562025.03 | 1.95 | 14.12 | 4.65 | 290.5 | 82 | 58 | — | |
| ARPG-XLType=AR (Parallel), Param.=719 M, Steps=642025.03 | 1.99 | 45.09 | 4.57 | 340.6 | 80 | 61 | — | |
| VAR-d24Type=VAR, Param.=1.0 B, Steps=102025.03 | 2.09 | 49.94 | 22.29 | 312.9 | 82 | 59 | — | |
| RandAR-XXLType=AR (Parallel), Param.=1.4 B, Steps=882025.03 | 2.15 | 10.46 | 21.77 | 322 | 79 | 62 | — | |
| RandAR-XLType=AR (Parallel), Param.=775 M, Steps=882025.03 | 2.25 | 15.51 | 13.52 | 317.8 | 80 | 60 | — | |
| DiT-XL/2Type=Diffusion, Param.=675 M, Steps=2502025.03 | 2.27 | 0.91 | 2.14 | 278.2 | 83 | 57 | — | |
| ARPG-LType=AR (Parallel), Param.=320 M, Steps=322025.03 | 2.3 | 130.14 | 2.78 | 297.7 | 82 | 56 | — | |
| MAR-BType=Mask, Param.=208 M, Steps=1002025.03 | 2.31 | 1.71 | 1.47 | 281.7 | 82 | 57 | — | |
| PAR-XXLType=AR (Parallel), Param.=1.4 B, Steps=1472025.03 | 2.35 | 5.23 | 26.25 | 263.2 | 80 | 62 | — | |
| ARPG-LType=AR (Parallel), Param.=320 M, Steps=642025.03 | 2.37 | 67.47 | 2.64 | 293.7 | 82 | 55 | — | |
| VDM++params.=2B, steps=512, Classifier-free guidance=false2023.12 | 2.4 | — | — | 225.3 | — | — | — | |
| RandAR-LType=AR (Parallel), Param.=343 M, Steps=882025.03 | 2.55 | 25.3 | 7.32 | 288.8 | 81 | 58 | — | |
| W.A.L.T-LCost (Iter x BS)=437kx1024, params.=460M, steps=50, Classifier-free guidance=false2023.12 | 2.56 | — | — | 215.1 | — | — | — | |
| VAR-d20Type=VAR, Param.=600 M, Steps=102025.03 | 2.57 | 75.38 | 15.97 | 302.6 | 83 | 56 | — | |
| NAR-XXLType=AR (Parallel), Param.=1.5 B, Steps=312025.03 | 2.58 | 15.23 | 26.25 | 293.5 | 82 | 57 | — | |
| PAR-XLType=AR (Parallel), Param.=775 M, Steps=1472025.03 | 2.61 | 7.91 | 17.13 | 259.2 | 80 | 62 | — | |
| LlamaGen-XLType=AR, Param.=775 M, Steps=5762025.03 | 2.62 | 2.46 | 17.11 | 244.1 | 80 | 57 | — | |
| LlamaGen-XXLType=AR, Param.=1.4 B, Steps=5762025.03 | 2.62 | 1.58 | 26.22 | 244.1 | 80 | 57 | — | |
| NAR-XLType=AR (Parallel), Param.=816 M, Steps=312025.03 | 2.7 | 23.97 | 17.13 | 277.5 | 81 | 58 | — | |
| simple diffusionCost (Iter x BS)=500Kx2048, params.=2B, steps=512, Classifier-free guidance=false2023.12 | 2.77 | — | — | 211.8 | — | — | — | |
| NAR-LType=AR (Parallel), Param.=372 M, Steps=312025.03 | 3.06 | 42.71 | 10.25 | 263.9 | 81 | 53 | — | |
| LlamaGen-LType=AR, Param.=343 M, Steps=5762025.03 | 3.07 | 4.33 | 10.23 | 256.1 | 83 | 52 | — | |
| VAR-d16Type=VAR, Param.=310 M, Steps=102025.03 | 3.3 | 123.21 | 10.85 | 274.4 | 84 | 51 | — | |
| RINCost (Iter x BS)=600kx1024, params.=410M, steps=1000, Classifier-free guidance=false2023.12 | 3.42 | — | — | 182 | — | — | — | |
| PAR-LType=AR (Parallel), Param.=343 M, Steps=1472025.03 | 3.76 | 14.77 | 10.25 | 218.9 | 81 | 60 | — | |
| DiT-L/2Type=Diffusion, Param.=458 M, Steps=2502025.03 | 5.02 | 1.32 | 1.62 | 167.2 | 75 | 57 | — | |
| MaskDiTCost (Iter x BS)=1200kx1024, params.=736M, steps=40, Classifier-free guidance=false2023.12 | 5.69 | — | — | 178 | — | — | — | |
| MaskGITType=Mask, Param.=227 M, Steps=82025.03 | 6.18 | 46.18 | 1.71 | 182.1 | 80 | 51 | — | |
| MDTCost (Iter x BS)=6500kx256, params.=676M, steps=250, Classifier-free guidance=false2023.12 | 6.23 | — | — | 143 | — | — | — | |
| BigGAN-deepparams.=160M, steps=1, Classifier-free guidance=false2023.12 | 6.95 | — | — | 171.4 | — | — | — | |
| ADMparams.=608M, steps=2000, Classifier-free guidance=false2023.12 | 7.49 | — | — | 127.5 | — | — | — | |
| DiT-XL/2Cost (Iter x BS)=7000kx256, params.=675M, steps=250, Classifier-free guidance=false2023.12 | 9.62 | — | — | 121.5 | — | — | — | |
| TFMQ-DMBits (W/A)=4/82023.11 | 10.29 | — | — | 221.82 | — | — | 7.35 | |
| PTQDBits (W/A)=4/82023.11 | 10.4 | — | — | 214.73 | — | — | 12.63 | |
| TFMQ-DMBits (W/A)=4/322023.11 | 10.5 | — | — | 223.81 | — | — | 7.98 | |
| LDM-4Cost (Iter x BS)=178kx1200, params.=400M, steps=250, Classifier-free guidance=false2023.12 | 10.56 | — | — | 103.5 | — | — | — | |
| Q-DiffusionBits (W/A)=4/82023.11 | 10.68 | — | — | 212.51 | — | — | 14.85 | |
| TFMQ-DMBits (W/A)=8/82023.11 | 10.79 | — | — | 198.86 | — | — | 7.65 | |
| Full Prec.Bits (W/A)=32/322023.11 | 10.91 | — | — | 235.64 | — | — | 7.67 | |
| PTQDBits (W/A)=4/322023.11 | 11.65 | — | — | 201.78 | — | — | 9.06 | |
| Q-DiffusionBits (W/A)=4/322023.11 | 11.87 | — | — | 213.56 | — | — | 8.76 | |
| PTQDBits (W/A)=8/82023.11 | 11.94 | — | — | 153.92 | — | — | 8.03 | |
| PTQ4DMBits (W/A)=8/82023.11 | 12.59 | — | — | 161.75 | — | — | — | |
| Q-DiffusionBits (W/A)=8/82023.11 | 12.8 | — | — | 187.65 | — | — | 9.87 |