Google Colab use big Dataset

I want to train yolov8 using colab, but there is a problem I am facing now. I am using bdd100k which I have changed a little bit as a dataset, there is 63k data for train, when I start yolo training, I get a mount error in the part where it counts the labels in yolo and this sequence never ends and then it throws it. obviously I pay money to use the gpu and I cannot use it the way I want. the process of counting this data takes an extremely long time, do you have any suggestions for this?