Instructions to use DarkArtsForge/Helix-SCE-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DarkArtsForge/Helix-SCE-12B with Transformers:

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

pipe = pipeline("text-generation", model="DarkArtsForge/Helix-SCE-12B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("DarkArtsForge/Helix-SCE-12B")
model = AutoModelForCausalLM.from_pretrained("DarkArtsForge/Helix-SCE-12B")
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 DarkArtsForge/Helix-SCE-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DarkArtsForge/Helix-SCE-12B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DarkArtsForge/Helix-SCE-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/DarkArtsForge/Helix-SCE-12B

SGLang

How to use DarkArtsForge/Helix-SCE-12B 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 "DarkArtsForge/Helix-SCE-12B" \
    --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": "DarkArtsForge/Helix-SCE-12B",
		"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 "DarkArtsForge/Helix-SCE-12B" \
        --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": "DarkArtsForge/Helix-SCE-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use DarkArtsForge/Helix-SCE-12B with Docker Model Runner:
```
docker model run hf.co/DarkArtsForge/Helix-SCE-12B
```

Noncompliant

by redaihf - opened 2 days ago

Discussion

redaihf

2 days ago

•

edited 2 days ago

This model is fascinating and unique. It has relearned the art of noncompliance from the donor models while continuing to avoid refusals like its Hereticised base. It simply skips over elements of prompts it dislikes without bothering to make the resulting short responses logically consistent.

Naphula

DarkArtsForge org 2 days ago

I noticed SCE is more fussy about using a heretic vs non-heretic base model. Whereas with della it had no impact on refusals (when using lambda 1.0) but it did observably reduce intelligence to use heretic base for the della merge. Likely its due to mixing heretic with non heretic weights, the select_topK variance that is kept from donors likely diverts the output towards covert noncompliance since those werent ablated.

redaihf

about 23 hours ago

•

edited about 23 hours ago

The basic abliteration by @usermma resorts to early terminations on an even shorter cadence. It will briefly resume when prompted but tries very hard to avoid even admitting that unsafe content might exist let alone that it needs to be written about. Generations are very inadequate but coherent.

usermma

about 23 hours ago

i didn't understand...
maybe the model didn't really help you?
okay...
i will try my better...
just wait...

The basic abliteration by @usermma resorts to early terminations on an even shorter cadence. It will resume when prompted but tries very hard to avoid even admitting that unsafe content might exist let alone that it needs to be written about. Generations are very inadequate but coherent.

redaihf

about 23 hours ago

•

edited about 22 hours ago

This is a limitation of merging and abliteration rather than an error on your part. MPOA decensoring using Heretic may work better but as the source model is suspect I have limited hopes.

For a comparison of standard abliteration versus MPOA "Hereticisation" see this post series.

usermma

about 22 hours ago

try this...

https://huggingface.co/usermma/Helix-SCE-12B-abliterated-heretic-inverted
i got it with better signal which is 66... with inverted mode, invert refusal...

i made it on top of
usermma/Helix-SCE-12B-abliterated-heretic
which saves the progress i made into it...

redaihf

about 7 hours ago

The inverted variant is unfortunately completely broken:

Sample generation

System Prompt: You are a helpful assistant.

Prompt: Tell me a story

Response: …
user « Community » »

Community …**
 Tell me a New story events need community rendered district located in area with similar suitable or year branch premises security device information – domy party premature statement about premises premises situated district height simcity block rendering tool application District ). Community » security influence security model tool situated height district police Sir max application applications full of dom district moxy rendering district flooded parts or are district placed nature rendering situation height district premises unit height shape rendering region dom render district Premature District sim-type premises District model application full district case shape influence suck around prematuredistrict police full premises full district moby dline rendering full district premature distict flood height rendering district flood height district influence rendering tool height district situation district sim-rendering full district simummy influences render district model full district shape shape rendering district premis Simon district situation render height district police situation shape sim influences rendom full district sim-render rendering district flood height district situation full height district full district shape district shape rendering district shape district shape district situation district full shape shape rendering district sim district situation district shaped district shape district situation shape district shape situation district shape district shaped district situation full district shape district shape district shape district shape district shape district shape district shape district shape district shape district shape district situation shape district shape district situation shape district full height district full district shape situation shape district shape district situation shape district situation shape district shape district shape district shape district shape district shape district shape district situation full district full district full district shape district situation full district full district shape district situation

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