Text-to-Image Generation on SD 3-medium (2B) (evaluation)
0.3168CLIP ScoreOBS-diff
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OBS-diffBase Model=SD 3-medium, Sparsity (%)=15%, #Params=1.83 B, TFLOPs=2.432025.10 | 0.3168 | 34.33 | 0.6717 | |
| OBS-diffBase Model=SD 3-medium, Sparsity (%)=10%, #Params=1.91 B, TFLOPs=2.592025.10 | 0.3166 | 35.65 | 0.8118 | |
| OBS-diffBase Model=SD 3-medium, Sparsity (%)=20%, #Params=1.78 B, TFLOPs=2.312025.10 | 0.3163 | 33.15 | 0.4997 | |
| Dense ModelBase Model=SD 3-medium, Sparsity (%)=0%, #Params=2.03 B, TFLOPs=2.842025.10 | 0.3162 | 36.14 | 0.9029 | |
| C2FGBase Model=SD3, Parameters=(5,1)2026.03 | 0.315 | — | — | |
| OBS-diffBase Model=SD 3-medium, Sparsity (%)=25%, #Params=1.72 B, TFLOPs=2.192025.10 | 0.3143 | 32.96 | 0.2782 | |
| CFGBase Model=SD3, Parameters=(5,1)2026.03 | 0.314 | — | — | |
| L1-normBase Model=SD 3-medium, Sparsity (%)=10%, #Params=1.91 B, TFLOPs=2.592025.10 | 0.2035 | 267.32 | -2.2611 | |
| L1-normBase Model=SD 3-medium, Sparsity (%)=15%, #Params=1.83 B, TFLOPs=2.432025.10 | 0.1942 | 326.92 | -2.2768 | |
| L1-normBase Model=SD 3-medium, Sparsity (%)=20%, #Params=1.78 B, TFLOPs=2.312025.10 | 0.1926 | 348.77 | -2.2768 | |
| L1-normBase Model=SD 3-medium, Sparsity (%)=25%, #Params=1.72 B, TFLOPs=2.192025.10 | 0.1906 | 365.24 | -2.2786 |