Qwen-2511-edit LoRA
Style LoRA for image editing, trained on FLUX.2-klein-base-9B.
Details
- Base model:
Qwen/Qwen-Image-Edit-2511
- Architecture: qwen_image_edit_plus:2511
- Trigger word:
t01nspc_style
- Network: LoRA (linear=32, conv=16)
- Training steps: 8000 (completed at step 8000)
- Learning rate: 0.0001
- Optimizer: adamw8bit
- Dtype: bf16
- Resolution: [512, 768, 1024]
- Content/Style: content
Training Parameters
| Parameter | Value |
|---|
| Batch size | 1 |
| Gradient accumulation | 1 |
| Noise scheduler | flowmatch |
| Timestep type | weighted |
| Loss type | mse |
| Caption dropout | 0.05 |
| Weight decay | 0.0001 |
| Diff output preservation | False |
Checkpoints
Last 7 checkpoints (steps 4500–7500) + final model in checkpoints/.
