To improve scalability and usability, we should formalize the offloading API and refactor it into a modular class.
The current implementation in dattri/algorithm/block_projected_if/offload serves as an excellent starting point for our codebase. Efficient offloading is also a critical component for extending dattri’s applicability to large-scale settings, such as massive datasets.
To improve scalability and usability, we should formalize the offloading API and refactor it into a modular class.
The current implementation in dattri/algorithm/block_projected_if/offload serves as an excellent starting point for our codebase. Efficient offloading is also a critical component for extending dattri’s applicability to large-scale settings, such as massive datasets.