What percentage of the original dataset is composed of ham messages before undersampling?

What percentage of the original dataset is composed of ham messages before undersampling?: a task in SmolDataEnvs-harbor-train-sandbox: 📊 SmolDataEnvs: Harbor (train) (Harbor dataset). Files (in /home/user/input, no subfolders): - spam.csv