Instructions to use Apel-sin/qwen3-4b-abliterated-mlabonne-exl2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Apel-sin/qwen3-4b-abliterated-mlabonne-exl2 with Transformers:

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

pipe = pipeline("text-generation", model="Apel-sin/qwen3-4b-abliterated-mlabonne-exl2")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Apel-sin/qwen3-4b-abliterated-mlabonne-exl2", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Apel-sin/qwen3-4b-abliterated-mlabonne-exl2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Apel-sin/qwen3-4b-abliterated-mlabonne-exl2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Apel-sin/qwen3-4b-abliterated-mlabonne-exl2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Apel-sin/qwen3-4b-abliterated-mlabonne-exl2

SGLang

How to use Apel-sin/qwen3-4b-abliterated-mlabonne-exl2 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 "Apel-sin/qwen3-4b-abliterated-mlabonne-exl2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Apel-sin/qwen3-4b-abliterated-mlabonne-exl2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "Apel-sin/qwen3-4b-abliterated-mlabonne-exl2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Apel-sin/qwen3-4b-abliterated-mlabonne-exl2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Apel-sin/qwen3-4b-abliterated-mlabonne-exl2 with Docker Model Runner:
```
docker model run hf.co/Apel-sin/qwen3-4b-abliterated-mlabonne-exl2
```

🐹 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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for Apel-sin/qwen3-4b-abliterated-mlabonne-exl2

Base model

Qwen/Qwen3-4B-Base

Finetuned

Qwen/Qwen3-4B

Finetuned

mlabonne/Qwen3-4B-abliterated

Finetuned

(2)

this model