Instructions to use moondream/moondream-2b-2025-04-14-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use moondream/moondream-2b-2025-04-14-4bit with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="moondream/moondream-2b-2025-04-14-4bit",
	filename="moondream2-mmproj-f16.gguf",
)

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

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use moondream/moondream-2b-2025-04-14-4bit with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf moondream/moondream-2b-2025-04-14-4bit:F16
# Run inference directly in the terminal:
llama-cli -hf moondream/moondream-2b-2025-04-14-4bit:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf moondream/moondream-2b-2025-04-14-4bit:F16
# Run inference directly in the terminal:
llama-cli -hf moondream/moondream-2b-2025-04-14-4bit:F16

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 moondream/moondream-2b-2025-04-14-4bit:F16
# Run inference directly in the terminal:
./llama-cli -hf moondream/moondream-2b-2025-04-14-4bit:F16

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 moondream/moondream-2b-2025-04-14-4bit:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf moondream/moondream-2b-2025-04-14-4bit:F16

Use Docker

docker model run hf.co/moondream/moondream-2b-2025-04-14-4bit:F16

LM Studio
Jan

vLLM

How to use moondream/moondream-2b-2025-04-14-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "moondream/moondream-2b-2025-04-14-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "moondream/moondream-2b-2025-04-14-4bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/moondream/moondream-2b-2025-04-14-4bit:F16

Ollama
How to use moondream/moondream-2b-2025-04-14-4bit with Ollama:
```
ollama run hf.co/moondream/moondream-2b-2025-04-14-4bit:F16
```

Unsloth Studio

How to use moondream/moondream-2b-2025-04-14-4bit 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 moondream/moondream-2b-2025-04-14-4bit 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 moondream/moondream-2b-2025-04-14-4bit to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for moondream/moondream-2b-2025-04-14-4bit to start chatting

Atomic Chat new
Docker Model Runner
How to use moondream/moondream-2b-2025-04-14-4bit with Docker Model Runner:
```
docker model run hf.co/moondream/moondream-2b-2025-04-14-4bit:F16
```

Lemonade

How to use moondream/moondream-2b-2025-04-14-4bit with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull moondream/moondream-2b-2025-04-14-4bit:F16

Run and chat with the model

lemonade run user.moondream-2b-2025-04-14-4bit-F16

List all available models

lemonade list

snowclipsed commited on Jun 9, 2025

Commit

8cb4b5e

1 Parent(s): e15c30f

remove gsize check, replace with quantized_linear directly

Browse files

Files changed (1) hide show

moondream.py +6 -10

moondream.py CHANGED Viewed

@@ -77,36 +77,32 @@ class MoondreamModel(nn.Module):
         self.vision = build_vision_model(config.vision, dtype)
         self.text = build_text_model(config.text, dtype)
-        # Region Model
-        linear_cls = (
-            QuantizedLinear if config.region.group_size is not None else nn.Linear
-        )
         self.region = nn.ModuleDict(
             {
-                "coord_encoder": linear_cls(
                     config.region.coord_feat_dim, config.region.dim, dtype=dtype
                 ),
                 "coord_decoder": nn.ModuleDict(
                     {
-                        "fc1": linear_cls(
                             config.region.dim, config.region.inner_dim, dtype=dtype
                         ),
-                        "fc2": linear_cls(
                             config.region.inner_dim,
                             config.region.coord_out_dim,
                             dtype=dtype,
                         ),
                     }
                 ),
-                "size_encoder": linear_cls(
                     config.region.size_feat_dim, config.region.dim, dtype=dtype
                 ),
                 "size_decoder": nn.ModuleDict(
                     {
-                        "fc1": linear_cls(
                             config.region.dim, config.region.inner_dim, dtype=dtype
                         ),
-                        "fc2": linear_cls(
                             config.region.inner_dim,
                             config.region.size_out_dim,
                             dtype=dtype,

         self.vision = build_vision_model(config.vision, dtype)
         self.text = build_text_model(config.text, dtype)
         self.region = nn.ModuleDict(
             {
+                "coord_encoder": QuantizedLinear(
                     config.region.coord_feat_dim, config.region.dim, dtype=dtype
                 ),
                 "coord_decoder": nn.ModuleDict(
                     {
+                        "fc1": QuantizedLinear(
                             config.region.dim, config.region.inner_dim, dtype=dtype
                         ),
+                        "fc2": QuantizedLinear(
                             config.region.inner_dim,
                             config.region.coord_out_dim,
                             dtype=dtype,
                         ),
                     }
                 ),
+                "size_encoder": QuantizedLinear(
                     config.region.size_feat_dim, config.region.dim, dtype=dtype
                 ),
                 "size_decoder": nn.ModuleDict(
                     {
+                        "fc1": QuantizedLinear(
                             config.region.dim, config.region.inner_dim, dtype=dtype
                         ),
+                        "fc2": QuantizedLinear(
                             config.region.inner_dim,
                             config.region.size_out_dim,
                             dtype=dtype,