Move the model into robotwin/, as in THUSI-Lab/PrismWAM

README.md

@@ -14,15 +14,16 @@ World Action Models**. It is released for evaluation only.
 
 [Code](https://github.com/THUSI-Lab/PrismWAM) · [Project page](https://prism-wam.github.io/) · [arXiv](https://arxiv.org/abs/2610.26564) · [PrismWAM checkpoints](https://huggingface.co/THUSI-Lab/PrismWAM)
 
-| Benchmark | Actions | Result |
-| --- | --- | ---: |
-| RoboTwin 2.0, Clean → Randomized | 14-D absolute joints | 83.67 |
+| Folder | Benchmark | Actions | Result |
+| --- | --- | --- | ---: |
+| `robotwin/` | RoboTwin 2.0, Clean → Randomized | 14-D absolute joints | 83.67 |
 
 Average success rate in % over the 50 tasks in clean and randomized scenes, 100 episodes per task;
 trained on clean scenes only.
 
-This repository is one model folder: `config.yaml`, `policy.pt`, the RGB and depth VAEs, and the
-action statistics. On first use the code also fetches the T5 text encoder of
+The model folder is laid out like those of [THUSI-Lab/PrismWAM](https://huggingface.co/THUSI-Lab/PrismWAM):
+`config.yaml`, `policy.pt`, the RGB and depth VAEs, and the action statistics. On first use the
+code also fetches the T5 text encoder of
 [Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B) and the CLIP image encoder of
 [Wan2.1-I2V-14B-480P](https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-480P).
 
@@ -31,13 +32,13 @@ action statistics. On first use the code also fetches the T5 text encoder of
 Install the [code](https://github.com/THUSI-Lab/PrismWAM), then from the repository root:
 
 ```bash
-hf download THUSI-Lab/PrismWAM-Star --local-dir weights/prismwam_star
+hf download THUSI-Lab/PrismWAM-Star --include "robotwin/*" --local-dir weights/prismwam_star
 ```
 
 ```python
 from prismwam import load_pretrained
 
-policy = load_pretrained("prismwam_star")
+policy = load_pretrained("prismwam_star/robotwin")   # uses weights/prismwam_star/robotwin
 observation["prompt"] = policy.format_prompt("lift the bottle with the right arm")
 out = policy.predict(observation, seed=0)        # out["action"]: 32 x 14
 ```