NotImplementedError when processing video frames with demo_inference.py

I encountered a NotImplementedError while running the demo_inference.py script with a video input. The script loads the pose and YOLO models successfully, but fails during the image processing step with the following message: ===============================> Finish Model Running. NotImplementedError() An error as above occurs when processing the images, please check it

Steps: PYTHONPATH=$PYTHONPATH:$(pwd) python scripts/demo_inference.py
--cfg configs/coco/resnet/256x192_res50_lr1e-3_1x.yaml
--checkpoint pretrained_models/fast_res50_256x192.pth
--video input/cycle.mp4
--outdir output
--save_video
--vis_fast

Environment Info • Python version: 3.12 • Torchvision warning shows usage of deprecated pretrained parameter • FFMPEG fallback due to unsupported codec h264 • OS: macOS (Homebrew + Miniconda) • Backend: agg version v2.2 • CUDA not available, fallback to CPU

Additional Notes • The video loads, models initialize correctly, and progress reaches 100% (444/444 frames). • Final output halts with the NotImplementedError() without a traceback. • Might be related to vis_fast or save_video features post-processing.

Would appreciate any insights or help fixing this issue. Let me know if more logs or details are needed.