---

name: image-training description: "Routes Lumina image-model training, finetuning, data preparation, resume, and DreamBooth-style adaptation tasks." disable-model-invocation: true metadata: disco-role: operating license: MIT

Image Training

Use this subskill when the user wants to train, finetune, resume, or prepare data for the Lumina image branches. It covers the Lumina-T2I, Lumina-Next-T2I, and Lumina-Next-T2I-Mini training entry points and the DreamBooth-style SD3 adaptation path.

Include here

• Lumina-T2I training from JourneyDB-style image/caption manifests.
• Lumina-Next-T2I training and the mini training route.
• Resume and initialization behavior for checkpoint-based training.
• DreamBooth-style SD3 adaptation in the mini branch.
• Data-cache and local-diffusers/offline preparation for training runs.

Exclude or route elsewhere

• Pure inference, demo launch, or checkpoint conversion: use image-generation.
• Audio/music demos: use audio-music.
• Visual anagrams: use visual-anagrams.
• ImageNet benchmark training and sampling: use imagenet-training.

Read first

• references/workflows.md for the canonical training routes and launch patterns.
• references/data-formats.md for JourneyDB manifests and other training data layouts.
• references/troubleshooting.md for training-specific backend and resume failures.
• scripts/check_training_data.py before launching if the manifest or config is uncertain.

Fast routing hints

• If the user says torchrun, srun, resume, cache_data_on_disk, JourneyDB, or DreamBooth, use this subskill.
• If the user only needs inference from a checkpoint, stay in image-generation.