...
SeFi-Image is intended for research and creative text-to-image generation, including prompt following, bilingual text rendering, style exploration, and model development. The Base checkpoints are suitable starting points for fine-tuning and analysis. Turbo checkpoints are intended for fast generation. The RL checkpoint is intended for stronger alignment-oriented generation.
| Family | Model | Checkpoint | TR | TE | VAE | Steps | Guidance | time (285H) | test v1 | test v2 | test v3 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Base | SeFi-Image-1B-Base | SeFi-Image/SeFi-Image-1B-Base | 1.1 | 2.1 | 0.08 | 50 | 4.0 | 2m 18.09s | SeFi Image 1B Base | SeFi Image 1B Base test v2 | SeFi Image 1B Base test v3 |
|
| Base | SeFi-Image-2B-Base | SeFi-Image/SeFi-Image-2B-Base | 2.1 | 2.1 | 0.08 | 50 | 4.0 | 3m 50.27s | SeFi Image 2B Base |
| ||
| Base | SeFi-Image-5B-Base | SeFi-Image/SeFi-Image-5B-Base | 4.9 | 4.4 | 0.08 | 50 | 4.0 | 8m 5.20s | SeFi Image 5B Base |
| ||
| RL | SeFi-Image-5B-RL | SeFi-Image/SeFi-Image-5B-RL | 4.9 | 4.4 | 0.08 | 50 | 4.0 | 8m 2.03s | SeFi Image 5B RL | SeFi Image 5B RL test v2 | SeFi Image 5B RL test v3 |
|
| Turbo | SeFi-Image-1B-turbo | SeFi-Image/SeFi-Image-1B-turbo | 1.1 | 2.1 | 0.08 | 4 | 1.0 | SeFi Image 1B Turbo | SeFi Image 1B Turbo test v2 | SeFi Image 1B Turbo test v3 |
| |
| Turbo | SeFi-Image-2B-turbo | SeFi-Image/SeFi-Image-2B-turbo | 2.1 | 2.1 | 0.08 | 4 | 1.0 | 10.86s | SeFi Image 2B Turbo |
| ||
| Turbo | SeFi-Image-5B-turbo | SeFi-Image/SeFi-Image-5B-turbo | 4.9 | 4.4 | 0.08 | 4 | 1.0 | 23.72s | ||||
| SeFi Image 5B Turbo | SeFi Image 5B Turbo test v2 | SeFi Image 5B Turbo test v3 |
|
Total parameters count (Text Encoder + Transformer + VAE)
...