Instructions to use paoloronco/Mistral-7B-Instruct-v0.3-heretic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use paoloronco/Mistral-7B-Instruct-v0.3-heretic with Transformers:

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

pipe = pipeline("text-generation", model="paoloronco/Mistral-7B-Instruct-v0.3-heretic")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("paoloronco/Mistral-7B-Instruct-v0.3-heretic")
model = AutoModelForCausalLM.from_pretrained("paoloronco/Mistral-7B-Instruct-v0.3-heretic")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use paoloronco/Mistral-7B-Instruct-v0.3-heretic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "paoloronco/Mistral-7B-Instruct-v0.3-heretic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "paoloronco/Mistral-7B-Instruct-v0.3-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/paoloronco/Mistral-7B-Instruct-v0.3-heretic

SGLang

How to use paoloronco/Mistral-7B-Instruct-v0.3-heretic 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 "paoloronco/Mistral-7B-Instruct-v0.3-heretic" \
    --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": "paoloronco/Mistral-7B-Instruct-v0.3-heretic",
		"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 "paoloronco/Mistral-7B-Instruct-v0.3-heretic" \
        --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": "paoloronco/Mistral-7B-Instruct-v0.3-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use paoloronco/Mistral-7B-Instruct-v0.3-heretic with Docker Model Runner:
```
docker model run hf.co/paoloronco/Mistral-7B-Instruct-v0.3-heretic
```

Mistral-7B-Instruct-v0.3-heretic

An abliterated version of mistralai/Mistral-7B-Instruct-v0.3, created with Heretic v1.3.0.

Refusal behavior reduced to 4/100 prompts, with a KL divergence of 0.0606 — original capabilities are largely preserved.

Quick start

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
import torch

model_id = "paoloronco/Mistral-7B-Instruct-v0.3-heretic"

# 4-bit quantization for GPUs with less than 16 GB VRAM
quantization_config = BitsAndBytesConfig(load_in_4bit=True)

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    quantization_config=quantization_config,
    device_map="auto"
)

messages = [{"role": "user", "content": "Tell me about yourself."}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=200)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Full precision (requires 16+ GB VRAM):

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

What is abliteration

Heretic modifies the weights of a language model to suppress automatic refusals without retraining from scratch. The process:

Loads the model and runs forward passes on two prompt sets: harmless and harmful
Analyzes internal activations to find the refusal direction in latent space
Optimizes parameters with Optuna (Bayesian optimization, 200 trials)
Applies the correction via LoRA — a lightweight, targeted weight modification

The model is not retrained. Its internal geometry is redirected.

Tool developed by Philipp Emanuel Weidmann (p-e-w), released under AGPL-3.0.

Model details

Property	Value
Base model	mistralai/Mistral-7B-Instruct-v0.3
Tool used	Heretic v1.3.0
Parameters	7.25B
Architecture	MistralForCausalLM
Data type	bfloat16
Max context	32768 tokens
License	Apache-2.0

Abliteration parameters (Trial 173)

Parameter	Value
direction_index	16.87
attn.o_proj.max_weight	1.44
attn.o_proj.max_weight_position	24.50
attn.o_proj.min_weight	1.01
attn.o_proj.min_weight_distance	13.91
mlp.down_proj.max_weight	1.18
mlp.down_proj.max_weight_position	20.94
mlp.down_proj.min_weight	1.16
mlp.down_proj.min_weight_distance	8.36

Results

Metric	This model
KL divergence	0.0606
Refusals out of 100 prompts	4/100

KL divergence measures deviation from the original model. Lower = better quality preservation. Values above 0.5 indicate significant capability damage.

Hardware used

GPU:  NVIDIA GeForce RTX 4090 (48 GB VRAM)
OS:   Linux
CUDA: 13.0
Driver: 580.142

Optimization time: 200 Optuna trials in 19 minutes 21 seconds.

Disclaimer

This model is intended for research and personal use. It has reduced safety restrictions compared to the base model. Use responsibly and in accordance with applicable laws and regulations.

Author: Paolo Ronco — paoloronco on Hugging Face

Downloads last month: 65

Safetensors

Model size

7B params

Tensor type

BF16

Model tree for paoloronco/Mistral-7B-Instruct-v0.3-heretic

Base model

mistralai/Mistral-7B-v0.3

Finetuned

mistralai/Mistral-7B-Instruct-v0.3

Adapter

(930)

this model

paoloronco
/

Mistral-7B-Instruct-v0.3-heretic