Instructions to use second-state/medgemma-4b-it-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use second-state/medgemma-4b-it-GGUF with Transformers:

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

pipe = pipeline("image-text-to-text", model="second-state/medgemma-4b-it-GGUF")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("second-state/medgemma-4b-it-GGUF")
model = AutoModelForMultimodalLM.from_pretrained("second-state/medgemma-4b-it-GGUF")

llama-cpp-python

How to use second-state/medgemma-4b-it-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="second-state/medgemma-4b-it-GGUF",
	filename="medgemma-4b-it-Q2_K.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": [
				{
					"type": "text",
					"text": "Describe this image in one sentence."
				},
				{
					"type": "image_url",
					"image_url": {
						"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
					}
				}
			]
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use second-state/medgemma-4b-it-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf second-state/medgemma-4b-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf second-state/medgemma-4b-it-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf second-state/medgemma-4b-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf second-state/medgemma-4b-it-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 second-state/medgemma-4b-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf second-state/medgemma-4b-it-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 second-state/medgemma-4b-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf second-state/medgemma-4b-it-GGUF:Q4_K_M

Use Docker

docker model run hf.co/second-state/medgemma-4b-it-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use second-state/medgemma-4b-it-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "second-state/medgemma-4b-it-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "second-state/medgemma-4b-it-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/second-state/medgemma-4b-it-GGUF:Q4_K_M

SGLang

How to use second-state/medgemma-4b-it-GGUF 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 "second-state/medgemma-4b-it-GGUF" \
    --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": "second-state/medgemma-4b-it-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "second-state/medgemma-4b-it-GGUF" \
        --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": "second-state/medgemma-4b-it-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Ollama
How to use second-state/medgemma-4b-it-GGUF with Ollama:
```
ollama run hf.co/second-state/medgemma-4b-it-GGUF:Q4_K_M
```

Unsloth Studio

How to use second-state/medgemma-4b-it-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 second-state/medgemma-4b-it-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 second-state/medgemma-4b-it-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for second-state/medgemma-4b-it-GGUF to start chatting

Atomic Chat new
Docker Model Runner
How to use second-state/medgemma-4b-it-GGUF with Docker Model Runner:
```
docker model run hf.co/second-state/medgemma-4b-it-GGUF:Q4_K_M
```

Lemonade

How to use second-state/medgemma-4b-it-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull second-state/medgemma-4b-it-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.medgemma-4b-it-GGUF-Q4_K_M

List all available models

lemonade list

apepkuss79 commited on May 22, 2025

Commit

b0298ba

verified ·

1 Parent(s): a88f54c

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +107 -1

README.md CHANGED Viewed

@@ -1,3 +1,109 @@
 ---
-license: apache-2.0
 ---

 ---
+base_model: google/medgemma-4b-it
+inference: false
+library_name: transformers
+license: other
+license_name: health-ai-developer-foundations
+license_link: https://developers.google.com/health-ai-developer-foundations/terms
+model_creator: Google
+model_name: medgemma-4b-it
+quantized_by: Second State Inc.
+pipeline_tag: image-text-to-text
+tags:
+  - medical
+  - radiology
+  - clinical-reasoning
+  - dermatology
+  - pathology
+  - ophthalmology
+  - chest-x-ray
 ---
+<!-- header start -->
+<!-- 200823 -->
+<div style="width: auto; margin-left: auto; margin-right: auto">
+<img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;">
+</div>
+<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
+<!-- header end -->
+# medgemma-4b-it-GGUF
+## Original Model
+[google/medgemma-4b-it](https://huggingface.co/google/medgemma-4b-it)
+## Run with LlamaEdge
+- LlamaEdge version: [v0.18.5](https://github.com/LlamaEdge/LlamaEdge/releases/tag/0.18.5) and above
+- Prompt template
+  - Prompt type: `gemma-3`
+  - Prompt string
+    ```text
+    <bos><start_of_turn>user
+    {user_message}<end_of_turn>
+    <start_of_turn>model
+    {model_message}<end_of_turn>model
+    ```
+- Context size: `128000`
+- Run as LlamaEdge service
+  - Chat
+    ```bash
+    wasmedge --dir .:. --nn-preload default:GGML:AUTO:medgemma-4b-it-Q5_K_M.gguf \
+      llama-api-server.wasm \
+      --prompt-template gemma-3 \
+      --ctx-size 128000 \
+      --model-name medgemma-4b
+    ```
+  - Images
+    **Note that input images are required to be normalized to 896 x 896 resolution and encoded to 256 tokens each**
+    ```bash
+    wasmedge --dir .:. --nn-preload default:GGML:AUTO:medgemma-4b-it-Q5_K_M.gguf \
+      llama-api-server.wasm \
+      --prompt-template gemma-3 \
+      --llava-mmproj medgemma-4b-it-mmproj-f16.gguf \
+      --ctx-size 128000 \
+      --model-name medgemma-4b
+    ```
+- Run as LlamaEdge command app
+  ```bash
+  wasmedge --dir .:. \
+    --nn-preload default:GGML:AUTO:medgemma-4b-it-Q5_K_M.gguf \
+    llama-chat.wasm \
+    --prompt-template gemma-3 \
+    --ctx-size 128000
+  ```
+## Quantized GGUF Models
+| Name | Quant method | Bits | Size | Use case |
+| ---- | ---- | ---- | ---- | ----- |
+| [medgemma-4b-it-Q2_K.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q2_K.gguf)     | Q2_K   | 2 | 1.73 GB| smallest, significant quality loss - not recommended for most purposes |
+| [medgemma-4b-it-Q3_K_L.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q3_K_L.gguf) | Q3_K_L | 3 | 2.24 GB| small, substantial quality loss |
+| [medgemma-4b-it-Q3_K_M.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q3_K_M.gguf) | Q3_K_M | 3 | 2.10 GB| very small, high quality loss |
+| [medgemma-4b-it-Q3_K_S.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q3_K_S.gguf) | Q3_K_S | 3 | 1.94 GB| very small, high quality loss |
+| [medgemma-4b-it-Q4_0.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q4_0.gguf)     | Q4_0   | 4 | 2.36 GB| legacy; small, very high quality loss - prefer using Q3_K_M |
+| [medgemma-4b-it-Q4_K_M.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q4_K_M.gguf) | Q4_K_M | 4 | 2.49 GB| medium, balanced quality - recommended |
+| [medgemma-4b-it-Q4_K_S.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q4_K_S.gguf) | Q4_K_S | 4 | 2.38 GB| small, greater quality loss |
+| [medgemma-4b-it-Q5_0.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q5_0.gguf)     | Q5_0   | 5 | 2.76 GB| legacy; medium, balanced quality - prefer using Q4_K_M |
+| [medgemma-4b-it-Q5_K_M.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q5_K_M.gguf) | Q5_K_M | 5 | 2.83 GB| large, very low quality loss - recommended |
+| [medgemma-4b-it-Q5_K_S.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q5_K_S.gguf) | Q5_K_S | 5 | 2.76 GB| large, low quality loss - recommended |
+| [medgemma-4b-it-Q6_K.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q6_K.gguf)     | Q6_K   | 6 | 3.19 GB| very large, extremely low quality loss |
+| [medgemma-4b-it-Q8_0.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-Q8_0.gguf)     | Q8_0   | 8 | 4.13 GB| very large, extremely low quality loss - not recommended |
+| [medgemma-4b-it-f16.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-f16.gguf)       | f16   | 16 | 7.77 GB|  |
+| [medgemma-4b-it-mmproj-f16.gguf](https://huggingface.co/second-state/medgemma-4b-it-GGUF/blob/main/medgemma-4b-it-mmproj-f16.gguf)       | f16   | 16 | 851 MB|  |
+*Quantized with llama.cpp b5452*