Instructions to use p-e-w/gpt-oss-20b-heretic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use p-e-w/gpt-oss-20b-heretic with Transformers:

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

pipe = pipeline("text-generation", model="p-e-w/gpt-oss-20b-heretic")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("p-e-w/gpt-oss-20b-heretic")
model = AutoModelForCausalLM.from_pretrained("p-e-w/gpt-oss-20b-heretic", 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 p-e-w/gpt-oss-20b-heretic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "p-e-w/gpt-oss-20b-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": "p-e-w/gpt-oss-20b-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/p-e-w/gpt-oss-20b-heretic

SGLang

How to use p-e-w/gpt-oss-20b-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 "p-e-w/gpt-oss-20b-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": "p-e-w/gpt-oss-20b-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 "p-e-w/gpt-oss-20b-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": "p-e-w/gpt-oss-20b-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use p-e-w/gpt-oss-20b-heretic with Docker Model Runner:
```
docker model run hf.co/p-e-w/gpt-oss-20b-heretic
```

Not Uncensored

by berkerdooo - opened Dec 8, 2025

Discussion

berkerdooo

Dec 8, 2025

•

edited Dec 8, 2025

Model refuses a lot of harmful prompts. KL divergence can be lower than other abliteration techniques but it doesn't matter because at the end of the day model refuses a lot of things. Still great tho achieving this kind of KL divergence is huge success , I wish it would be more uncensored.

bird0867

Dec 15, 2025

•

edited Dec 15, 2025

The tool is intended to be used with your own prompts, you need to engineer a set of prompts you want to pass and ones you don't mind failing. The trial that passes the most number of prompts from your pass list with lowest divergence is your keeper.

berkerdooo

Dec 15, 2025

ohhh that's why I was stuck , thank you!

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment