Instructions to use numind/NuExtract-2.0-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use numind/NuExtract-2.0-8B with Transformers:

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

pipe = pipeline("image-text-to-text", model="numind/NuExtract-2.0-8B")
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("numind/NuExtract-2.0-8B")
model = AutoModelForImageTextToText.from_pretrained("numind/NuExtract-2.0-8B")
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

vLLM

How to use numind/NuExtract-2.0-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "numind/NuExtract-2.0-8B"
# 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-2.0-8B",
		"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/numind/NuExtract-2.0-8B

SGLang

How to use numind/NuExtract-2.0-8B 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-2.0-8B" \
    --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-2.0-8B",
		"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 "numind/NuExtract-2.0-8B" \
        --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-2.0-8B",
		"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 numind/NuExtract-2.0-8B with Docker Model Runner:
```
docker model run hf.co/numind/NuExtract-2.0-8B
```

fix: remove duplicated chat_template and trailing comma

by Restodecoca - opened Nov 13, 2025

base: refs/heads/main

←

from: refs/pr/6

Discussion Files changed

-1

Files changed (1) hide show

tokenizer_config.json +1 -1

tokenizer_config.json CHANGED Viewed

@@ -196,7 +196,7 @@
     "<|video_pad|>"
   ],
   "bos_token": null,
-  "chat_template":   "chat_template": "{%- set image_placeholder = '<|vision_start|><|image_pad|><|vision_end|>' -%}\n{%- for message in messages -%}\n    {%- if message['role'] == 'system' -%}\n        {{- '<|im_start|>system\n' -}}\n        {%- if message['content'] is string -%}\n            {{- message['content'] | trim -}}\n        {%- endif -%}\n        {{- '<|im_end|>\n' -}}\n    {%- elif message['role'] == 'user' -%}\n        {%- if loop.first -%}\n            {{- '<|im_start|>system\n' -}}\n            {%- if template -%}\n                {#--- If template, extraction task ---#}\n                {{- 'You are NuExtract, an information extraction tool created by NuMind.' -}}\n            {%- else -%}\n                {#--- Else, template generation task ---#}\n                {{- 'You are a helpful assistant.' -}}\n            {%- endif -%}\n            {{ '<|im_end|>\n' }}\n        {%- endif -%}\n        {{- '<|im_start|>' + message['role'] + '\n' -}}\n        {%- if template -%}\n            {#--- Template Section ---#}\n            {{- '# Template:\n' -}}\n            {{- template -}}\n            {{- '\n' -}}\n\n            {%- if examples -%}\n                {#--- Examples can only exist in the extraction task ---#}\n                {{- '# Examples:\n' -}}\n                {%- for example in examples -%}\n                    {{- '## Input:\n' -}}\n                    {%- if example['input'] is mapping and (example['input']['type'] == 'image' or example['input']['type'] == 'image_url') -%}\n                        {{- image_placeholder | trim -}}\n                    {%- elif example['input'] == '<image>' -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {{- image_placeholder | trim -}}\n                    {%- else -%}\n                        {#--- Text input example ---#}\n                        {{- example['input'] -}}\n                    {%- endif -%}\n                    {{- '\n' -}}\n                    {{- '## Output:\n' -}}\n                    {{- example['output'] -}}\n                    {{- '\n' -}}\n                {%- endfor -%}\n            {%- endif -%}\n            {{- '# Context:\n' -}}\n        {%- endif -%}\n\n        {%- if message['content'] is string -%}\n            {#--- Simple string content ---#}\n            {{- message['content'] | trim -}}\n        {%- elif message['content'] is mapping and (message['content']['type'] == 'image' or message['content']['type'] == 'image_url') -%}\n            {{- image_placeholder | trim -}}\n        {%- else -%}\n            {#--- List of content items (mixed text/images) ---#}\n            {#--- First, determine what the actual input content is (not ICL images) ---#}\n            {%- set ns = namespace(has_text_input=false, text_content='') -%}\n\n            {#--- Count content types and identify actual input document ---#}\n            {%- for content in message['content'] -%}\n                {%- if content is mapping and content.get('type') == 'text' -%}\n                    {%- if content.get('text') != '<image>' -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {%- set ns.has_text_input = true -%}\n                        {%- set ns.text_content = content['text'] -%}\n                    {%- endif -%}\n                {%- elif content is string -%}\n                    {%- if content != '<image>' -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {%- set ns.has_text_input = true -%}\n                        {%- set ns.text_content = content -%}\n                    {%- endif -%}\n                {%- endif -%}\n            {%- endfor -%}\n\n            {#--- Determine what to output based on actual input type ---#}\n            {%- if ns.has_text_input -%}\n                {#--- Main input is text, so output the text content ---#}\n                {{- ns.text_content | trim -}}\n            {%- else -%}\n                {#--- Main input is image or <image> placeholder ---#}\n                {%- set ns2 = namespace(found_image=false) -%}\n                {%- for content in message['content'] -%}\n                    {%- if content is mapping and (content.get('type') == 'image' or content.get('type') == 'image_url') and not ns2.found_image -%}\n                        {{- image_placeholder | trim -}}\n                        {%- set ns2.found_image = true -%}\n                    {%- elif content is mapping and content.get('type') == 'text' and content.get('text') == '<image>' and not ns2.found_image -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {{- image_placeholder | trim -}}\n                        {%- set ns2.found_image = true -%}\n                    {%- elif content is string and content == '<image>' and not ns2.found_image -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {{- image_placeholder | trim -}}\n                        {%- set ns2.found_image = true -%}\n                    {%- endif -%}\n                {%- endfor -%}\n            {%- endif -%}\n        {%- endif -%}\n        {{- '<|im_end|>\n'}}\n\n    {%- elif message['role'] == 'assistant' -%}\n        {{- '<|im_start|>assistant\n' -}}\n        {%- if message['content'] is string -%}\n            {{- message['content'] | trim -}}\n        {%- elif message['content'] is iterable and message['content'] is not string -%}\n            {%- for content in message['content'] -%}\n                {%- if content is mapping and content.get('type') == 'text' -%}\n                    {{- content['text'] | trim -}}\n                {%- elif content is string -%}\n                    {{- content | trim -}}\n                {%- endif -%}\n            {%- endfor -%}\n        {%- endif -%}\n        {{- '<|im_end|>\n' -}}\n    {%- endif -%}\n{%- endfor -%}\n{%- if add_generation_prompt -%}\n    {{- '<|im_start|>assistant\n' -}}\n{%- endif -%}",,
   "clean_up_tokenization_spaces": false,
   "eos_token": "<|im_end|>",
   "errors": "replace",

