Request for Training Code and Evaluation Pipeline

[jee019]

Hi,

Thank you for open-sourcing this impressive work. I’ve successfully run the provided demos.

However, I couldn't find the training scripts, datasets, or evaluation scripts in the repository.

Could you please clarify:

1. Is the training code for the models publicly available?
2. Where can we access the full training/evaluation pipeline, especially for supervised fine-tuning or the two-stage pretraining strategy mentioned in the technical report?
3. If these are not yet released, are there any plans to open-source them in the near future?

Looking forward to your response.
Thanks again!