Instructions to use Auroraventures/cipher-simpo-merged with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Auroraventures/cipher-simpo-merged with Transformers:

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

pipe = pipeline("text-generation", model="Auroraventures/cipher-simpo-merged")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("Auroraventures/cipher-simpo-merged")
model = AutoModelForImageTextToText.from_pretrained("Auroraventures/cipher-simpo-merged")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Auroraventures/cipher-simpo-merged with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Auroraventures/cipher-simpo-merged

SGLang

How to use Auroraventures/cipher-simpo-merged 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 "Auroraventures/cipher-simpo-merged" \
    --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": "Auroraventures/cipher-simpo-merged",
		"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 "Auroraventures/cipher-simpo-merged" \
        --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": "Auroraventures/cipher-simpo-merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Auroraventures/cipher-simpo-merged with Docker Model Runner:
```
docker model run hf.co/Auroraventures/cipher-simpo-merged
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Cipher SimPO 🦑⚔️

The kraken learns to refuse the obvious.

Cipher-SimPO-Merged is the anti-slop checkpoint of the Cipher series. Starting from the SFT base, this model is tuned with SimPO (Simple Preference Optimization, Meng et al. 2024) against preference pairs where the rejected response is a textbook Tailwind / lenis.stop() / copy-paste hero and the chosen response is the Awwwards-grade equivalent.

The result is a model that refuses to reach for the easy cliché.

🧠 Base: Auroraventures/cipher-sft-merged
🎯 Method: SimPO with a length-normalized preference objective, γ=0.5, β=2.0
🚫 Rejects: cdn.tailwindcss.com, lenis.stop(), <div class="container mx-auto"> boilerplate, Bootstrap defaults
✅ Prefers: Bespoke CSS tokens, ScrollTrigger timelines, SplitText reveals, bespoke typography, dark-mode palettes
🔀 Continued as: cipher-sft25-real-merged (v3, with real scraped SFT on top)

Quickstart

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
tok = AutoTokenizer.from_pretrained("Auroraventures/cipher-simpo-merged")
m = AutoModelForCausalLM.from_pretrained(
    "Auroraventures/cipher-simpo-merged",
    torch_dtype=torch.bfloat16, device_map="auto",
)
msgs = [
  {"role":"system","content":"You are Cipher, the Code Kraken. Emit complete single-file HTML."},
  {"role":"user","content":"Build a minimal dark-mode hero with GSAP stagger entry — no frameworks."},
]
p = tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True)
print(tok.decode(m.generate(**tok(p,return_tensors='pt').to(m.device), max_new_tokens=2048)[0]))

GGUF sibling: cipher-simpo-merged-Q4_K_M-GGUF.

Slop taxonomy (what SimPO penalizes)

Reject	Prefer
`cdn.tailwindcss.com` script tag	Hand-rolled CSS tokens + `@media`
`lenis.stop()` / `lenis.start()` misuse	Native Lenis on page + GSAP ScrollTrigger integration
`container mx-auto` / `text-center text-white`	CSS custom properties + fluid typography
Bootstrap grid	CSS Grid + `clamp()`
`.fadeIn` / `.fade-up` generic utility classes	Bespoke keyframes named by intent

Pipeline position

cipher-sft-merged ──► [SimPO anti-slop] ──► cipher-simpo-merged (YOU) ──► cipher-sft25-real-merged

License

CC-BY-NC-4.0. Base model: Gemma-4 terms apply.

Built with 🦑 by Aurora Ventures.

Downloads last month: 16

Safetensors

Model size

33B params

Tensor type

BF16

Model tree for Auroraventures/cipher-simpo-merged

Base model

Auroraventures/cipher-sft-merged

Finetuned

(1)

this model

Adapters

1 model

Finetunes

1 model

Quantizations

1 model

Paper for Auroraventures/cipher-simpo-merged

SimPO: Simple Preference Optimization with a Reference-Free Reward

Paper • 2405.14734 • Published May 23, 2024 • 12