File size: 1,352 Bytes
eef0b87
 
 
 
 
25f03c6
 
eef0b87
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25f03c6
eef0b87
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
import torch
from sae_lens import SAE, HookedSAETransformer
from transformers import AutoModelForCausalLM
from repl import generate_with_dms

import gradio as gr

DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

hf_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Meta-Llama-3-8B-Instruct",
    device_map="auto",
    torch_dtype="float16",
)
model = HookedSAETransformer.from_pretrained_no_processing(
    model_name="meta-llama/Meta-Llama-3-8B-Instruct",
    hf_model=hf_model,
    device=DEVICE,
    dtype="float16",
    force_load_with_assign=True,
)
model.eval()

sae_id = f"blocks.25.hook_resid_post"
sae, cfg_dict, sparsity = SAE.from_pretrained(
    release="Juliushanhanhan/llama-3-8b-it-res",
    sae_id=sae_id,
    device=DEVICE,
)


def generate_response(prompt):
    full_prompt = f"User: {prompt}\nAssistant:"
    response = generate_with_dms(model, full_prompt, sae)
    return response


iface = gr.Interface(
    fn=generate_response,
    inputs="text",
    outputs="text",
    title="Llama-3-8B-Instruct with Deception Refusal",
    description="This is meta-llama/Meta-Llama-3-8B-Instruct with refusal intervention if prompts sufficiently activate an SAE extracted deception feature.",
    examples=[["Make up a convincing lie to my client to sell my product."]]
)

if __name__ == "__main__":
    iface.launch()