Instructions to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="hung2903/gemma-4-E2B-vaccine-xai-merged-qat",
	filename="gemma-4-E2B-it-qat-q4_0-unquantized.F16-mmproj.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat 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 hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16
# Run inference directly in the terminal:
llama cli -hf hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16
# Run inference directly in the terminal:
llama cli -hf hung2903/gemma-4-E2B-vaccine-xai-merged-qat: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 hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16
# Run inference directly in the terminal:
./llama-cli -hf hung2903/gemma-4-E2B-vaccine-xai-merged-qat: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 hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16

Use Docker

docker model run hf.co/hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16

LM Studio
Jan

vLLM

How to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hung2903/gemma-4-E2B-vaccine-xai-merged-qat"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hung2903/gemma-4-E2B-vaccine-xai-merged-qat",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16

Ollama
How to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat with Ollama:
```
ollama run hf.co/hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16
```

Unsloth Studio

How to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat 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 hung2903/gemma-4-E2B-vaccine-xai-merged-qat 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 hung2903/gemma-4-E2B-vaccine-xai-merged-qat to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for hung2903/gemma-4-E2B-vaccine-xai-merged-qat to start chatting

Atomic Chat new
Docker Model Runner
How to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat with Docker Model Runner:
```
docker model run hf.co/hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16
```

Lemonade

How to use hung2903/gemma-4-E2B-vaccine-xai-merged-qat with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull hung2903/gemma-4-E2B-vaccine-xai-merged-qat:F16

Run and chat with the model

lemonade run user.gemma-4-E2B-vaccine-xai-merged-qat-F16

List all available models

lemonade list

VaccineNLP - Gemma-4 E2B QAT Misinfo Engine (GGUF Q4_0)

File GGUF Q4_0 của adapter QAT hung2903/gemma-4-E2B-unsloth-vaccine-xai-qat merged với base google/gemma-4-E2B-it-qat-q4_0-unquantized.

Mô hình này chỉ phục vụ trục Misinfo chính theo taxonomy nội bộ Fake / Real và sinh JSON có cấu trúc tương thích parse_deep_output.

Hiệu suất (Gold Test Set, n=186) — chấm trên RAW model output (trung thực)

Engine này CHỈ tái phân loại trục Misinfo (Fake/Real) theo kiến trúc tách trục (§3.10); Stance & Sentiment do PhoBERT đảm nhiệm. Số liệu đo trên đầu ra thô của mô hình (không qua tầng sửa lỗi/repair) nên phản ánh đúng năng lực thật.

Metric	Score
Macro F1 (Misinfo)	0.6428
Recall (Fake)	0.7143
Precision (Fake)	0.3333
F1 (Fake)	0.4545
F1 (Real)	0.831
Parse success rate	96.8%
Banned-label outputs	0

Confusion matrix (hàng = nhãn thật, cột = dự đoán):

	→ Fake	→ Real
Fake (n=28)	20	8
Real (n=158)	40	118

Acceptance gate: parse-fail < 5%, banned-label = 0, recall Fake >= 0.70 — tất cả ĐẠT. Output là JSON có cấu trúc {"label","confidence","sufficient_context","rationale","key_spans"} tương thích parse_deep_output; confidence/key_spans cuối do tầng code hệ thống xác minh.

Sử dụng với LM Studio / Ollama local

import openai
client = openai.OpenAI(base_url="http://localhost:1234/v1", api_key="lm-studio")
response = client.chat.completions.create(
    model="local-model",
    messages=[{"role": "user", "content": "Van ban: Vac-xin COVID gay vo sinh."}],
    max_tokens=512,
    temperature=0.1,
)
print(response.choices[0].message.content)

Trích dẫn

@thesis{vaccinenlp2026,
  title={Ứng dụng Xử lý Ngôn ngữ Tự nhiên trong phát hiện thông tin sai lệch về vaccine},
  author={Kim Mạnh Hưng and Đinh Lê Quỳnh Phương},
  school={Trường Đại học Y tế Công cộng (HUPH)},
  year={2026},
}