Upload README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# SpatialMirage Stage1 to Stage2 Checkpoint
|
| 2 |
+
|
| 3 |
+
Private checkpoint package for the Stage2 SpatialEdit camera-control run initialized from Stage1.
|
| 4 |
+
|
| 5 |
+
The main trainable safetensors file is split to stay under the HuggingFace single-file limit.
|
| 6 |
+
|
| 7 |
+
Reassemble it with:
|
| 8 |
+
|
| 9 |
+
```bash
|
| 10 |
+
bash model/reassemble_trainable.sh
|
| 11 |
+
```
|
| 12 |
+
|
| 13 |
+
Contents:
|
| 14 |
+
|
| 15 |
+
- split `model/trainable.safetensors.part-*`
|
| 16 |
+
- `model/trainable_params.json`
|
| 17 |
+
- full resume state under `resume_state/`
|
| 18 |
+
- Stage2 training config and run script under `configs/` and `runs/`
|
| 19 |
+
|
| 20 |
+
Source checkpoint:
|
| 21 |
+
|
| 22 |
+
`/mnt/dhwfile/raise/user/caoyunqi/checkpoints/spatialedit_stage2_camera_243k_old_notext_from_stage1vggt_60000_20260624/from_stage1vggt_1node8gpu/0060000`
|