datasets/README.md

Datasets Manifest

This directory is intentionally lightweight on the local machine. It records which datasets should be downloaded on the remote cluster under:

/home/l/liyj/shiying/hku_diffusion_dllm/datasets

Use ../scripts/download_assets.py on the remote side to materialize datasets. Large datasets such as Waymo or full OpenWebText should only be pulled after confirming storage and job allocation.