Instructions to use narukijima/stabilizer-mini-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use narukijima/stabilizer-mini-v1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="narukijima/stabilizer-mini-v1", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("narukijima/stabilizer-mini-v1", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("narukijima/stabilizer-mini-v1", trust_remote_code=True, device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use narukijima/stabilizer-mini-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "narukijima/stabilizer-mini-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "narukijima/stabilizer-mini-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/narukijima/stabilizer-mini-v1

SGLang

How to use narukijima/stabilizer-mini-v1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "narukijima/stabilizer-mini-v1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "narukijima/stabilizer-mini-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "narukijima/stabilizer-mini-v1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "narukijima/stabilizer-mini-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use narukijima/stabilizer-mini-v1 with Docker Model Runner:
```
docker model run hf.co/narukijima/stabilizer-mini-v1
```

stabilizer-mini-v1

Overview

This is a test model.

Technical notes

Base: openai/gpt-oss-20b (bf16)
Steering: rank-1 delta on Q/K/V across 24 layers (RMSNorm-aware)
Concept vector: concept_vec_v15k.pt, shape [24, 6, 2880], gain=0.5
Checkpoint: single baked weights (no LoRA/adapters; knowledge ≈ base)
Data used: neutral_examples=86376, pairs_used=14400
Source files: narukijima/stabilizer → S_instruction_pairs_en.jsonl, S_instruction_pairs_ja.jsonl
Inference: use base tokenizer & chat template

Quick inference

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
M = "narukijima/stabilizer-mini-v1"
tok = AutoTokenizer.from_pretrained(M, trust_remote_code=True)
mdl = AutoModelForCausalLM.from_pretrained(
    M, torch_dtype=torch.bfloat16, device_map='auto', trust_remote_code=True
)
msgs = [{"role":"user","content":"test"}]
p = tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True)
out = mdl.generate(**tok(p, return_tensors='pt').to(mdl.device),
                   max_new_tokens=64, do_sample=True, temperature=0.7)
print(tok.decode(out[0], skip_special_tokens=True))

Downloads last month: 19

Safetensors

Model size

21B params

Tensor type

BF16

Model tree for narukijima/stabilizer-mini-v1

Base model

openai/gpt-oss-20b

Finetuned

(546)

this model