Update README.md

README.md

@@ -36,21 +36,15 @@ checkpoints/
     vit_l_issu.pth
 ```
 
-Each `flows/*/checkpoint_best.ckpt` is a plain `{"state_dict": ...}` dict with
-the trained flow-mixture parameters only. The frozen DINOv3 backbone weights are
-**not** included — they are gated `timm` weights and must be obtained directly
-from Hugging Face (see the code repo's environment setup). The evaluation code
-rebuilds the backbone from `timm` and loads these checkpoints with
-`strict=False`, so the absent backbone keys are expected.
 
 | Checkpoint | Backbone | Feature level | Training data |
 |---|---|---|---|
 | `convnext_l_res4_cityscapes` | DINOv3 ConvNeXt-L | res4 | Cityscapes (train) |
 | `convnext_l_res5_cityscapes` | DINOv3 ConvNeXt-L | res5 | Cityscapes (train) |
-| `vit_l_res5_cityscapes`      | DINOv3 ViT-L     | last block (no closing LN) | Cityscapes (train) |
+| `vit_l_res5_cityscapes`      | DINOv3 ViT-L     | last block | Cityscapes (train) |
 | `convnext_l_res4_issu`       | DINOv3 ConvNeXt-L | res4 | ISSU-static (train) |
 | `convnext_l_res5_issu`       | DINOv3 ConvNeXt-L | res5 | ISSU-static (train) |
-| `vit_l_res5_issu`            | DINOv3 ViT-L     | last block (no closing LN) | ISSU-static (train) |
+| `vit_l_res5_issu`            | DINOv3 ViT-L     | last block | ISSU-static (train) |
 
 ## Usage
 
@@ -68,16 +62,8 @@ snapshot_download(
 )
 ```
 
-Then run evaluation as described in the code repo, e.g.:
 
-```bash
-python eval_smiyc_benchmark.py \
-  --config configs/eval/convnext_l_res5_cityscapes.yaml \
-  --checkpoint checkpoints/flows/convnext_l_res5_cityscapes/checkpoint_best.ckpt
-```
 
 ## License
 
-Released for the purpose of peer review and research reproducibility only. The
-DINOv3 backbones these flows are trained on top of are governed by their own
-licenses on Hugging Face and are not redistributed here.
+Released for the purpose of peer review and research reproducibility only.