Huge size of weights consuming whole 5gb disk in kaggle

I face a strange issue while i save the weights for segmentation problem with encoder resnet50 …
Should this happen ? are the size of weights so big .
I just used 128 size that too with mixed precision.