Releasing the training code

[richiejp]

We are considering releasing the training code or at least code suitable for fine-tuning. However we have to consider what information we should withhold to help our business offerings.

The training code in its present state; we wouldn't release without thorough review to remove sensitive information. I can not give a date on when we may release some training code as LocalVQE is still a WIP and there are new directions to try.

What would motivate us to release the training code is if we see strong evidence that the community will participate in improving the training and create more value for our products. As is always the case with Open Source it's worth sharing when giving up the advantage from having some secrets is outweighed by how much more value you will get from the community.

Originally requested in #2

[thinkingwanglin]

Thank you for carefully considering the release of training or fine-tuning code and for clearly explaining your current thinking!
We fully understand and respect the balance your team is striking between protecting business interests and fostering open collaboration. Given that LocalVQE is still under active development, it’s entirely reasonable to handle sensitive information with caution.
It’s very encouraging to see your openness to community co-creation—that alone means a lot to us.
Looking forward to staying in touch, and wishing you continued success with the LocalVQE project!

[mpisat]

Let me use this open issue to say congrats for releasing localvqe models. It is working very well and I'm very happy with the results.