mlabonne/NeuralLlama-3-8B-Instruct-abliterated

Llama-3-8B-Instruct-abliterated-dpomix

This model is an experimental DPO fine-tune of an abliterated Llama 3 8B Instruct model on the full mlabonne/orpo-dpo-mix-40k dataset. It improves Llama 3 8B Instruct's performance while being uncensored.

🔎 Applications

This is an uncensored model. You can use it for any application that doesn't require alignment, like role-playing.

Tested on LM Studio using the "Llama 3" preset.

⚡ Quantization

• GGUF: https://huggingface.co/mlabonne/Llama-3-8B-Instruct-abliterated-dpomix-GGUF

🏆 Evaluation

Open LLM Leaderboard

This model improves the performance of the abliterated source model and recovers the MMLU that was lost in the abliteration process.

Nous

Model	Average	AGIEval	GPT4All	TruthfulQA	Bigbench
mlabonne/Llama-3-8B-Instruct-abliterated-dpomix 📄	52.26	41.6	69.95	54.22	43.26
meta-llama/Meta-Llama-3-8B-Instruct 📄	51.34	41.22	69.86	51.65	42.64
failspy/Meta-Llama-3-8B-Instruct-abliterated-v3 📄	51.21	40.23	69.5	52.44	42.69
abacusai/Llama-3-Smaug-8B 📄	49.65	37.15	69.12	51.66	40.67
mlabonne/OrpoLlama-3-8B 📄	48.63	34.17	70.59	52.39	37.36
meta-llama/Meta-Llama-3-8B 📄	45.42	31.1	69.95	43.91	36.7

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "mlabonne/Llama-3-8B-Instruct-abliterated-dpomix"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])