README.md

---

license: mit
language:

- multilingual

tags:

- nlp

base_model: microsoft/Phi-3.5-mini-instruct

---

# NuExtract-v1.5 by NuMind 🔥

NuExtract is a fine-tuning of [Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct), trained on a private high-quality dataset for structured information extraction. It supports long documents and several languages (English, French, Spanish, German, Portuguese, and Italian).
To use the model, provide an input text and a JSON template describing the information you need to extract.

Note: This model is trained to prioritize pure extraction, so in most cases all text generated by the model is present as is in the original text.

Try it here: <space-link>

We also provide a tiny (0.5B) version which is based on Qwen2.5-0.5B: [NuExtract-tiny-v1.5](https://huggingface.co/numind/NuExtract-tiny-v1.5)

**Checkout other models by NuMind:**

- NuExtract Version 1 Models: [0.5B](https://huggingface.co/numind/NuExtract-tiny), [3.8B](https://huggingface.co/numind/NuExtract), [7B](https://huggingface.co/numind/NuExtract-large)
- SOTA Zero-shot NER Model [NuNER Zero](https://huggingface.co/numind/NuNER_Zero)
- SOTA Multilingual Entity Recognition Foundation Model: [link](https://huggingface.co/numind/entity-recognition-multilingual-general-sota-v1)
- SOTA Sentiment Analysis Foundation Model: [English](https://huggingface.co/numind/generic-sentiment-v1), [Multilingual](https://huggingface.co/numind/generic-sentiment-multi-v1)

## Benchmark

Include result plots.

## Usage

To use the model:

```python
import json
from transformers import AutoModelForCausalLM, AutoTokenizer

def predict_NuExtract(model, tokenizer, texts, template, batch_size=1, max_length=10_000, max_new_tokens=4_000):
    template = json.dumps(json.loads(template), indent=4)
    prompts = [f"""<|input|>\n### Template:\n{template}\n### Text:\n{text}\n\n<|output|>""" for text in texts]
    
    outputs = []
    with torch.no_grad():
		    for i in range(0, len(prompts), batch_size):
				    batch_prompts = prompts[i:i+batch_size]
				    batch_encodings = tokenizer(batch_prompts, return_tensors="pt", truncation=True, padding=True, max_length=max_length).to(model.device)
    
				    pred_ids = model.generate(**batch_encodings, max_new_tokens=max_new_tokens)
				    outputs += tokenizer.batch_decode(pred_ids, skip_special_tokens=True)

    return [output.split("<|output|>")[1] for output in outputs]

model_name = "numind/NuExtract-v1.5"
device = "cuda"
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, trust_remote_code=True).to(device).eval()
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)

text = """We introduce Mistral 7B, a 7–billion-parameter language model engineered for
superior performance and efficiency. Mistral 7B outperforms the best open 13B
model (Llama 2) across all evaluated benchmarks, and the best released 34B
model (Llama 1) in reasoning, mathematics, and code generation. Our model
leverages grouped-query attention (GQA) for faster inference, coupled with sliding
window attention (SWA) to effectively handle sequences of arbitrary length with a
reduced inference cost. We also provide a model fine-tuned to follow instructions,
Mistral 7B – Instruct, that surpasses Llama 2 13B – chat model both on human and
automated benchmarks. Our models are released under the Apache 2.0 license.
Code: <https://github.com/mistralai/mistral-src>
Webpage: <https://mistral.ai/news/announcing-mistral-7b/>"""

template = """{
    "Model": {
        "Name": "",
        "Number of parameters": "",
        "Number of max token": "",
        "Architecture": []
    },
    "Usage": {
        "Use case": [],
        "Licence": ""
    }
}"""

prediction = predict_NuExtract(model, tokenizer, [text], template)[0]
print(prediction)

```