Instructions to use RKNNAI/RK182X-VLM-UI-TARS-2B-SFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use RKNNAI/RK182X-VLM-UI-TARS-2B-SFT 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 RKNNAI/RK182X-VLM-UI-TARS-2B-SFT
# Run inference directly in the terminal:
llama cli -hf RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf RKNNAI/RK182X-VLM-UI-TARS-2B-SFT
# Run inference directly in the terminal:
llama cli -hf RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

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 RKNNAI/RK182X-VLM-UI-TARS-2B-SFT
# Run inference directly in the terminal:
./llama-cli -hf RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

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 RKNNAI/RK182X-VLM-UI-TARS-2B-SFT
# Run inference directly in the terminal:
./build/bin/llama-cli -hf RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

Use Docker

docker model run hf.co/RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

LM Studio
Jan

vLLM

How to use RKNNAI/RK182X-VLM-UI-TARS-2B-SFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RKNNAI/RK182X-VLM-UI-TARS-2B-SFT"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RKNNAI/RK182X-VLM-UI-TARS-2B-SFT",
		"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/RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

Ollama
How to use RKNNAI/RK182X-VLM-UI-TARS-2B-SFT with Ollama:
```
ollama run hf.co/RKNNAI/RK182X-VLM-UI-TARS-2B-SFT
```
Unsloth Desktop
Docker Model Runner
How to use RKNNAI/RK182X-VLM-UI-TARS-2B-SFT with Docker Model Runner:
```
docker model run hf.co/RKNNAI/RK182X-VLM-UI-TARS-2B-SFT
```

Lemonade

How to use RKNNAI/RK182X-VLM-UI-TARS-2B-SFT with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

Run and chat with the model

lemonade run user.RK182X-VLM-UI-TARS-2B-SFT-{{QUANT_TAG}}

List all available models

lemonade list

Atomic Chat

RK182X-VLM-UI-TARS-2B-SFT

English | 简体中文

1. Model Overview

This repository provides an RKNN model converted from ByteDance-Seed/UI-TARS-2B-SFT.

Model ID: RKNNAI/RK182X-VLM-UI-TARS-2B-SFT
Model Display Name: RK182X-VLM-UI-TARS-2B-SFT
RKNN Runtime Version: v1.1.0
Source Model: ByteDance-Seed/UI-TARS-2B-SFT
Model Type: VLM
Chip Label: RK182X
Supported Chips: RK1828

Available Models

RKNN Runtime Version	Configuration Directory	Supported Chips	Resolution	Quantization	NPU Cores	Context Length (tokens)	KVCache
`v1.1.0`	UI-TARS-2B-SFT-448x448-w4a16-8-27648	RK1828	Vision: 448x448	Vision: w4a16; LLM: w4a16	Vision: 8; LLM: 8	LLM: 27648 (27k)	LLM: fp16

2. Files

File or Directory	Description
`LICENSE`	Source model license
`NOTICE`	License, usage requirements, and attribution
`<configuration-directory>/`	Matching model files
Subdirectory `README.md`	Configuration documentation (English)
Subdirectory `README_CN.md`	Configuration documentation (Chinese)
Subdirectory `config.json`	Model configuration and file manifest
Subdirectory `SHA256SUMS`	SHA-256 checksums of the files in this directory (excluding this file)

3. Download

Download the full repository from ModelScope:

modelscope download --model RKNNAI/RK182X-VLM-UI-TARS-2B-SFT --revision v1.1.0 --local_dir ./RK182X-VLM-UI-TARS-2B-SFT

Download the full repository from Hugging Face:

hf download RKNNAI/RK182X-VLM-UI-TARS-2B-SFT --revision v1.1.0 --local-dir ./RK182X-VLM-UI-TARS-2B-SFT

Download a specific configuration from ModelScope:

from modelscope import snapshot_download

snapshot_download(
    "RKNNAI/RK182X-VLM-UI-TARS-2B-SFT",
    revision="v1.1.0",
    allow_patterns=["README.md", "README_CN.md", "LICENSE", "NOTICE", "UI-TARS-2B-SFT-448x448-w4a16-8-27648/**"],
    local_dir="./RK182X-VLM-UI-TARS-2B-SFT",
)

Download a specific configuration from Hugging Face:

hf download RKNNAI/RK182X-VLM-UI-TARS-2B-SFT --revision v1.1.0 --include "README.md" "README_CN.md" "LICENSE" "NOTICE" "UI-TARS-2B-SFT-448x448-w4a16-8-27648/**" --local-dir ./RK182X-VLM-UI-TARS-2B-SFT

4. SHA-256 Verification

Run in the configuration directory:

cd ./RK182X-VLM-UI-TARS-2B-SFT/UI-TARS-2B-SFT-448x448-w4a16-8-27648
sha256sum -c SHA256SUMS

Deploy only after all entries show OK.

5. Compatibility and Limitations

Supported Chips: RK1828.
Use the matching RKNN Runtime and driver. Check the device-side version with rknn-smi -v before running.
Use the files from the selected configuration directory together.

6. Copyright and License

Source model license: Apache License 2.0, See LICENSE for the full text and NOTICE for attribution and conversion modifications.
The artifacts in this repository are converted derivatives of the source weights, not unmodified upstream weights. They remain subject to the source model license and usage restrictions.
The applicable RKNN Toolkit and RKNN Runtime licenses also apply; runtime libraries are not included in this package.

Downloads last month: 182

GGUF

Model size

0 params

Architecture

qwen2vl

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Model tree for RKNNAI/RK182X-VLM-UI-TARS-2B-SFT

Base model

ByteDance-Seed/UI-TARS-2B-SFT

Quantized

(7)

this model