Official implementation of the paper "InSpire: Vision-Language-Action Models with Intrinsic Spatial Reasoning".
Note: We are doing our best to improve this work. If you have any questions or suggestions, please feel free to create an issue in this repo or contact us at shihan.wu.koorye@outlook.com.
[Project] [ArXiv] [PDF] [Inspire]
- 🔥May 23, 2025: Our paper has been updated for better clarity and readability. The optimized version is now available on arXiv.
- 🔥May 21, 2025: The code is released and the paper is now available on arXiv.
| Model | Dataset | Checkpoint |
|---|---|---|
| FAST | Libero90 | Download |
| InspireVLA-FAST | Libero90 | Download |
- Clone the repository.
git clone https://github.com/Koorye/Inspire-FAST.git- Install dependencies.
cd LIBERO
pip install -r requirements.txt
pip install -e .
cd ..
cd lerobot-v2
pip install -e .
cd ..
pip install -e .- Download pretrained checkpoints.
huggingface-cli download InspireVLA/fast-libero-90
huggingface-cli download InspireVLA/fast-inspire-libero-90- Run evaluation.
task_suite_names=(
libero_90
libero_goal
libero_spatial
libero_object
libero_10
)
for task_suite_name in "${task_suite_names[@]}"; do
XLA_PYTHON_CLIENT_PREALLOCATE=false python scripts/parallel_libero_evaluator.py \
--config-name your_config_name \
--checkpoint-dir your/checkpoint/path \
--task-suite-name $task_suite_name
done- Prepare your dataset.
See Dataset Preparation.
-
Update the config file
src/training/config.py, changing eachrepo_idin the_CONFIGSlist. -
Train Baseline.
XLA_PYTHON_CLIENT_MEM_FRACTION=0.9 python scripts/train.py \
pi0_fast_libero \
--exp-name=pi0_fast_libero \
--resumeXLA_PYTHON_CLIENT_MEM_FRACTION=0.9 python scripts/train.py \
pi0_fast_real \
--exp-name=pi0_fast_real \
--resume- Train Inspire on LIBERO.
XLA_PYTHON_CLIENT_MEM_FRACTION=0.9 python scripts/train_fast_vqa_libero.py \
pi0_fast_libero \
--exp-name=pi0_fast_libero \
--resume- Train Inspire on real-world.
XLA_PYTHON_CLIENT_MEM_FRACTION=0.9 python scripts/train_fast_vqa_real.py \
pi0_fast_real \
--exp-name=pi0_fast_real \
--resumeOur work is built upon the following open-source projects: LIBERO, MiniVLA, Pi-0. We thank the authors for releasing their code. If you use our model and code, please consider citing these works as well.