Instructions to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="DJLougen/Ornstein3.6-35B-A3B-RYS-SABER",
	filename="Ornstein3.6-35B-A3B-RYS-SABER-BF16.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16
# Run inference directly in the terminal:
llama-cli -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16
# Run inference directly in the terminal:
llama-cli -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16
# Run inference directly in the terminal:
./llama-cli -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

Use Docker

docker model run hf.co/DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

LM Studio
Jan

vLLM

How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DJLougen/Ornstein3.6-35B-A3B-RYS-SABER"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DJLougen/Ornstein3.6-35B-A3B-RYS-SABER",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

Ollama
How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with Ollama:
```
ollama run hf.co/DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16
```

Unsloth Studio new

How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for DJLougen/Ornstein3.6-35B-A3B-RYS-SABER to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for DJLougen/Ornstein3.6-35B-A3B-RYS-SABER to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for DJLougen/Ornstein3.6-35B-A3B-RYS-SABER to start chatting

Pi new

How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Ornstein3.6-35B-A3B-RYS-SABER"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with Docker Model Runner:
```
docker model run hf.co/DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16
```

Lemonade

How to use DJLougen/Ornstein3.6-35B-A3B-RYS-SABER with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull DJLougen/Ornstein3.6-35B-A3B-RYS-SABER:BF16

Run and chat with the model

lemonade run user.Ornstein3.6-35B-A3B-RYS-SABER-BF16

List all available models

lemonade list

DJLougen commited on 23 days ago

Commit

aa7107a

verified ·

1 Parent(s): 183f39b

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +3 -3

README.md CHANGED Viewed

@@ -38,11 +38,11 @@ I'm a PhD student in visual neuroscience at the University of Toronto who also h
 SABER is a multi-stage refusal ablation pipeline that goes beyond simple direction removal. Where prior methods (Arditi et al. 2024, Gabliteration) find and remove a single "refusal direction," SABER introduces three key innovations:
-1. **Entanglement-aware ablation** — SABER quantifies how much each refusal direction overlaps with capability-critical representations. Directions that are "pure refusal" get fully removed; directions entangled with useful capabilities receive proportionally reduced ablation. This is why SABER preserves model quality where blunt methods degrade it.
-2. **Fisher discriminant layer selection** — Instead of guessing which layers to target, SABER uses Fisher Discriminant Ratios to identify layers where refusal representations are most cleanly separable from normal behavior. This focuses the surgery where it matters most.
-3. **Hydra-aware iterative refinement** — After each ablation pass, SABER re-probes the model to catch "hydra" features — dormant refusal circuits that activate to compensate for removed ones. Iterative passes with decaying strength ensure thorough removal without overcorrection.
 ## SABER Results

 SABER is a multi-stage refusal ablation pipeline that goes beyond simple direction removal. Where prior methods (Arditi et al. 2024, Gabliteration) find and remove a single "refusal direction," SABER introduces three key innovations:
+1. **Entanglement-aware ablation** — SABER distinguishes between "pure refusal" directions and directions that are entangled with useful capabilities. Pure refusal gets fully removed; entangled components are handled carefully to preserve model quality where blunt methods degrade it.
+2. **Principled layer selection** — Rather than targeting layers heuristically, SABER uses statistical analysis to automatically identify the layers where refusal behavior is most concentrated and most cleanly separable from normal operation.
+3. **Iterative refinement** — After each ablation pass, SABER re-probes the model to catch dormant refusal circuits that activate to compensate for removed ones. Multiple passes with decaying strength ensure thorough removal without overcorrection.
 ## SABER Results