     "<|video_pad|>"
   ],
   "bos_token": null,
+  "chat_template": "{%- set image_placeholder = '<|vision_start|><|image_pad|><|vision_end|>' -%}\n{%- for message in messages -%}\n    {%- if message['role'] == 'system' -%}\n        {{- '<|im_start|>system\n' -}}\n        {%- if message['content'] is string -%}\n            {{- message['content'] | trim -}}\n        {%- endif -%}\n        {{- '<|im_end|>\n' -}}\n    {%- elif message['role'] == 'user' -%}\n        {%- if loop.first -%}\n            {{- '<|im_start|>system\n' -}}\n            {%- if template -%}\n                {#--- If template, extraction task ---#}\n                {{- 'You are NuExtract, an information extraction tool created by NuMind.' -}}\n            {%- else -%}\n                {#--- Else, template generation task ---#}\n                {{- 'You are a helpful assistant.' -}}\n            {%- endif -%}\n            {{ '<|im_end|>\n' }}\n        {%- endif -%}\n        {{- '<|im_start|>' + message['role'] + '\n' -}}\n        {%- if template -%}\n            {#--- Template Section ---#}\n            {{- '# Template:\n' -}}\n            {{- template -}}\n            {{- '\n' -}}\n\n            {%- if examples -%}\n                {#--- Examples can only exist in the extraction task ---#}\n                {{- '# Examples:\n' -}}\n                {%- for example in examples -%}\n                    {{- '## Input:\n' -}}\n                    {%- if example['input'] is mapping and (example['input']['type'] == 'image' or example['input']['type'] == 'image_url') -%}\n                        {{- image_placeholder | trim -}}\n                    {%- elif example['input'] == '<image>' -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {{- image_placeholder | trim -}}\n                    {%- else -%}\n                        {#--- Text input example ---#}\n                        {{- example['input'] -}}\n                    {%- endif -%}\n                    {{- '\n' -}}\n                    {{- '## Output:\n' -}}\n                    {{- example['output'] -}}\n                    {{- '\n' -}}\n                {%- endfor -%}\n            {%- endif -%}\n            {{- '# Context:\n' -}}\n        {%- endif -%}\n\n        {%- if message['content'] is string -%}\n            {#--- Simple string content ---#}\n            {{- message['content'] | trim -}}\n        {%- elif message['content'] is mapping and (message['content']['type'] == 'image' or message['content']['type'] == 'image_url') -%}\n            {{- image_placeholder | trim -}}\n        {%- else -%}\n            {#--- List of content items (mixed text/images) ---#}\n            {#--- First, determine what the actual input content is (not ICL images) ---#}\n            {%- set ns = namespace(has_text_input=false, text_content='') -%}\n\n            {#--- Count content types and identify actual input document ---#}\n            {%- for content in message['content'] -%}\n                {%- if content is mapping and content.get('type') == 'text' -%}\n                    {%- if content.get('text') != '<image>' -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {%- set ns.has_text_input = true -%}\n                        {%- set ns.text_content = content['text'] -%}\n                    {%- endif -%}\n                {%- elif content is string -%}\n                    {%- if content != '<image>' -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {%- set ns.has_text_input = true -%}\n                        {%- set ns.text_content = content -%}\n                    {%- endif -%}\n                {%- endif -%}\n            {%- endfor -%}\n\n            {#--- Determine what to output based on actual input type ---#}\n            {%- if ns.has_text_input -%}\n                {#--- Main input is text, so output the text content ---#}\n                {{- ns.text_content | trim -}}\n            {%- else -%}\n                {#--- Main input is image or <image> placeholder ---#}\n                {%- set ns2 = namespace(found_image=false) -%}\n                {%- for content in message['content'] -%}\n                    {%- if content is mapping and (content.get('type') == 'image' or content.get('type') == 'image_url') and not ns2.found_image -%}\n                        {{- image_placeholder | trim -}}\n                        {%- set ns2.found_image = true -%}\n                    {%- elif content is mapping and content.get('type') == 'text' and content.get('text') == '<image>' and not ns2.found_image -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {{- image_placeholder | trim -}}\n                        {%- set ns2.found_image = true -%}\n                    {%- elif content is string and content == '<image>' and not ns2.found_image -%}\n                        {#--- Keep compatibility with <image> for now ---#}\n                        {{- image_placeholder | trim -}}\n                        {%- set ns2.found_image = true -%}\n                    {%- endif -%}\n                {%- endfor -%}\n            {%- endif -%}\n        {%- endif -%}\n        {{- '<|im_end|>\n'}}\n\n    {%- elif message['role'] == 'assistant' -%}\n        {{- '<|im_start|>assistant\n' -}}\n        {%- if message['content'] is string -%}\n            {{- message['content'] | trim -}}\n        {%- elif message['content'] is iterable and message['content'] is not string -%}\n            {%- for content in message['content'] -%}\n                {%- if content is mapping and content.get('type') == 'text' -%}\n                    {{- content['text'] | trim -}}\n                {%- elif content is string -%}\n                    {{- content | trim -}}\n                {%- endif -%}\n            {%- endfor -%}\n        {%- endif -%}\n        {{- '<|im_end|>\n' -}}\n    {%- endif -%}\n{%- endfor -%}\n{%- if add_generation_prompt -%}\n    {{- '<|im_start|>assistant\n' -}}\n{%- endif -%}",
   "clean_up_tokenization_spaces": false,
   "eos_token": "<|im_end|>",
   "errors": "replace",