PixlStash adds a ranked tag-review queue to clean image datasets before LoRA training
Quick-Confection96 · reddit · 2026-07-21
PixlStash adds a ranked tag-review queue for cleaning training data
The author says their open-source, self-hosted image database PixlStash now includes a tag-review feature to clean up bad auto-tags before training or evaluation.
The workflow is meant to help with common issues in image datasets:
- auto-taggers often produce wrong tags
- systematic mistakes can be hard to spot when browsing image-by-image
- manual cleanup via booru-style editors or text-file search/replace is slow
The new review queue ranks the least certain tag suggestions first, so users can fix the most suspicious labels before they contaminate a LoRA or tag-model training set. It also supports:
- queue-based review instead of a single linear list
- returning to unfinished tag reviews later
- remembering rejected labels
- locking a dataset so captions, tags, and scores become read-only
The author asks whether tag cleaning is still important for LoRA training with newer models.
Related event: PixlStash Introduces Tag Review Queue for Data Cleaning(2 posts)→
More from Companies & People
- FactoryAI gave back its first millions, then shipped Droid CLI two years later — matanSF · 2026-07-22
- A post says AI teams should drop the research scientist vs engineer split — jsuarez · 2026-07-22
- YC startup Vendo launches an open-source customization layer for user-built micro-apps — ycombinator · 2026-07-22
- Prescience launches publicly as an AI-native health insurance company — ycombinator · 2026-07-22
- SF AI crowd swaps poker for a bullet and bughouse chess night — Jackyhuang · 2026-07-22
- Polymarket puts Anthropic’s year-end IPO odds at 64% amid patent suit — Polymarket · 2026-07-22