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

Libraries

How to use numind/NuExtract with Transformers:

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

pipe = pipeline("text-generation", model="numind/NuExtract", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("numind/NuExtract", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("numind/NuExtract", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use numind/NuExtract with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/numind/NuExtract

SGLang

How to use numind/NuExtract 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 "numind/NuExtract" \
    --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": "numind/NuExtract",
		"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 "numind/NuExtract" \
        --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": "numind/NuExtract",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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

how to improve accuracy and pdf size limit

#18

by sanamulani1108 - opened Sep 2, 2025

Discussion

sanamulani1108

Sep 2, 2025

•

edited Sep 2, 2025

I have a 135-page PDF of drawings. Each page contains 2–3 sections, and each section represents an individual drawing. I need to extract information from each drawing separately. When I tested with a 6-page pdf, all sections were detected correctly, but the extracted parameter values were often inaccurate. How can I improve the accuracy of the extraction?

What is the maximum number of pages allowed when uploading a PDF for processing in nuextract platform?

liamcripwell

Sep 15, 2025

Hello! You can improve the accuracy by either modifying your template to make it more descriptive, or add in-context examples to help show the model how your task works ("Example Set" tab on the platform).
The maximum input is ~30k tokens, so it depends on the resolution of your documents. You can increase/lower resolution under the "Settings" tab in the platform.

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