---
license: apache-2.0
tags:
- robotics
- lingbot-va
- unitree-g1
- world-model
---

# g1_fdmv2_openLidPotato_500step — LingBot-VA G1 post-trained transformer

Fine-tuned `transformer` for LingBot-VA on Unitree G1 (Dex1), task
`bobchenyx/g1_dex1_open_lid_add_potato_033`:
*"Open the pot's lid and put the potato inside the pot."*

- Base: `robbyant/lingbot-va-base`
- Post-training: 50 demos, single task, lr 1e-5, **FDM v2 recipe** —
  mutually-exclusive per-microstep regime (rank-synced coin `fdm_prob=0.5`:
  FDM video-only L_fdm Eq.13 `lambda_fdm=1.0` OR standard IDM L_dyn+L_inv;
  one forward, one backward). Optimizer **step 500** of a 2000-step run
  (training still in progress).
- This repo contains **only `transformer/`** — `vae/`, `text_encoder/`,
  `tokenizer/` are unchanged from `robbyant/lingbot-va-base`.

## Assemble an eval-ready checkpoint

```bash
hf download robbyant/lingbot-va-base                   --local-dir lingbot-va-base
hf download armanakbari4/g1_fdmv2_openLidPotato_500step --local-dir g1_olp_500_dl

mkdir -p g1_olp_500
ln -sf $(realpath g1_olp_500_dl/transformer)  g1_olp_500/transformer
ln -sf $(realpath lingbot-va-base/vae)        g1_olp_500/vae
ln -sf $(realpath lingbot-va-base/text_encoder) g1_olp_500/text_encoder
ln -sf $(realpath lingbot-va-base/tokenizer)  g1_olp_500/tokenizer
```

Serve with `CONFIG_NAME=g1_openlidpotato MODEL_PATH=g1_olp_500`.
`transformer/config.json` has `attn_mode: torch` (inference-ready).