# ppo_agent

zero2robot live-demo policy. Loaded by the site at `/models/ppo_agent.onnx` and provisioned at build time by `site/scripts/fetch_models.py` (sha256-verified).

| field | value |
|---|---|
| chapter | `ch2.1-ppo` |
| contract | `v1` |
| seed | `0` |
| tier | `cpu-laptop` |
| exporter | `curriculum/phase2_reinforcement/ch2.1_ppo/export_ppo_onnx.py` |
| sha256 | `abbf577b80eef8a7f6d2df44d526720ec20a447486de45f574c2e52f2a399c69` |

Reproduce: run the exporter above at the recorded seed on the recorded tier; the export is CPU-deterministic, so the bytes (and this sha256) reproduce exactly.