Questions about the training process

Thank you for your great work! I have some questions about the two-stage training process since the training code has not been released yet. The cross-task triplet curation framework yields layout-preserved triplets and layout-shifted triplets. In content-style disentanglement training, do you use layout-preserved triplets without text prompt to train the model for layout-preserved generation and use layout-shifted triplets with text prompts to train the model for layout-shifted generation?

The issue does not name any files, tests, or implementation entry points, and notes that the training code has not been released. Start by checking the repository documentation and any released training materials for the two-stage process. The issue is done when the triplet and prompt usage is documented or an authoritative answer is provided.

Written by the indexing model from the issue text.