metadata
license: mit
tags:
- lerobot
- act
- libero
- robot-learning
act-libero-baseline-2026-03-29-005757-1k
備份自 ID Lab lerobot-vla-fbagent 專案，含完整中間 checkpoint。
訓練資訊
| 項目 | 值 |
|---|---|
| Policy Type | act |
| Dataset | HuggingFaceVLA/libero |
| Max Steps | 1,000 |
| Checkpoints | 1 個 |
| Seed | ? |
| Batch Size | ? |
| Learning Rate | ? |
| Num Workers | ? |
| Chunk Size | 100 |
| Git Commit | d440da2 |
| Source Dir | 00-57-57_act |
復現指令
git clone --recurse-submodules git@github.com:KunHsiang/lerobot-vla-fbagent.git
cd lerobot-vla-fbagent
git checkout d440da2
uv sync
HF_HUB_ENABLE_HF_TRANSFER=1 .venv/bin/python scripts/train_with_fb.py \
--policy.type=act \
--policy.repo_id=kunhsiang/act-libero-baseline-2026-03-29-005757-1k \
--dataset.repo_id=HuggingFaceVLA/libero \
--training.seed=? \
--training.batch_size=? \
--training.learning_rate=? \
--training.num_workers=?
Checkpoint 結構
checkpoints/
001000/ # step 1,000
...
001000/ # final (1,000 steps)
每個 step 目錄包含:
pretrained_model/model.safetensors— 模型權重pretrained_model/config.json— 模型設定pretrained_model/train_config.json— 完整訓練設定training_state/optimizer_state.safetensors— Optimizer 狀態（可 resume）training_state/rng_state.safetensors— RNG 狀態（可 resume）
Resume 訓練
.venv/bin/python scripts/train_with_fb.py \
--policy.type=act \
--resume=true \
--policy.pretrained_path=checkpoints/001000/pretrained_model