compress in parts (on linux)
tar -I 'zstd -9' -cvf - tiffs/ | split -b 4600M - disfor.tar.zst.
upload to hf
uvx --from "huggingface-hub" hf auth login uvx --from "huggingface-hub" hf upload JR-DIGITAL/DISFOR ./data . --repo-type dataset --include tar.zst
(have to wait a bit, nothing will be printed to stdout while data is prepared)