你们有对比多lumina-next-2B模型跟SD3-2B模型的速度吗？我这边跑了下，慢了一倍，会是什么原因呢？

Alpha-VLLM / Lumina-T2X

Lumina-T2X is a unified framework for Text to Any Modality Generation

MIT License

1.82k stars 74 forks source link

Open heart-du opened 1 week ago

gaopengpjlab commented 1 week ago

The depth and dimension of SD3-2B are smaller than Lumina-Next-2B. That's why SD3-2B is faster.