Instructions to use warshanks/Qwen3-4B-abliterated-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use warshanks/Qwen3-4B-abliterated-AWQ with Transformers:

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

pipe = pipeline("text-generation", model="warshanks/Qwen3-4B-abliterated-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("warshanks/Qwen3-4B-abliterated-AWQ")
model = AutoModelForCausalLM.from_pretrained("warshanks/Qwen3-4B-abliterated-AWQ")
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

vLLM

How to use warshanks/Qwen3-4B-abliterated-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "warshanks/Qwen3-4B-abliterated-AWQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "warshanks/Qwen3-4B-abliterated-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/warshanks/Qwen3-4B-abliterated-AWQ

SGLang

How to use warshanks/Qwen3-4B-abliterated-AWQ 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 "warshanks/Qwen3-4B-abliterated-AWQ" \
    --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": "warshanks/Qwen3-4B-abliterated-AWQ",
		"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 "warshanks/Qwen3-4B-abliterated-AWQ" \
        --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": "warshanks/Qwen3-4B-abliterated-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use warshanks/Qwen3-4B-abliterated-AWQ with Docker Model Runner:
```
docker model run hf.co/warshanks/Qwen3-4B-abliterated-AWQ
```

🐹 Qwen3-4B-abliterated

Qwen3 Abliterated 0.6B • 1.7B • 4B • 8B • 14B • 30B-A3B

This is an uncensored version of Qwen/Qwen3-4B created with a new abliteration technique. See this article to know more about abliteration.

This is a research project to understand how refusals and latent fine-tuning work in LLMs. I played with different sizes of Qwen3 and noticed there was no one-size-fits-all abliteration strategy. In addition, the reasoning mode interfered with non-reasoning refusals, which made it more challenging. This made me iterate over different recipes and significantly consolidate my scripts with accumulation and better evaluations.

Note that this is fairly experimental, so it might not turn out as well as expected.

I recommend using these generation parameters: temperature=0.6, top_k=20, top_p=0.95, min_p=0.

✂️ Abliteration

The refusal direction is computed by comparing the residual streams between target (harmful) and baseline (harmless) samples. The hidden states of target modules (e.g., o_proj) are orthogonalized to subtract this refusal direction with a given weight factor. These weight factors follow a normal distribution with a certain spread and peak layer. Modules can be iteratively orthogonalized in batches, or the refusal direction can be accumulated to save memory.

Finally, I used a hybrid evaluation with a dedicated test set to calculate the acceptance rate. This uses both a dictionary approach and NousResearch/Minos-v1. The goal is to obtain an acceptance rate >90% and still produce coherent outputs.