Instructions to use TengfeiLiuCoder/RefCaptioner with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TengfeiLiuCoder/RefCaptioner with Transformers:

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

pipe = pipeline("image-text-to-text", model="TengfeiLiuCoder/RefCaptioner")
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("TengfeiLiuCoder/RefCaptioner")
model = AutoModelForMultimodalLM.from_pretrained("TengfeiLiuCoder/RefCaptioner", 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=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use TengfeiLiuCoder/RefCaptioner with vLLM:

Install from pip and serve model

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

SGLang

How to use TengfeiLiuCoder/RefCaptioner 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 "TengfeiLiuCoder/RefCaptioner" \
    --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": "TengfeiLiuCoder/RefCaptioner",
		"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 "TengfeiLiuCoder/RefCaptioner" \
        --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": "TengfeiLiuCoder/RefCaptioner",
		"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"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use TengfeiLiuCoder/RefCaptioner with Docker Model Runner:
```
docker model run hf.co/TengfeiLiuCoder/RefCaptioner
```

TengfeiLiuCoder commited on 3 days ago

Commit

f85622f

verified ·

1 Parent(s): 403ebeb

Add task and method overview figures

Browse files

Files changed (4) hide show

.gitattributes +2 -0
README.md +11 -2
framework.png +3 -0
teaser.png +3 -0

.gitattributes CHANGED Viewed

@@ -6,3 +6,5 @@
 *.gguf filter=lfs diff=lfs merge=lfs -text
 captioncase_01.png filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text

 *.gguf filter=lfs diff=lfs merge=lfs -text
 captioncase_01.png filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text
+teaser.png filter=lfs diff=lfs merge=lfs -text
+framework.png filter=lfs diff=lfs merge=lfs -text

README.md CHANGED Viewed

@@ -20,6 +20,12 @@ RefCaptioner is an 8B vision-language model for **multi-reference image-grounded
 The released weights are a merged Hugging Face checkpoint based on [Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct). No PEFT adapter is required at inference time.
 ## Model details
 | Item | Value |
@@ -31,6 +37,11 @@ The released weights are a merged Hugging Face checkpoint based on [Qwen3-VL-8B-
 | Thinking mode | Disabled |
 | Weight format | Merged `safetensors` |
 ## Installation
 The tested environment for loading RefCaptioner and running inference is provided in `requirement.txt`. Use Python 3.10 and install it with:
@@ -39,9 +50,7 @@ The tested environment for loading RefCaptioner and running inference is provide
 pip install -r requirement.txt
 ```
-RefCaptioner uses the standard Qwen3-VL inference stack and does not require a task-specific runtime. Any environment that can run Qwen3-VL with a compatible Transformers release can be used directly.
-The benchmark environment used `torch==2.9.0`, `transformers==4.57.6`, `accelerate==1.14.0`, `qwen-vl-utils==0.0.14`, and `av==17.1.0`.
 ## Quick start

 The released weights are a merged Hugging Face checkpoint based on [Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct). No PEFT adapter is required at inference time.
+## Task overview
+![Multi-reference image-grounded video captioning](teaser.png)
+RefCaptioner grounds relevant reference images to local phrases in a video caption while rejecting distractor references.
 ## Model details
 | Item | Value |
 | Thinking mode | Disabled |
 | Weight format | Merged `safetensors` |
+## Method overview
+![RefCaptioner training framework](framework.png)
+Training Framework
 ## Installation
 The tested environment for loading RefCaptioner and running inference is provided in `requirement.txt`. Use Python 3.10 and install it with:
 pip install -r requirement.txt
 ```
 ## Quick start

framework.png ADDED Viewed

Git LFS Details

SHA256: 0ea48392e6c91f713035ff1904c3705d971b83f7b281de1df800a3330dc625bb
Pointer size: 131 Bytes
Size of remote file: 580 kB

teaser.png ADDED Viewed

Git LFS Details

SHA256: b4157be5baf79a750e800352ca13dfd8e72468f9576b188636786352741618d1
Pointer size: 132 Bytes
Size of remote file: 1.96 MB