Tensorflow rust needs love

Hello, I have trying tensorflow rust and my conclusion is that it lacks many functionality.

For instance, some as simple as saving training state to disk and restoring it (checkpointing) is difficult. By contrast, in Tensorflow java it is as simple as Session.save/Session.restore.

Rust should be the favorite language for production workloads over C++. Rust is simpler and less error prone. Of course, Java can be an option for those who prefer easier programming, but with larger memory consumption and therefore higher cloud fees.

Please, give more love to Tensorflow rust.

PD: I will contribute to the extent I can.

Best regards.