Training abnormally slow

[7W7W7W]

Ultralytics YOLOv8.2.29 🚀 Python-3.11.9 torch-2.4.0+cu121 CUDA:0 (NVIDIA GeForce RTX 3060 Ti, 8192MiB)
Mamba-YOLO-T， It took over 30 hours, but Mamba-YOLO-L couldn't run at all. Is this normal？

[EthanW-coder]

Unfortunately, Mamba's training time on visual tasks is not comparable to CNN's framework, and we are working to address these issues in the future. In fact, on the COCO dataset, we used 8×H800 training, about which we elaborate in the paper, and more details will be published after the mid-draft of the paper.

[YangJae96]

Does mamba based yolo network has advantange in terms of inference speed, but not training time?