README.md

metadata

license: cc-by-nc-sa-4.0
datasets:
  - taigasan/e6-visual-ratings
base_model:
  - lodestones/tagger-experiment

Aesthetic Classifier Trainer

Community:

https://discord.gg/nm3mhfhP

Quick start:

1. source venv/bin/activate
2. pip install -r requirements.txt
3. WANDB_API_KEY= python train.py

Notes:

• tagger_proto.safetensors came from https://huggingface.co/lodestones/tagger-experiment
• embed_cache.pt is included so training can run turn-key without fetching local samples for embedding cache build.

Credits:

• All community members who are contributing ratings via https://taigasan-e6-visual-ratings.hf.space/
• Thessalo https://huggingface.co/Thouph
• Taigasan https://huggingface.co/taigasan