sk Skill · whitecircle
data
Get a dataset into the right shape for a Halo training method, and pre-process it (tokenize / pack / shard) when worth it. USER-INVOKED ONLY. Use when the user asks "what format does SFT/DPO/GRPO/reward/embedding expect", "prepare/tokenize/pack/shard my dataset", "load from S3/HF Hub", "which collator", "padding-free / completions-only", or hits a data-loading error (wrong schema, embedding OOB at step 0, presharded…
Open on skills.sh ↗read 2026-09-17
- installs 8w
- 0
- 30-day movement
- starts with the next reading
- Related entries
- 4
- Connections
- 4
Python
- Host repository
- whitecircle/halo
- Allowed tools
- Read, Bash, Grep, Glob
- Host stars
- 20
- Host language
- Python