Simplest way to do this would be stopping the generation loop after every individual step is completed, then inserting your next scripted step. I'm not sure if this is doable with just templating.

I didn't ask for anything. They randomly emailed me with this.

Hugging Face put up a similar dataset (only 1 million rows), but they unfortunately removed it :( This is an attempt to undo that injustice, and provide double the amount of data. The data gathering process is fully legal, as per the Bluesky Terms & Service. huggingface.co/datasets/alp...