Instructions to use lvyufeng/PaddleOCR-VL-0.9B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lvyufeng/PaddleOCR-VL-0.9B with Transformers:

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

pipe = pipeline("image-text-to-text", model="lvyufeng/PaddleOCR-VL-0.9B")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("lvyufeng/PaddleOCR-VL-0.9B", dtype="auto")

MindSpore

How to use lvyufeng/PaddleOCR-VL-0.9B with MindSpore:

# No code snippets available yet for this library.

# To use this model, check the repository files and the library's documentation.

# Want to help? PRs adding snippets are welcome at:
# https://github.com/huggingface/huggingface.js

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use lvyufeng/PaddleOCR-VL-0.9B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lvyufeng/PaddleOCR-VL-0.9B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lvyufeng/PaddleOCR-VL-0.9B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/lvyufeng/PaddleOCR-VL-0.9B

SGLang

How to use lvyufeng/PaddleOCR-VL-0.9B 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 "lvyufeng/PaddleOCR-VL-0.9B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lvyufeng/PaddleOCR-VL-0.9B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "lvyufeng/PaddleOCR-VL-0.9B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lvyufeng/PaddleOCR-VL-0.9B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use lvyufeng/PaddleOCR-VL-0.9B with Docker Model Runner:
```
docker model run hf.co/lvyufeng/PaddleOCR-VL-0.9B
```

prompt

by tadkt - opened Oct 20, 2025

Discussion

tadkt

Oct 20, 2025

Orginal paper said that there are 4 prompt for 4 different objects

lvyufeng

Owner Oct 20, 2025

sure, you can use the prompts below：

query = "OCR:"
query = "Table Recognition:"
query = "Chart Recognition:"
query = "Formula Recognition:"

tadkt

Oct 21, 2025

For table recognition I use the prompt "Table Recognition:" but it does not output OTSL style like in the paper noticed. Is this the problem with the checkpoint/prompt or the paper?

lvyufeng

Owner Oct 21, 2025

For table recognition I use the prompt "Table Recognition:" but it does not output OTSL style like in the paper noticed. Is this the problem with the checkpoint/prompt or the paper?

it seems the preprocess should be exactly same as PaddleX, i will analyze the code when i have free time

tadkt

Oct 22, 2025

•

edited Oct 22, 2025

Hi, the prompt "OTSL Table Recognition:" works better for me

OTSL Table Recognition:

<fcel>Millions of dollars and shares except per share data<fcel>Year Ended December 31<lcel><lcel><nl><ucel><fcel>2008<fcel>2007<fcel>2006<nl><fcel>Revenue:<ecel><ecel><ecel><nl><fcel>Services<fcel>$ 13,391<fcel>$ 11,256<fcel>$ 9,643<nl><fcel>Product sales<fcel>4,888<fcel>4,008<fcel>3,312<nl><fcel>Total revenue<fcel>18,279<fcel>15,264<fcel>12,955<nl><fcel>Operating costs and expenses:<ecel><ecel><ecel><nl><fcel>Cost of services<fcel>10,079<fcel>8,167<fcel>6,751<nl><fcel>Cost of sales<fcel>3,970<fcel>3,358<fcel>2,675<nl><fcel>Revenue and administrative<fcel>282<fcel>293<fcel>342<nl><fcel>Gain on sale of business assets, net<fcel>(62)<fcel>(52)<fcel>(58)<nl><fcel>Total operating costs and expenses<fcel>14,269<fcel>11,766<fcel>9,710<nl><fcel>Operating income<fcel>4,010<fcel>3,498<fcel>3,245<nl><fcel>Interest expense<fcel>(160)<fcel>(154)<fcel>(165)<nl><fcel>Interest income<fcel>39<fcel>124<fcel>129<nl><fcel>Other, net<fcel>(726)<fcel>(8)<fcel>(10)<nl><fcel>Income from continuing operations before income taxes and minority interest provision for income taxes<fcel>3,163<fcel>3,460<fcel>3,199<nl><fcel>Minority interest in net income of subsidiaries<fcel>(1,211)<fcel>(907)<fcel>(1,003)<nl><fcel>Income from continuing operations<fcel>9<fcel>(29)<fcel>(19)<nl><fcel>Income from continuing operations, net of income tax (provision) benefit of $3, $15), and $ (183)<fcel>1,961<fcel>2,524<fcel>2,177<nl><fcel>Net income<fcel>(423)<fcel>975<fcel>171<nl><fcel>Net income<fcel>$ 1,538<fcel>$ 3,499<fcel>$ 2,348<nl><fcel>Basic income (loss) per share:<ecel><ecel><ecel><nl><fcel>Income from continuing operations<fcel>$ 2,24<fcel>$ 2,76<fcel>$ 2,15<nl><fcel>Income (loss) from discontinued operations, net<fcel>(0.49)<fcel>1.07<fcel>0.16<nl><fcel>Net income per share<fcel>$ 1.75<fcel>$ 3.83<fcel>$ 2.31<nl><fcel>Diluted income (loss) per share:<ecel><ecel><ecel><nl><fcel>Income from continuing operations<fcel>$ 2.17<fcel>$ 2.66<fcel>$ 2.07<nl><fcel>Income (loss) from discontinued operations, net<fcel>(0.47)<fcel>1.02<fcel>0.16<nl><fcel>Net income per share<fcel>$ 1.70<fcel>$ 3.68<fcel>$ 2.23<nl><fcel>Basic weighted average common shares outstanding<fcel>877<fcel>913<fcel>1,014<nl><fcel>Diluted weighted average common shares outstanding<fcel>904<fcel>950<fcel>1,054<nl>

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment