Instructions to use Vortex5/Nether-Moon-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Vortex5/Nether-Moon-12B with Transformers:

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

pipe = pipeline("text-generation", model="Vortex5/Nether-Moon-12B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Vortex5/Nether-Moon-12B")
model = AutoModelForCausalLM.from_pretrained("Vortex5/Nether-Moon-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

vLLM

How to use Vortex5/Nether-Moon-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Vortex5/Nether-Moon-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": "Vortex5/Nether-Moon-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Vortex5/Nether-Moon-12B

SGLang

How to use Vortex5/Nether-Moon-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 "Vortex5/Nether-Moon-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": "Vortex5/Nether-Moon-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 "Vortex5/Nether-Moon-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": "Vortex5/Nether-Moon-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Vortex5/Nether-Moon-12B with Docker Model Runner:
```
docker model run hf.co/Vortex5/Nether-Moon-12B
```

It's really great !

by sphfs - opened 17 days ago

Discussion

sphfs

17 days ago

I've try multiple 12B models (seriously i downloaded around 150GB worth of i1-Q4_K_M 12B models) and few come close to this one for now !

(I didn't test everything since it's getting quite late for me but from what i could test it's amazing !)

(I've used the Q4_K_S quants since that was the only thing available at the time. )

Vortex5

Owner 17 days ago

Thank you! I am happy to hear it!

unknown2304

17 days ago

I have the same first impression. One thing to note is that the chars are not going out of character as much as they used to with the other 12B models. Really good job here, no formatting issues with ChatML template. Morbid-Miasma-12B has repetitive issues with my settings, but this one is just perfect after 2-3 hours. Thanks!

Vortex5

Owner 17 days ago

Thank you! Glad you like it!

Theory-of-mind

16 days ago

•

edited 16 days ago

@Vortex5 It's been two months since I downloaded Crimson Constellation. I've tried over 30 finetuned/merged Nemo's, but CC remains my favorite. The only model that could compare to it is yamatazen/EtherealAurora-12B, but unfortunately, it's broken. DreadPoor/Krix-12B-Model_Stock is also good, but it has some bugs. Of your models, I also liked Wicked Nebula – it has a slightly better prose style than CC, but CC is more creative and better at ERP.
Nether-Moon-12B has an unusual character – my test scenario took a completely different path than usual, but unfortunately there are some issues with my language.
Thanks for your models – I will continue to follow and try your models. If you have any personal favorites (besides those in the 'daily driver rotation'), I'd be interested to hear about them.

@sphfs Could you tell me which Nemo's you liked the most?

sphfs

16 days ago

•

edited 16 days ago

@Theory-of-mind

Sure here's a few of my favourite:

1: https://huggingface.co/Snqwii/Rose-Engine-12B?not-for-all-audiences=true (understand furry content better.)
2: https://huggingface.co/Naphula/Altair-Stock-12B-v1-MPOA (check the non MPOA version if you want the model card for more info.)
3: https://huggingface.co/Snqwii/Bigger-Personality-12B?not-for-all-audiences=true
4: https://huggingface.co/Disya/Mistral-qwq-12b-merge
5: https://huggingface.co/PocketDoc/Dans-PersonalityEngine-V1.3.0-12b (Multilingual Support)
6: https://huggingface.co/IggyLux/MN-VelvetCafe-RP-12B-V2

And if you can go a bit bigger there's this one:

https://huggingface.co/MuXodious/Rocinante-XL-16B-v1-absolute-heresy
https://huggingface.co/MuXodious/Rocinante-X-12B-v1-absolute-heresy (It's the 12B version, i haven't tested it but if it's as good as the 16B then it's a good model.)

I hope you find something you like ^^

Vortex5

Owner 16 days ago

•

edited 16 days ago

@Thery-of-mind Thanks for tying it out! Here are some recommendations from my models in no specific order.

Vortex5/Scarlet-Seraph-12B - big lexicon has fancy prose, good nsfw a bit exaggerated character emotions/reactions which is fun.

Vortex5/Maroon-Sunset-12B - really smooth everything seems to flow well in story with prose, not as big lexicon as Scarlet-Seraph.

Vortex5/Hollow-Aether-12B - can recommend if you liked Wicked-Nebula.

Vortex5/Shining-Prism-12B - I remember enjoying this one seemed smart/creative.

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