Recommended HPC templates and best practices for faster DeePTB training and large-scale prediction

### Details

Hi Developers,

I have a few questions
1. Do you have a recommended job-script template for running DeePTB efficiently on HPC systems?
2. For large systems, is there a recommended way to reduce memory usage during prediction, especially when building the Hamiltonian or computing band structure/transport properties?
3. Are there recommended settings for batch size, dataloader workers, or threading to improve performance on clusters?

Thank you