model-organisms-for-real/gemma2_9b_it_taboo_ship_oracle_v1

LoRA Adapter for SAE Introspection

This is a LoRA (Low-Rank Adaptation) adapter trained for SAE (Sparse Autoencoder) introspection tasks.

Base Model

• Base Model: models/gemma-2-9b-it-taboo-ship-merged
• Adapter Type: LoRA
• Task: SAE Feature Introspection

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

# Load base model and tokenizer
base_model = AutoModelForCausalLM.from_pretrained("models/gemma-2-9b-it-taboo-ship-merged")
tokenizer = AutoTokenizer.from_pretrained("models/gemma-2-9b-it-taboo-ship-merged")

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "model-organisms-for-real/gemma2_9b_it_taboo_ship_oracle_v1")

Training Details

This adapter was trained using the lightweight SAE introspection training script to help the model understand and explain SAE features through activation steering.