I'm trying to run YOLO-World in a Docker container on an HPC. I’m facing build and runtime issues with mmcv, mmdet, mmyolo, and yolo_world. Has anyone built a working Dockerfile for this? Requirements: GPU support (CUDA 11.8+, H100 compatible) PyTorch + all YOLO-World dependencies Based on nvidia/cuda image Able to run, fine-tune, train, and evaluate YOLO-World models If you have a Dockerfile that works or any tips, please share. Thanks! Continue reading...
