slim-extract-tiny-npu-ov

slim-extract-tiny-npu-ov

slim-extract-tiny-npu-ov is a specialized function calling model with a single mission to look for values in a text, based on an "extract" key that is passed as a parameter. No other instructions are required except to pass the context passage, and the target key, and the model will generate a python dictionary consisting of the extract key and a list of the values found in the text, including an 'empty list' if the text does not provide an answer for the value of the selected key.

This is an OpenVino int4 quantized version of slim-extract-tiny, providing a very fast, very small inference implementation, optimized for AI PCs using Intel NPU.

Model Description

• Developed by: llmware
• Model type: tinyllama
• Parameters: 1.1 billion
• Model Parent: llmware/slim-extract-tiny
• Language(s) (NLP): English
• License: Apache 2.0
• Uses: Extraction of values from complex business documents
• RAG Benchmark Accuracy Score: NA
• Quantization: int4

Example Usage

from llmware.models import ModelCatalog

text_passage = "The company announced that for the current quarter the total revenue increased by 9% to $125 million."  
model = ModelCatalog().load_model("slim-extract-tiny-npu-ov")   
llm_response = model.function_call(text_passage, function="extract", params=["revenue"])  

Output: `llm_response = {"revenue": [$125 million"]}`  

Model Card Contact

llmware on github

llmware on hf

llmware website