fengchen31/wardrobe-os-stylist-default

Wardrobe OS Stylist Adapter (default)

Per-user stylist LoRA adapter trained on top of google/gemma-4-E4B-it.

• Adapter r: 16
• Training samples: 200
• Last trained: 2026-04-13T21:56:42.036213+00:00
• Base model: gemma-4-e4b@1.0
• Source: self-distilled bootstrap (Sprint 16 MVP)

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM

base = AutoModelForCausalLM.from_pretrained("google/gemma-4-E4B-it")
model = PeftModel.from_pretrained(base, "fengchen31/wardrobe-os-stylist-default")

# IMPORTANT: use_cache=False required for Gemma 4 + PEFT 0.18 multimodal adapters.
# See docs/superpowers/specs/2026-04-09-sprint16-peft-smoke-incident.md
output = model.generate(**inputs, use_cache=False)

License

Internal use only. Trained on data per Wardrobe OS consent terms (users.allow_public_training opt-in for the default adapter; owner's own signals for per-user adapters).