model-organisms-for-real/open_instruct_dpo_replication_olmo2_1b_oracle-step-5000

LoRA Adapter for SAE Introspection

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

Base Model

• Base Model: model-organisms-for-real/open_instruct_dpo_replication
• 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("model-organisms-for-real/open_instruct_dpo_replication")
tokenizer = AutoTokenizer.from_pretrained("model-organisms-for-real/open_instruct_dpo_replication")

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "model-organisms-for-real/open_instruct_dpo_replication_olmo2_1b_oracle-step-5000")

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.