Instructions to use ngkhoi/vietron-4b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ngkhoi/vietron-4b with Transformers:

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

pipe = pipeline("text-generation", model="ngkhoi/vietron-4b")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("ngkhoi/vietron-4b")
model = AutoModelForMultimodalLM.from_pretrained("ngkhoi/vietron-4b", device_map="auto")
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?"}
        ]
    },
]
inputs = processor.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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use ngkhoi/vietron-4b 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 ngkhoi/vietron-4b:Q8_0
# Run inference directly in the terminal:
llama cli -hf ngkhoi/vietron-4b:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ngkhoi/vietron-4b:Q8_0
# Run inference directly in the terminal:
llama cli -hf ngkhoi/vietron-4b:Q8_0

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf ngkhoi/vietron-4b:Q8_0
# Run inference directly in the terminal:
./build/bin/llama cli -hf ngkhoi/vietron-4b:Q8_0

Use Docker

docker model run hf.co/ngkhoi/vietron-4b:Q8_0

Llama new
LM Studio
Jan

vLLM

How to use ngkhoi/vietron-4b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ngkhoi/vietron-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": "ngkhoi/vietron-4b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ngkhoi/vietron-4b:Q8_0

SGLang

How to use ngkhoi/vietron-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 "ngkhoi/vietron-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": "ngkhoi/vietron-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 "ngkhoi/vietron-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": "ngkhoi/vietron-4b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use ngkhoi/vietron-4b with Ollama:
```
ollama run hf.co/ngkhoi/vietron-4b:Q8_0
```
Unsloth Desktop
Docker Model Runner
How to use ngkhoi/vietron-4b with Docker Model Runner:
```
docker model run hf.co/ngkhoi/vietron-4b:Q8_0
```

Lemonade

How to use ngkhoi/vietron-4b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ngkhoi/vietron-4b:Q8_0

Run and chat with the model

lemonade run user.vietron-4b-Q8_0

List all available models

lemonade list

Atomic Chat
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

VIETRON 4B - Fine-tuned Vietnamese model

This is my first proper fine-tuned model, but still, the model may generate false informations or mistakes.

VieTron 4B is a Large Language Model (LLM) that has been extensively fine-tuned for Vietnamese users. With a 4-billion-parameter scale, VieTron is designed to be a smart, friendly AI assistant with a deep understanding of Vietnamese culture and education.

Details

Trained on high-quality Vietnamese datasets that cover most fields and topics.

More thoughtful: the model is trained with instruction to give response step by step (or CoT), the model will not only generate results but the reasoning steps behind the results.

More natural response style: the datasets also includes the natural Vietnamese conversation, making the model's response more "human".

Model info

~4 billion parameters

Currently I've only uploaded the initial version, quantized Q8_0 GGUF format to test the model. I will provide more quantized GGUF formats in the future as the model is getting better.

Usage

LM Studio recommended: the easiest way to run inference. Search ^{ngkhoi/vietron-4b} and download to use this model.

Limitations & Ethical Considerations

Knowledge Cutoff: VieTron's knowledge is limited to its training data. The model may not be aware of the latest events.

Hallucination Potential: Like all LLMs, VieTron can generate incorrect information. Please verify important facts.

Contributions

This project is developed solely by me so any contributions to this project are truly welcome!

Logs

26/10/2025: Continuation of fine-tuning.

Downloads last month: 44

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for ngkhoi/vietron-4b

Quantizations

2 models