Spaces:

---

openenv-community

/

origami_env

like 0

Running

Rewrote training notebook to properly use the OpenEnv architecture:

• Connects to HF Space server URL for all simulation and reward computation
• No local origami_server imports — fully decoupled client/server
• GET /tasks to fetch task definitions from server
• WebSocket /ws for reset + step (simulation + shape matching)
• valid_fold reward: local JSON validation (fast, no server needed)
• shape_match reward: calls server via WebSocket per completion