Missing It Takes Two metadata + Robotics training code?

[lmsdss]

Hi, thanks for open-sourcing this great work and releasing the dataset!

I downloaded the data from HuggingFace (Haoyuwu/MultiWorldData) and ran preprocess/untar_chunks.sh. A couple of questions:

1. It Takes Two metadata (missing)
   Video/action files extracted fine, but I couldn't find the metadata CSVs needed for training:

Expected: data/ittakestwo_release/480P_toy/metadata/train_metadata.csv and val_metadata.csv (per train_ua_480P_toy.yaml)
Also for eval: data/ittakestwo_release/480P_eval/metadata/metadata.csv
Robotics metadata is included (robots_release/metadata/), but It Takes Two only has per-clip .mp4/.csv under chunk*/. I checked HuggingFace and ModelScope and didn't find these files.

2. Robotics training code
   The repo includes It Takes Two training (ittakestwo/train.py, train_ua_480P_toy.yaml), but I only see Robotics inference (robots/parallel_inference.py). Is Robotics training code planned to be released, or should we adapt the It Takes Two training pipeline ourselves?

Thanks!