Load sharded data into a distributed array

Reads fragmented datasets (.npy shards, Parquet, HDF5, custom binary) into a cuPyNumeric distributed array when row counts vary across files or standard loaders don't fit.

Best for: Data engineers distributing irregular or custom-layout datasets across GPU/CPU clusters.

Engineering / pipelines-dataatomicfor-engineersneeds-integrationfrom-file

Source

Creator's repository · nvidia/skills

View on GitHub

License: CC-BY-4.0 OR Apache-2.0

Security

Verified — safe to install
Passed all 3 independent security checks
Checked by 3 independent security firms
Does it try to trick the AI?NoSAFE · Gen Agent Trust Hub
Does it sneak in hidden code?NoNo alerts · Socket
Does it have known bugs?NoLow risk · Snyk