ANWM

Aerial Navigation World Model checkpoint released with the paper Aerial World Model for Long-horizon Visual Generation and Navigation in 3D Space.

• Code: https://github.com/EmbodiedCity/ANWM.code
• Dataset: EmbodiedCity/ANWM-Dataset

Files

File	Steps	Notes
0200000.pth.tar	200,000	Paper / released model (config/anwm.yaml)

Quick start

Clone the code, then download this checkpoint into the path expected by infer.py:

git clone https://github.com/EmbodiedCity/ANWM.code.git
cd ANWM.code

huggingface-cli download EmbodiedCity/ANWM 0200000.pth.tar \
  --local-dir logs/anwm_cdit_airvln/checkpoints

# optional check
python scripts/check_environment.py --require-checkpoint

# inference
torchrun --standalone --nproc_per_node=1 infer.py \
  --exp config/anwm.yaml \
  --ckp 0200000 \
  --datasets airvln_16 \
  --eval_type rollout \
  --rollout_fps_values 1,4

This places the weight at:

logs/anwm_cdit_airvln/checkpoints/0200000.pth.tar

Related

• Code: https://github.com/EmbodiedCity/ANWM.code
• Dataset: https://huggingface.co/datasets/EmbodiedCity/ANWM-Dataset
• Model: https://huggingface.co/EmbodiedCity/ANWM
• Paper: https://arxiv.org/abs/2512.21887