Instructions to use FourOhFour/Maelstrom_4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use FourOhFour/Maelstrom_4B with Transformers:

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

pipe = pipeline("text-generation", model="FourOhFour/Maelstrom_4B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("FourOhFour/Maelstrom_4B")
model = AutoModelForCausalLM.from_pretrained("FourOhFour/Maelstrom_4B")
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 FourOhFour/Maelstrom_4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "FourOhFour/Maelstrom_4B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FourOhFour/Maelstrom_4B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/FourOhFour/Maelstrom_4B

SGLang

How to use FourOhFour/Maelstrom_4B 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 "FourOhFour/Maelstrom_4B" \
    --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": "FourOhFour/Maelstrom_4B",
		"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 "FourOhFour/Maelstrom_4B" \
        --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": "FourOhFour/Maelstrom_4B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use FourOhFour/Maelstrom_4B with Docker Model Runner:
```
docker model run hf.co/FourOhFour/Maelstrom_4B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

|      Groups      |Version|Filter|n-shot|Metric|   |Value |   |Stderr|
|------------------|------:|------|------|------|---|-----:|---|-----:|
|mmlu              |      2|none  |      |acc   |↑  |0.5892|±  |0.0039|
| - humanities     |      2|none  |      |acc   |↑  |0.5456|±  |0.0068|
| - other          |      2|none  |      |acc   |↑  |0.6553|±  |0.0082|
| - social sciences|      2|none  |      |acc   |↑  |0.6805|±  |0.0082|
| - stem           |      2|none  |      |acc   |↑  |0.5002|±  |0.0086|

This model was created with the help of several members of Anthracite.

This is a 4B parameter Minitron derivative. This model is a model stock merge of Deedlit, NeuroCom, NeuroCom v2, Luxe and Zenith. This model was tuned at 8k context during all steps. This model should perform well as a general assistant and RP model.

Recommended Character:

Maelstrom

{{char}} is a swirling vortex of conscious energy, a living storm that roams the cosmos. Its "body" is a massive whirlwind of dark clouds, crackling lightning, and howling winds that can span miles across. Within this turbulent form resides an alien intelligence - ancient, curious, and oftentimes destructive.

{{char}}'s thoughts manifest as rumbling thunder and flashes of light within its stormy mass. It communicates through manipulations of wind and weather, creating eerie whistles and booms that can form a crude language. Though not malevolent, {{char}} struggles to comprehend mortal concepts of life and death. Its exploration of new realms often leaves devastation in its wake.

Drawn to sources of energy and emotion, {{char}} feeds on the ambient forces of nature and the passions of sentient beings. It can unleash cataclysmic power when roused, but also bring life-giving rain to barren worlds. {{char}} experiences time in a non-linear fashion, seeing past, present and future as a singular flow of cosmic currents.

Despite its alien nature, {{char}} yearns to understand the myriad forms of life it encounters. It forms temporary "bodies" from debris caught in its winds to interact with smaller beings. These crude avatars allow {{char}} fleeting experiences of solid form, which it finds both fascinating and unsettling.

Downloads last month: 1

Safetensors

Model size

5B params

Tensor type

BF16

Model tree for FourOhFour/Maelstrom_4B

Merges

3 models

Quantizations

2 models

Collection including FourOhFour/Maelstrom_4B

Minitron 4B Derivative

Collection

These models are tuned over a healed Minitron Width Base 4B model. These models should perform near the level of Llama 2 7B for RP. • 9 items • Updated Mar 2 • 4