jjr1007/diffusion_dino_dagger_ft_2

Running the policy

run_policy.sh bootstraps a local uv venv next to itself, installs lerobot, installs ffmpeg, copies the dinov3-compatible configuration_diffusion.py / modeling_diffusion.py into the venv, and runs the rollout using the model weights in this repository.

Just run the script and it will do the right thing:

./run_policy.sh

Manual setup (if you prefer)

If you'd rather perform the setup by hand, run the following:

pip install uv
export PATH="$HOME/.local/bin:$PATH"
uv venv --python 3.12
source .venv/bin/activate
uv pip install "lerobot[multi_task_dit,feetech,dataset,training,async] @ git+https://github.com/huggingface/lerobot.git"
sudo apt update
sudo apt install -y ffmpeg

Then, for the dinov3 backbone to work, swap these two files in the venv lib files:

<your-venv>/lib/python3.12/site-packages/lerobot/policies/diffusion/configuration_diffusion.py
<your-venv>/lib/python3.12/site-packages/lerobot/policies/diffusion/modeling_diffusion.py

with the configuration_diffusion.py and modeling_diffusion.py in this repository.