Text-to-Image Generation on Z-Image 1024x1024 (test)
0.982ImageRewardSpectral Progressive Diffusion (LoRA)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Spectral Progressive Diffusion (LoRA)Fine-tuning type=LoRA, Number of resolution stages (S)=2, Speedup (s)=1.65×, TFLOPs=3132.032026.05 | 0.982 | 0.699 | 5.72 | 0.725 | 0.731 | |
| Z-ImageSampling steps=50, Speedup (s)=1.00×, TFLOPs=4941.232026.05 | 0.965 | 0.7 | 5.41 | 0.731 | 0.745 | |
| Spectral Progressive Diffusion (LoRA)Fine-tuning type=LoRA, Number of resolution stages (S)=3, Speedup (s)=1.74×, TFLOPs=2871.092026.05 | 0.954 | 0.697 | 5.75 | 0.717 | 0.728 | |
| Spectral Progressive Diffusion (TF)Fine-tuning type=TF, Number of resolution stages (S)=2, Speedup (s)=1.65×, TFLOPs=3132.032026.05 | 0.904 | 0.688 | 5.87 | 0.658 | 0.73 | |
| Spectral Progressive Diffusion (TF)Fine-tuning type=TF, Number of resolution stages (S)=3, Speedup (s)=1.74×, TFLOPs=2871.092026.05 | 0.875 | 0.69 | 5.59 | 0.65 | 0.682 |