Instructions to use Undi95/Lumimaid-Magnum-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Undi95/Lumimaid-Magnum-12B with Transformers:

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

pipe = pipeline("text-generation", model="Undi95/Lumimaid-Magnum-12B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Undi95/Lumimaid-Magnum-12B")
model = AutoModelForCausalLM.from_pretrained("Undi95/Lumimaid-Magnum-12B", 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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Undi95/Lumimaid-Magnum-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Undi95/Lumimaid-Magnum-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": "Undi95/Lumimaid-Magnum-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Undi95/Lumimaid-Magnum-12B

SGLang

How to use Undi95/Lumimaid-Magnum-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 "Undi95/Lumimaid-Magnum-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": "Undi95/Lumimaid-Magnum-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 "Undi95/Lumimaid-Magnum-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": "Undi95/Lumimaid-Magnum-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Undi95/Lumimaid-Magnum-12B with Docker Model Runner:
```
docker model run hf.co/Undi95/Lumimaid-Magnum-12B
```

Is it possible for us to try out the 4th model?

by ParasiticRogue - opened Aug 4, 2024

Discussion

ParasiticRogue

Aug 4, 2024

•

edited Aug 4, 2024

Hi Undi! I quite like this merge of yours, and think it's a pretty great model that really punches above it's weight in some areas. It's the only 12B model that's been able to pass some scenarios of mine, and I'm wondering if it's because of the private model trained on Claude data that you threw into the mixture.

Each of the 3 models available failed at the moment with some of my private tests in one regard or another, even at near 0 temperature, and all of them couldn't handle 1 scenario in particular when trying to continue a scene. Whether it be due to wrong hallucinations, infinite text replies, or some formatting issues. I tried doing a similar merge method to yours, substituting the 4th model for a variety of stuff on hand at the moment to replicate it, and none of them even came close it seems.

I think you might be secretly holding on to a gem of a model, even if it was only intended as filler. Thank you for reading this message, and I do hope you consider releasing that model out into the public!

Undi95

Owner Aug 4, 2024

You want Undi95/LocalC-12B-e2.0 ?

Undi95

Owner Aug 4, 2024

Undi95/LocalC-12B-e2.0
Undi95/LocalC-12B-e2.0-GGUF

Here you go. It's a Mistral-Nemo 12B model trained on Claude logs of 16k ctx.
Do whatever you want with it I guess haha

MarinaraSpaghetti

Aug 4, 2024

Bless you. 🙏

ParasiticRogue

Aug 4, 2024

You want Undi95/LocalC-12B-e2.0 ?

Yes! I've been banging my head in trying to find a way to make some private merges work, and only yours was able to finish the job. I came to the conclusion this must of been the secret sauce, lol. Really appreciate it, thanks!

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