refresh meta & README

README.md

@@ -1,3 +1,46 @@
-# LeRobot v2 Dataset
+---
+tags:
+  - robotics
+  - imitation-learning
+  - lerobot
+  - teleoperation
+license: apache-2.0
+pretty_name: LeRobot v2 Dataset
+---
 
-Uploaded with `upload_min.py`.
+# LeRobot v2 Dataset Upload
+
+This dataset follows the **LeRobot v2** layout with `data/`, `videos/`, and `meta/` folders.
+
+## Summary
+- **Version**: `2.0`
+- **FPS**: `12.0`
+- **Action shape**: `[-1, 14]`
+- **Cameras**: `wrist_cam_left, wrist_cam_right, overhead_cam, worms_eye_cam`
+- **Chunks**: `chunk-000`
+- **Episodes**: `2`
+- **Total frames**: `335`
+
+## Structure
+
+```
+
+dataset_root/
+├── data/
+│   └── chunk-000/
+│       ├── episode_000000.parquet
+│       └── ...
+├── videos/
+│   └── chunk-000/
+│       └── observation.images.wrist_cam_left/episode_000000.mp4
+│       └── observation.images.wrist_cam_right/episode_000000.mp4
+│       └── observation.images.overhead_cam/episode_000000.mp4
+│       └── observation.images.worms_eye_cam/episode_000000.mp4
+└── meta/
+    ├── info.json
+    └── episodes.jsonl
+```
+
+## Loading
+
+You can read Parquet steps and corresponding MP4 frames in your own loader, or integrate with your LeRobot-based pipeline.

meta/episodes.jsonl

@@ -1,3 +1,2 @@
-{"episode": "episode_000000", "chunk": "chunk-000", "length": 303, "split": "train"}
-{"episode": "episode_000001", "chunk": "chunk-000", "length": 142, "split": "train"}
-{"episode": "episode_000002", "chunk": "chunk-000", "length": 1009, "split": "train"}
+{"episode": "episode_000000", "chunk": "chunk-000", "length": 271, "split": "train"}
+{"episode": "episode_000001", "chunk": "chunk-000", "length": 64, "split": "train"}

meta/info.json

@@ -16,7 +16,7 @@
       "dtype": "float32",
       "shape": [
         -1,
-        "state_dim"
+        14
       ]
     },
     "timestamp": {