Your training code may be causing high internet costs in Google Colab due to:

1.Frequent Checkpoint Saving:

Saving the checkpoint after every epoch can increase disk I/O operations and might sync with your Google Drive (if mounted), consuming bandwidth. Consider saving checkpoints less frequently, such as every 5 or 10 epochs.

2.Visualization:

Frequent visualizations, especially when training models, can use significant resources. Reduce the frequency of visualizations or save plots locally instead of displaying them.