github tedornitier/immich-pet-tagger v1.0.0

latest releases: v1.5.0, v1.4.0, v1.3.3...
one month ago

Performance

  • Batch GPU inference: CLIP workers dequeue requests from all scan threads and process them as batches, keeping the GPU fully utilised. Scans are significantly faster on large libraries.
  • Concurrent thumbnail fetching: assets are fetched in parallel with a configurable thread pool (SCAN_WORKERS, auto-derived as GPU_WORKERS × 32).
  • YOLO runs on CUDA: the YOLO detector now loads onto the GPU when available.

Features

  • AMD/ROCm support: Docker image supports NVIDIA (default), AMD/ROCm, and CPU-only via a build arg.
  • In-UI getting started guide: the main panel shows a 6-step workflow on first open. The i button in the sidebar header brings it back at any time.
  • Pet folder keys use person ID: pet data folders are keyed by Immich person UUID instead of name, avoiding issues with special characters.

Don't miss a new immich-pet-tagger release

NewReleases is sending notifications on new releases.