LeHome Challenge 2026 — Submission

File Layout

submission_v4_global/
├── README.md                          # this file
├── residual_v4_global.py              # the policy module (drop into eval_policy/)
└── submission_models/
    ├── vla_backbone/                  # SmolVLA four_types 30K (LeRobot pretrained_model, ~865M)
    ├── residual_averaged.pt           # 40-garment averaged residual MLP (~286K)
    ├── dataset_meta/                  # LeRobot dataset metadata (stats.json etc.)
    └── hf_cache/                      # bundled SmolVLM2 weights for offline VLM load (~1.9G)
        └── hub/models--HuggingFaceTB--SmolVLM2-500M-Video-Instruct/
            ├── snapshots/<commit>/    # tokenizer + processor + model.safetensors
            └── refs/main              # commit hash file

How To Run (evaluator side)

1. Drop residual_v4_global.py into /opt/lehome-challenge/scripts/eval_policy/.
2. Add to scripts/eval_policy/__init__.py:
   from .residual_v4_global import ResidualV4GlobalPolicy
3. Set environment variables:
   bash

   1export LEHOME_VLA_POLICY_PATH=<path to submission_models/vla_backbone> 2export LEHOME_VLA_DATASET_ROOT=<path to submission_models/dataset_meta or any LeRobot dataset> 3export LEHOME_RESIDUAL_CHECKPOINT=<path to submission_models/residual_averaged.pt> 4export LEHOME_RESIDUAL_SCALE=0.03
   The wrapper sets HF_HOME automatically to the bundled hf_cache/ when it sees LEHOME_VLA_POLICY_PATH, so no network access is required even on a fully offline evaluator.
   Belt-and-suspenders — if the evaluator's launcher imports huggingface_hub before our wrapper module loads (rare but possible), the redirect may be too late. To be safe, set HF env vars before invoking python -m scripts.eval:
   bash

   1export HF_HOME="$LEHOME_VLA_POLICY_PATH/../hf_cache" 2export HF_HUB_CACHE="$HF_HOME/hub" 3export HF_HUB_OFFLINE=1 4export TRANSFORMERS_OFFLINE=1
4. Invoke evaluator:
   bash

   1python -m scripts.eval \ 2 --policy_type residual_v4_global \ 3 --policy_path "$LEHOME_VLA_POLICY_PATH" \ 4 --dataset_root "$LEHOME_VLA_DATASET_ROOT" \ 5 --garment_type <top_long|top_short|pant_long|pant_short> \ 6 --num_episodes 5 --max_steps 600 \ 7 --enable_cameras --device cpu --headless

Method Summary

• Backbone: SmolVLA, jointly-trained on 4 garment types for 30K steps. Frozen during residual RL.
• Residual: small state-only MLP (state_dim=12 → 256 → 256 → action_dim=12, 3 Linear+ReLU layers).
• Final action: clip(base_action + 0.03 * residual_mlp(state)).
• Training signal: sparse reward (1 if folding success at episode end, else 0).
• Training data: 40 Seen garments (10 per type × 4 types), 30 episodes per garment, on-policy PPO updates.
• Aggregation: weights averaged across 40 per-garment training runs to get a single global residual.
• Inference: deterministic — no exploration noise, no online updates.

Key Hyperparameters

Evaluation Results

Reference baselines

Notes on the run

• One garment (Top_Long_Seen_9) initially failed in the main parallel sweep with an Isaac Sim TiledCamera._annotators AttributeError (unrelated to the policy). It was retried in a fresh single-garment process and produced 4/5 success — that retry is included in the 150/240 figure above.
• Episode-level data per garment is in FINAL_SUMMARY.json and per-garment stdout logs are in eval_raw/.

Artefact Hashes

Reproducibility

Notes

• The residual_averaged.pt follows the format:
  python

  1torch.save({ 2 "state_dim": 12, 3 "action_dim": 12, 4 "hidden_dims": (256, 256), 5 "model_state_dict": <state-only MLP weights>, 6}, path)
• This submission is a single model (one residual checkpoint) handling all four garment types — not a per-type specialist ensemble.
• Inference path: LeRobotPolicy.select_action(observation) → adds 0.03 * residual_mlp(observation['observation.state']).

Contact