Instructions to use devingulliver/mamba-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use devingulliver/mamba-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="devingulliver/mamba-gguf",
	filename="mamba-1.4b/mamba-1.4b-f16.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use devingulliver/mamba-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf devingulliver/mamba-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf devingulliver/mamba-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf devingulliver/mamba-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf devingulliver/mamba-gguf:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf devingulliver/mamba-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf devingulliver/mamba-gguf:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf devingulliver/mamba-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf devingulliver/mamba-gguf:Q4_K_M

Use Docker

docker model run hf.co/devingulliver/mamba-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use devingulliver/mamba-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "devingulliver/mamba-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "devingulliver/mamba-gguf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/devingulliver/mamba-gguf:Q4_K_M

Ollama
How to use devingulliver/mamba-gguf with Ollama:
```
ollama run hf.co/devingulliver/mamba-gguf:Q4_K_M
```

Unsloth Studio new

How to use devingulliver/mamba-gguf with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for devingulliver/mamba-gguf to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for devingulliver/mamba-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for devingulliver/mamba-gguf to start chatting

Docker Model Runner
How to use devingulliver/mamba-gguf with Docker Model Runner:
```
docker model run hf.co/devingulliver/mamba-gguf:Q4_K_M
```

Lemonade

How to use devingulliver/mamba-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull devingulliver/mamba-gguf:Q4_K_M

Run and chat with the model

lemonade run user.mamba-gguf-Q4_K_M

List all available models

lemonade list

Mamba GGUF

These are the Mamba base models, converted to GGUF for use with llama.cpp, in a variety of precisions (2, 3, 4, 5, 6, 8, 16, and 32-bit).

Please click "Files and versions" at the top of the page to choose your desired model size, and then click the "📦LFS ↓" button next to your desired quantization.

Here is a table adapted from TheBloke explaining the various precisions:

Quant method	Use case
Q2_K	significant quality loss - not recommended for most purposes
Q3_K_S	very small, high quality loss
Q3_K_M	very small, high quality loss
Q3_K_L	small, substantial quality loss
Q4_0	legacy; small, very high quality loss - prefer using Q3_K_M
Q4_K_S	small, greater quality loss
Q4_K_M	medium, balanced quality - recommended
Q5_0	legacy; medium, balanced quality - prefer using Q4_K_M
Q5_K_S	large, low quality loss - recommended
Q5_K_M	large, very low quality loss - recommended
Q6_K	very large, extremely low quality loss
Q8_0	very large, extremely low quality loss - not recommended
F16	half precision - almost identical to the original
F32	original precision - recommended by the Mamba authors